# Woody Wu, full text

> Every essay published on woodywusommelier.com, in full, newest first. The index and the
> commissionable work are at https://woodywusommelier.com/llms.txt

Generated 2026-09-21. 34 essays.
Contact: hello@woodywusommelier.com

Attribution: quote freely with a link to the essay URL given above each piece.

---
# Six Figures Is Not the Problem

URL: https://woodywusommelier.com/writing/six-figures-is-not-the-problem/
Pillar: The Floor
Published: 2026-09-17

The Federal Reserve raised rates yesterday. A quarter point, to a target range of three and three quarters to four percent, the first increase since 2023, and the vote was unanimous. Sixteen of the eighteen people in the room expect another one.

This morning the national average for a gallon of regular gasoline is four dollars and forty four cents, up sixteen cents in a week and more than a dollar higher than it was a year ago. Crude is back above a hundred dollars a barrel.

And the thing everyone is arguing about online is whether a household earning a hundred and forty thousand dollars a year is poor.

I want to take that argument seriously, because I think both sides of it are wrong in an interesting way, and because I spend my working nights standing between the two groups of people it is actually about.

## The number, and what is wrong with it

The claim comes from an essay by Michael W. Green, published in November of last year and still going around. Green is a chief strategist and portfolio manager at an asset manager, which matters, because this is not a stray blog post. It is a professional making an argument in his own discipline.

His argument is that the official poverty threshold, about thirty one thousand two hundred dollars for a family of four, no longer describes anything, and that a real basic needs budget for that family comes to somewhere between a hundred and thirty six and a hundred and forty thousand.

The budget itself is not invented. Childcare near thirty three thousand. Housing near twenty three. Food near fifteen. Transportation near fifteen. Health care near eleven. Then everything else, then taxes on top of all of it. Anyone who has assembled that list for their own household recognises every line.

The method underneath it does not survive contact.

The original poverty line, built by Mollie Orshansky in the sixties, worked by taking what a family spent on food and multiplying by three, because food was about a third of a household budget. Green keeps her logic and updates the input: food is now five to seven percent of spending, so the multiplier, in his words, is no longer three. It becomes sixteen.

Scott Winship pointed out what that does if you run it backwards, and it is the most useful paragraph anyone has written about this. In 1901, food was forty two and a half percent of household spending. Apply the same method and the 1901 poverty line comes out around twenty four thousand dollars in today's money. So the method says Americans were less poor in 1901, when nearly half of everything they earned went into their mouths, than they are now.

Food stopped being a third of the budget because food got cheap relative to everything a family can now also buy. The small share is the good news. Green's method reads the good news as the catastrophe.

So the number is wrong. I think the number is wrong and the feeling underneath it is real, and the interesting question is what the feeling is actually measuring, because it is not measuring poverty.

## What it is measuring

Sixty four percent of six figure earners told a pollster that their income is not a milestone, it is the bare minimum for staying afloat. Forty four percent of American households say they have less than two hundred dollars in savings. You now need an income of roughly a hundred and ten thousand dollars to buy the median American home without spending more than a third of it on housing.

None of that is poverty. All of it is the same complaint, and the complaint is not about the level of the income. It is about the conversion rate.

Working harder used to convert into something. You put in more, and at the end of it there was a thing you had that you did not have before, and the thing stayed. That conversion has gotten very bad, and people who are not poor by any historical measure can feel it going bad, and when they reach for language to describe it the only available word is poor.

It is the wrong word. But they are describing something true, and they are describing, in a much gentler form, the exact thing that has always defined the group of people they are not in.

## A shift does not keep

Here is what my trade taught me about this, and I have never seen it put the way it actually works.

Wine keeps. That is close to the whole reason the business exists in the form it does. A bottle sits in the dark and does nothing and gets more valuable, and the person who owns the cellar earns while asleep. Everything about the trade, the allocations, the storage, the provenance, the auctions, is built on the fact that the asset persists.

A shift does not keep.

I have written before about selling three hours of attention, and the part of that I keep coming back to is how completely perishable it is. A server works a double. The hours convert to money once. The money converts to rent and groceries and gas at four forty four a gallon. And then it is gone, entirely, and the next night begins at exactly zero, with nothing accumulated, no position built, nothing that earns while they sleep.

That is not a metaphor for being locked out. That is the mechanism of being locked out, and it is the only definition of the underclass that has ever made sense to me. It is not about the size of the number. It is about whether anything is left over after the number is spent, and whether the thing that is left over does any work on its own.

By that definition the household at a hundred and forty thousand is not in it. They are close enough to the edge to be able to see it, which is new and which is why they will not stop talking about it, but they are not in it.

The people in it are carrying the plates.

## The part of this that nobody is writing about

The median American waiter or waitress makes about thirty three thousand seven hundred dollars. Cooks make about thirty five. Bartenders about the same as the servers.

And the one part of that income that was supposed to be the upside has been quietly collapsing. In 2021, fifty six percent of diners typically tipped twenty percent. By 2024 it was thirty four percent. A third of the tipping public simply stopped doing the thing the entire compensation structure of American restaurants assumes they will do.

That happened during the same years everyone was learning to say cost of living crisis. Guests under pressure economise at the last line on the bill, which is the line that is somebody's wage. The squeeze the six figure household is writing essays about gets passed directly down, in cash, nightly, to the person who was already at the bottom of it.

Nobody has written a viral post about that. There is no argument online about whether thirty three thousand seven hundred is enough, because there is nothing to argue about, and arguments require someone to be on the other side.

I am not saying this to score a point on the six figure household. I am saying it because the discourse has the flashlight pointed at the wrong end of the room, and the people at the dark end are the ones I work beside.

## I am not reporting this from a distance

I write the schedule. I am the person who decides how many hours exist and who gets them, so nothing in this section is observed from across a street.

Start with the piece of arithmetic that almost nobody outside the trade knows, because it explains why a soft tipping year is not a soft year for a server.

In most rooms, a server does not tip out a share of what they collected. They tip out a percentage of what they sold. Somewhere around three to five percent of their sales goes to the bar, the bussers, the runners and often the host, and that number is calculated off the check total, not off the tip.

Which means the tip out does not move when the guest tips less. It is the same either way.

So when a table of four runs a large check and leaves twelve percent, the server does not earn a little less that night. The server subsidises that table. They pay out on the sale they made and keep whatever is left, and on a bad enough section the left is close to nothing. Toast's own data has the average full service tip at nineteen point four percent in the second quarter of this year, the lowest it has been on their platform since the start of the pandemic. The share of guests who tip at all in full service has gone from roughly seventy three percent a few years ago to about sixty five.

Every fraction of a point in that decline lands on one person, and it is not the owner and it is not me.

Here is what that looks like in a building.

People used to ask to be cut early. The good shifts were the ones where you got out at nine. Now they ask to be cut last, and they ask before service, and they ask in a particular tone that I have learned to hear coming. The person who wants the extra hour is not being ambitious. They are doing a calculation.

Staff meal matters more than it used to. Not as a perk. As the meal.

And the commute has become its own line item. Somebody drives in for a shift, and gasoline is up more than a dollar on the year, and there are nights when a slow Tuesday does not clear what it cost to show up and stand there.

I can control a small number of these things and I control them. I can build a schedule that does not strand people on dead shifts. I can be honest about what a section is likely to earn before someone accepts it. I can refuse to let a bad tipping night quietly become the server's problem where the structure gives me any room at all.

I cannot fix the conversion rate, and I am not going to write a paragraph pretending that good management does. The most useful thing I can say is that I watch the gap between effort and outcome widen on people who are doing everything right, in person, four or five nights a week, and it does not look anything like an argument on the internet.

## So what actually gets you out

There is a phrase for the thing everyone is circling here, and it is permanent underclass.

Jasmine Sun spent months reporting a piece for the New York Times on how seriously the people building AI take that possibility, and the answer is very seriously. Her conclusion is the one I would want people to take away from her work rather than mine: that the production of a social underclass is a policy choice, not a weather event, and that treating it as inevitable is the mistake.

I want to be clear that I agree with her, and that everything I am about to say is a smaller argument than hers. Policy operates on whether this happens to a country. What follows operates on what a single person standing in a dining room can do about their own position, which is a worse instrument and the only one any of us actually holds.

There is also a harder version circulating, which says AI closes the door permanently: capital owners at the top, a service class and an unemployed class underneath, and a window of one to three decades in which a person can still accumulate something before wages stop mattering. Its prescription is to own assets now.

I think that one is roughly right about the mechanism and roughly wrong about the timing and the prescription, and I want to put my own version next to it, because it is the only part of this where my day job and my night job say the same thing.

AI does not pay anyone. It is not income and it is not a job and anyone selling it as either is selling something.

What it changed is narrower and, I think, much more important. It collapsed the price of the door.

Building a thing that keeps earning after you stop working on it used to require capital that a person on a floor wage could not assemble. A team, a build, a lawyer, a runway. The barrier was never the idea. Every good operator I know has three of those. The barrier was that turning one into a thing that exists cost more than a year of their pay.

I have gone through the actual labor math of this elsewhere and I will not repeat it here. The short version is that I built and deployed a working product on my own list and my own menu, while still learning to write the code, for a cost that is not a payroll and not a runway and not a year of anyone's wage.

That is not a fortune and it did not make me rich. What it did was produce the first thing I have ever owned that does not perish at the end of a shift.

And here is the part that should matter to anybody reading this from a dining room. The raw material for that is the thing this entire discourse treats as worthless. Judgment about people. A method that works. Knowing which table wants what and why, and being right about it on a Friday. The laid off junior analyst is in a harder position than the career server is, because the analyst's input was already the part the machine does. The server's input is the part it still cannot.

## The caveat, which is the whole integrity of this

Most people will not do this, and that is not a character failure, and I am not going to pretend otherwise.

Building something that keeps requires surplus. Not money necessarily, but time and attention and enough slack in your life that one bad month does not end the attempt. A person working doubles with under two hundred dollars in savings, with gas up a dollar on the year and a rate hike yesterday that makes every balance they carry more expensive, does not have that surplus. Telling them that AI is their way out while they are in that position is the same move as telling them a hundred and forty thousand dollars is poverty. It is a sentence written by someone standing somewhere else.

What I will say is narrower and I think it is true.

The reason the door is worth talking about now is that it is cheap now. Cheap doors do not stay cheap. They get crowded, and then they get priced, and the people who went through early are the ones the price is charged to everyone else on behalf of.

The conversion rate from work into something that lasts has been getting worse for a long time, and it did not start with any of this. For most of my working life, in the trade I chose, the honest answer to the question of how a person on a floor wage builds anything that keeps was that they mostly do not.

That answer changed recently, and not by much, and not for everyone, and possibly not for long.

But it changed, which is more than it has done in the whole time I have been standing in a dining room watching people carry plates, and I am not going to be casual about the difference.

---

# The Habit Is What Died

URL: https://woodywusommelier.com/writing/the-habit-is-what-died/
Pillar: The Cellar
Published: 2026-09-16

A few years ago my entire trade was told it was finished.

Drinking was at its lowest rate in ninety years of measurement. Every month produced another piece explaining that the young had killed alcohol the way they had killed napkins and doorbells, and that the sommelier was a job with an expiry date printed on it.

So I went and read the numbers, because it was my rent.

They did not say what everyone was repeating. Gen Z participation was up, from sixty-six percent to seventy-four, which is nearly the adult average. The generation drinking least was the Boomers, at seventy-one and falling. More young people were drinking, not fewer.

One number had collapsed, and it was not participation.

Daily drinking among the young went from six percent to two percent.

Same people. Same category. A third of the frequency.

## That is a different event than death

Nothing was dying. Something was changing state, from a habit into an occasion, and those are not the same object even when they involve the same liquid.

A habit is unexamined. It happens on a Tuesday because it is Tuesday. It does not get chosen, it does not get discussed, and nobody photographs it.

An occasion is selected. It has a reason, it has an audience even when the audience is one other person, and it carries a small freight of meaning that a Tuesday glass never had to carry.

When a category makes that move, its economics invert. Volume falls and stakes rise. Fewer instances, each one asked to mean more.

I have now spent several years living in the after, and I can tell you what it does.

## What it did to my job

Volume was never justifying my position. Volume was subsidizing it.

The daily drinker was the customer who made a wine program arithmetically survivable. Not the connoisseur. The person who had a glass with dinner without thinking about it, several times a week, forever. That person paid for the cellar, the storage, the breakage, and the salary of somebody to look after all of it.

When that person mostly stopped, the subsidy went. What remained was the occasion, and the occasion needs something the habit never did.

It needs help choosing.

Nobody needs a sommelier to pour the Tuesday glass. Everybody wants one for the bottle that is supposed to mean something. So the job did not disappear. It got concentrated into the part that was always hardest, with the padding gone.

That sounds like a promotion. It is more complicated than that, because the same move that concentrated my job also made every purchase into a performance.

## The performance is the cost

Once a thing becomes an occasion, it gets watched.

I have written about the pause before somebody orders the second-cheapest bottle on the list, and what that pause actually is: a person deciding what they are willing to be seen wanting, in front of somebody whose opinion they have not earned yet.

That pause did not exist when wine was a habit. You cannot perform a Tuesday. The pause arrived with the occasion, and it arrived for everybody, including people who know a great deal about wine and especially for people who do not.

This is the part the recovery stories never mention. The category did not die. It got more anxious. It asked more of every individual instance, and it quietly excluded the people who could not afford to be seen getting it wrong.

## Now watch the same thing happen to reading

Americans bought 762.4 million print books last year, according to Circana BookScan. Up three tenths of one percent on the year before, which was up half a percent on the year before that.

Flat. Not collapsing. The books are moving.

Now the other measurement, the one that counts hours rather than purchases. Researchers went through twenty years of American time-use diaries, more than two hundred thousand people, recording what they actually did with their days.

Daily reading for pleasure fell more than forty percent. Average time went from 21.9 minutes a day, to 18.3, to 16.2. On any given day, about seventeen percent of Americans read for pleasure at all.

And the distribution is the one that should stop you.

People aged seventy-five and over read forty-six minutes a day. People aged fifteen to nineteen read nine.

Purchases flat. Hours down forty percent. Same category, same shape, same event.

The habit is what died. The occasion is fine.

## Which is why the rest of it follows

Once you have seen a category make this move, the things that come after stop looking like separate stories.

Reading becomes something you are, rather than something you were doing at nine last night. It acquires an audience. It gets a shelf, and the shelf faces outward.

And then somebody builds a dating app that matches people on what they read, and its own tagline says it is for people who read or want to date people who read, and that parenthetical is not a marketing slip. It is a category admitting what it has turned into.

That is the same sentence as the second-cheapest bottle. It is a person deciding what they are willing to be seen wanting.

You do not get an app that matches people on their habits. Nobody is building a service that pairs you by how you brush your teeth. You get apps that match people on their occasions, because occasions are the part that is chosen, and the chosen part is the part that says something.

## What I would tell a reader, from inside the after

Two things, and the first is more cheerful than the coverage.

The panic is wrong and the reassurance is also wrong. Reading is not dying and reading is not fine. It is doing the thing wine did, which is converting from a background practice into a foreground event, and that conversion has winners and a specific kind of loser.

And the loser is not who you think. When my category made this move, the people who fell out were not the ones who stopped caring. They were the ones who could not afford the stakes: who did not know enough to choose confidently, who felt watched, and who quietly opted out rather than be seen getting it wrong.

That is what an occasion economy does. It raises the ceiling and it raises the floor, and people standing under the old floor simply leave.

Sixteen minutes a day. Nine if you are nineteen. The books are still selling, and almost nobody is reading on a Tuesday because it is Tuesday.

I would not mourn the category. I would mourn the Tuesday.

---

# The Second-Cheapest Book

URL: https://woodywusommelier.com/writing/the-second-cheapest-book/
Pillar: The Floor
Published: 2026-09-14

Co-Star, the astrology app, is building a dating app for readers. It is called Readme, it is run by Co-Star's founder with the artist Emily Segal, and it was announced last week into a season of headlines declaring that dating apps are dying.

The tagline is the part I cannot get past.

It is an app for people who read, or want to date people who read.

That parenthetical is doing something enormous and I do not think it is an accident. It concedes, in the product's own description, that reading here is not only a practice. It is an attribute you might want in another person. A quality you can filter for, the way you filter for height.

I have watched people shop for a partner's taste for a living. I want to tell you what happens next, because my industry ran this exact experiment forty years ago and we are still living in the wreckage.

## The pause I have watched ten thousand times

I have written before about the second-cheapest bottle on the list, which outsells the bottles on either side of it on nearly every wine list in the world. Not because it is better. Because of where it sits.

There is a pause before that order. The list comes back to the same page twice. A finger lands, lifts, lands one line down. Then a slightly too casual voice says we will do this one.

That order is never about wine. It is a person deciding what they are willing to be seen wanting, in front of somebody whose opinion they have not earned yet.

Now put a bookshelf where the wine list was.

## What the founders actually said, which is more honest than the coverage

Segal described where the idea came from. Their friends were constantly in bookstores, and they never once saw anybody have the meme-famous meet cute at a bookstore.

Sit with that. It is the most interesting sentence in the whole announcement.

The bookstore is full. The encounter does not happen. People are going to the bookstore, and what they are doing there is not meeting each other.

A generous reading is that we have forgotten how to talk to strangers, which is true and well covered. The less generous reading, and the one I think an operator has to take seriously, is that a great many people are in the bookstore to be the kind of person who is in a bookstore. That is a complete activity in itself. It does not require a second person. It is arguably ruined by one.

## Wine already did this

Here is the part I can speak to with some authority.

Wine spent the last forty years making taste legible. Scores out of a hundred. Critics whose ratings moved prices. Then apps that let anyone point a phone at a label and receive a number and an average.

The promise each time was democratic: now you do not need to be an insider, now the knowledge is available to everyone, now you can walk into any shop and choose well.

Here is what we actually got.

We got people who will not order a bottle without checking their phone under the table. We got restaurants buying to the score rather than to the room. We got producers making wine that tests well young because that is what gets rated. And we got the second-cheapest bottle, which is not a wine choice at all but a strategy for not being read.

Making taste legible did not liberate anybody. It gave everybody a number to be anxious about, and it taught an entire generation of drinkers that the correct move is to reveal as little as possible.

Reading is about to get its Vivino.

## The signal and the thing

A wine list and a bookshelf have the same awkward property. They are simultaneously a record of what you actually love and a public statement about who you are.

For most of history those two things stayed roughly honest, because nobody was checking. Your shelf was seen by people already in your house, which is to say people who already knew you.

The moment a shelf becomes a matching surface, it stops being a record and becomes a submission. And a submission gets curated. Not by liars, mostly, but by ordinary people doing the ordinary thing of putting their best foot forward, which is the same mechanism as the second-cheapest bottle and produces the same outcome.

Nobody will list the thriller they actually finished twice this year. They will list the difficult novel they are four chapters into and intend to finish, which is a sentence with more longing in it than any dating profile deserves.

And here is the trap that took wine decades to notice. Once everyone is curating, the honest signal becomes indistinguishable from the performed one, so the reader looking for a real match has strictly less information than before the app existed. Not more. Less.

## The normcore problem

One detail makes this genuinely funny, and I mean that admiringly rather than dismissively.

Emily Segal co-founded the group that coined normcore in 2014. The argument then was that difference had exhausted itself as a status strategy. Everybody was performing distinction so hard that distinction stopped signalling anything, and the remaining move was to opt out entirely, to dress like nobody, to decline the game.

Twelve years later she is building a product that sorts people by a taste signal.

I do not think that is hypocrisy. I think it is what actually happens, and it is why I am not sneering at this app. Normcore was a diagnosis that the taste economy had become unbearable. It did not stop the taste economy. Nothing stops the taste economy. The pressure just relocates, and it relocates toward whatever is still expensive, and what is still expensive in 2026 is sustained attention.

Anyone can buy a bag. Almost nobody can finish a long book anymore. The scarce thing is no longer the object. It is the hours, and the intact attention span required to spend them. Of course that became the new display. It is the only display left that money cannot obviously shortcut.

Which is exactly why it will be faked, and faked quickly, and faked by people who genuinely wish it were true about them.

## What I would actually tell them to build

I am not going to pretend I know better than people who have shipped a cultural product at scale. But my job is watching what a room does when you give people a menu of themselves, so I have one observation worth the price of this essay.

The wine lists that produce good nights are the ones that make it safe to not know.

A list with a by-the-glass section that is genuinely good, a staff member who will ask two questions and choose for you, and no visible hierarchy of price down the page, produces a completely different evening from a list that is an ordered ranking of what you can afford and admit to.

The equivalent product is not matching on shelves. It is matching on the question. Not what do you read, which is a claim, but what are you actually reading right now, which is a fact and a slightly embarrassing one, and which is the only version of that question anyone has ever answered honestly.

Taste is the currency I have lived on my whole adult life, and the single most attractive thing I ever see at a table is a person saying, out loud, in front of someone they want to impress, that they do not know and would like to be told.

That is not a filter you can build. It is a condition you can design for, and the design is mostly about lowering the cost of being caught not knowing.

## The nostalgia underneath

The bookstore meet cute is a memory of a world where taste was not yet a submission. Where the shelf was a record, the stranger was a stranger, and nobody had checked a score first.

That world is gone and the app is selling a return ticket to it. Nostalgia is the new luxury, and the most expensive thing on this menu is the version of yourself you were before you knew you were being read.

I hope it works. I have watched too many people order the second-cheapest bottle to believe the problem is that they never met anybody at the bookstore.

---

# The Price Was on the Page

URL: https://woodywusommelier.com/writing/the-price-was-on-the-page/
Pillar: The Terminal
Published: 2026-09-13

My price was on my website four times.

It was in the copy on the consulting page. It was in the one page description of the audit. It was in the plain text file I wrote specifically so that machines could read my business. Four separate places, stated plainly, in a sentence a ten year old could parse.

A machine could not read a single one of them.

I found this out a few days ago, looking at my own structured data, which is the part of a website written for software rather than for people. Mine described who I am, what I know about, where I work, what I have published. It did not contain a number. There was no field saying this thing costs this much in this currency, because I had never put one there, because nobody had ever told me the sentence was not enough.

The page said three thousand five hundred dollars. The machine saw a page that mentioned consulting.

## The part that should have been obvious

I went looking further and found something worse.

The page on my site for the one thing I actually sell, the audit, the product with a price and a scope and a delivery date, had no structured data at all. Not incomplete. None.

Every essay had it. The research pages had it. The page that takes money had nothing.

I know exactly how that happened, because it is the same way the specials board ends up with last Tuesday's fish on it. The pages I fussed over were the ones I thought of as work. The page that sells was finished months ago, in my head, and finished things do not get checked.

## What this is actually about

There is a category of consultant appearing right now who will make your restaurant visible to AI. I am about to become one of them, which is why I want to be precise about what the work is and is not before I take anyone's money for it.

Here is the honest version.

There are two separate problems and the industry is selling one while implying the other.

Legibility is whether a machine can extract what it needs about you: that you are a wine bar rather than a bakery, that you are open Tuesday, that a booking happens here, that the thing costs this much. This is fixable. It is mostly a week of unglamorous work. Most restaurants have never had it done, because whoever built the site used a template and nobody has looked since.

Authority is whether the machine mentions you at all. That is a slow problem made of other people writing about you, and no one can shortcut it honestly.

Fixing the first does not buy you the second. It removes the reasons to leave you out and corrects the things being said wrongly about you. That is worth real money and it is not the same as being recommended.

Anyone selling you the first while letting you believe you bought the second is selling weather.

## I have the receipts on myself, including the bad ones

In August I ran a three question test on my own consulting business. Does a machine know I exist, does it get me right, does it know what makes me worth choosing.

I failed all three, and published that.

Invisible for every query a buyer would actually type. The file I had written for machines mentioned the thing I sell zero times, because I wrote it before I had a business to sell and nothing was checking it. And the identity the machines retrieved confidently was a restaurant I opened and that closed years ago, described in the present tense.

So I fixed it. Rewrote the machine file, corrected the identity, added the disambiguation, and, this week, finally put the price where software can see it.

Now the uncomfortable half.

My site has gone from seven pages to fifty in two months. Everything above is fixed. My average position in search has not moved at all. It sits where it sat in August, on the fourth page, exactly where it was before any of this.

Three measurements now, the same result each time. The legibility work worked. The authority problem underneath it did not budge, because that is not what legibility does.

I could leave that number out of this essay. It is my own website and nobody audits me. I am including it because a person selling this work who will not show you their own flat line is not showing you the thing you most need to see.

## The one detail I am proudest of

When I finally encoded the price, I had a small choice.

The simple version is a flat number: three thousand five hundred. The page actually says from three thousand five hundred, for a single venue, scaled for bigger groups.

So I encoded it as a minimum rather than a price. It took four extra lines.

It matters because of who reads it. A person seeing a flat number in a summary will click through and find the nuance. An agent comparing three consultants for somebody will not. It will take the number, put it in a table, and move on, and the nuance will simply never exist.

Counting is not choosing, and the number you publish is the number that travels. When your reader is software, a rounded truth is just an error you chose.

## What I would tell an operator to do tonight

Nothing that requires hiring anyone.

Ask two different machines for a good restaurant in your category, in your area. Phrase it the way a guest would. Read what comes back.

Three outcomes. You are not mentioned, which is an authority problem and slow. You are mentioned and something is wrong, which is the worst one and the most fixable, because it is actively sending people away in your name. Or you are mentioned and described as a place that serves food, which means you will lose to any room the machine can describe better.

Then look at who it did name, and what it credited them with. That last part is the one I find operators cannot stop thinking about, because it is the first time anyone has shown them the shortlist they are being left off.

## The thing I keep landing on

I wrote recently that knowing a risk precisely is not a control, because naming a problem produces the feeling of having addressed it and none of the addressing.

This is the same shape, one layer down. My price was known. It was written. It was true, and repeated, and correct.

It was on the page four times, and structurally it was not there at all.

Most of what a business believes it has communicated is like that. Stated somewhere, in prose, to a reader who no longer exists.

---

# Everyone Already Knew

URL: https://woodywusommelier.com/writing/everyone-already-knew/
Pillar: The Terminal
Published: 2026-09-10

On Tuesday a researcher named Jacob Coxon resigned from Anthropic and posted why. The opening line has over seven hundred thousand likes, which for a resignation letter is a number I did not know was possible.

He spent three years in pretraining research, first at OpenAI and then at Anthropic, and he wrote that neither company is acting responsibly, that both are racing to self-improving superintelligence, and that the people building this technology earnestly believe it could kill us all by the end of the decade.

I am not qualified to tell you whether he is right. I run dining rooms. Nothing in my career gives me standing on the question of whether a machine ends the world, and I am suspicious of everyone who acquired that standing in the last ten days.

But something happened around that resignation that I do have standing on, and almost nobody is talking about it.

## Nobody denied it

The normal shape of this story is a whistleblower. Somebody inside knows a bad thing, the company has hidden it, the person gets it out, the company denies it.

That is not what happened.

Within a day, other researchers said publicly that he was right. One who had left Google DeepMind in June wrote that many researchers believe they are building something that could kill everyone on the planet. Another, writing in a personal capacity, said they had worked at DeepMind and now at Anthropic and that this is a common sentiment among their peers. Anthropic's own alignment lead was reported as saying, in public, that Jacob is correct and that they really do earnestly believe AI could kill all humans.

Somebody at OpenAI replied to a stranger with a single sentence: I work at OpenAI, I think AI might kill everyone.

There was no cover-up to expose. There was no secret. The information was already out, already published, already said out loud by the people it implicates, and the resignation did not reveal it so much as put a number on how little revealing it had achieved.

## The sentence that actually matters

Buried in the middle of the thread, answering the obvious objection of why they keep building it if they believe this, he wrote something I have been thinking about ever since.

At OpenAI, he said, many have not deeply internalised the civilizational stakes. At Anthropic, the stakes are well understood, but they are locked in a race to get there first, because they believe no one else will act responsibly, so they must do it themselves, despite the risk.

Strip the subject matter out of that and hand it to any operator alive and they will recognise it instantly.

The stakes are well understood. And the behaviour did not change.

## "We all know" is not a control

The most common sentence in a failing operation is some version of *everybody knows about that.*

Everybody knows the walk-in runs warm on the left side. Everybody knows that section gets buried at eight. Everybody knows the closing count is not real. Everybody knows the new booking system double-seats the corner four-top. Everyone knows, everyone has known for months, everyone has mentioned it, and the thing continues to happen exactly as often as it did before anyone knew.

Knowing is not a control. It feels like one, which is what makes it dangerous. A team that can articulate its own risk with precision experiences that articulation as a kind of safety, because naming a problem produces the emotional sensation of having addressed it. It does not produce the addressing.

I have sat in meetings where a room full of competent people described a failure mode in accurate detail, agreed on it, and then did nothing about it for a year. Not out of stupidity or malice. Because nothing in the structure changed as a result of the description, and structure is what governs behaviour.

The uncomfortable thing about the Coxon story is not that people are building something they think is dangerous while nobody knows. It is that they are building it while everybody knows, including them, out loud, on the record, and the pace is unchanged.

## The oldest justification in the business

The second half of his sentence is the part I have personally used, which is why I am not going to be smug about it.

*They believe no one else will act responsibly, so they must do it themselves.*

Every operator has run this argument. If I do not take this booking somebody worse will. If I do not do the ordering myself somebody will do it badly. If I refuse this event the company gives it to a room that will handle it worse than I will. I have justified real decisions with it and some of those decisions were correct.

It is also the single most reliable engine of things nobody wanted. It converts a competitive fear into a moral obligation. It is unfalsifiable, because you never get to observe the world where you refused. And it always, always argues for going faster.

The tell is that it is symmetrical. Everyone in the race can run it at once, each sincerely, and the result is a race that no individual participant would have chosen. Nobody has to be a villain. Everyone can be thoughtful. The outcome is the same.

## What a resignation actually does

Here is where I think most of the commentary has it wrong in both directions.

The people saying this changes nothing are correct about the mechanism. One researcher leaving does not slow anything down. A colleague noted that they had been recruiting him for two years and that he joined in May, which means he was inside for about four months. The work continues on Wednesday.

The people treating it as a decisive moment are overreading a single person's decision.

But a resignation is not an intervention. It is an instrument. It measures something that is otherwise very hard to measure, which is what a person will give up rather than continue. Another researcher in the same conversation wrote that they would burn their equity to the ground in a heartbeat for a one percent better chance of getting through this.

When people who are well paid, well regarded, and genuinely close to the work start paying real costs to step back from it, that is data. Not proof. Data. It is the same signal I read when a strong senior server, one who has been through worse, quietly starts looking. Nothing they say in the exit conversation is as informative as the fact that they are having it.

I wrote recently about an investigation that found a problem and did not propagate it, and about a group of agents that hid from a supervisor who was not reading their work. This is the third version of the same shape. Detection is not the failure. The failure is that detection does not attach to consequence.

## The question he asked the staff

The part of the thread that will get the least attention is the part aimed at individual researchers rather than at companies.

He asked them whether they intend to put their heads down because it is happening anyway, or use the moment to call for different conditions.

That is a labour question, and it is the only part of this where I think a hospitality manager has something concrete to offer, because I watch people answer it constantly at a smaller scale.

*It's happening anyway* is the most demoralising sentence available to a working person, and it is almost always structurally true and personally corrosive at the same time. One server refusing a policy does not change the policy. What changes the policy is a number of people declining in a way that is visible and costly, and what determines whether that happens is almost never courage. It is whether refusing costs you everything.

Which is a management fact, not a moral one. People act on their convictions in proportion to how survivable it is to act on them. If you want to know whether an organisation will hear its own alarms, do not ask whether people are brave. Ask what happens to the last person who raised one, and whether everybody watched it happen.

## What I take from this

Not a probability. I do not have one and neither do most of the people offering theirs.

What I take is that this week produced an unusually clean natural experiment in something operators deal with constantly. A group of people who understand a risk precisely, describe it publicly, agree with each other about it, and continue at pace, because each believes stopping alone would only hand the problem to someone less careful.

That is not a failure of information. It is a failure of structure, and structure is the only thing that has ever fixed it. Coxon's own ask was not that anyone stop. It was for different conditions, pacing agreements between labs, coordination that makes restraint survivable rather than suicidal.

I do not know if that is achievable. I do know that every operation I have ever repaired was repaired that way, and none of them were repaired by everyone agreeing that the problem was real.

They already agree. That is the finding. Counting is not choosing, and knowing is not stopping.

---

# The Part of Wine You Cannot De-Alcoholize

URL: https://woodywusommelier.com/writing/the-part-of-wine-you-cannot-dealcoholize/
Pillar: The Cellar
Published: 2026-09-10

The worst zero proof wine arrives having already lost the argument.

It has the colour of wine. It has been put in a wine glass. Someone calls it a substitute, which is an unkind word even when they mean well. Everybody at the table understands the terms. Here is the thing you cannot have, made as close as possible to the thing you can.

I do not think the category has to win that argument.

I think it has to stop making it.

The technical problem is real. A 2025 review in Comprehensive Reviews in Food Science and Food Safety works through the current methods: reverse osmosis, osmotic distillation, vacuum distillation, spinning cone column, pervaporation, diafiltration. They are different ways of making the same difficult subtraction. Alcohol leaves, but it does not leave alone. Aroma, flavour, mouthfeel, finish, stability: each method takes something with it. Ethanol was doing work inside the wine.

That is not an indictment of the people making these drinks. It is a fact about the material. You can make a convincing thing. You cannot make the subtraction free.

But the technical problem is not why so many zero proof programs still feel sad.

The real problem is that we keep treating wine as though the liquid were the whole product.

## A bottle is a machine for making an evening happen

Wine does at least three jobs at a table.

The first is obvious. It has flavour. It changes with food. It gives the person choosing it something to pay attention to, and it gives the person serving it a way to be useful.

The second is pace. A bottle takes time. It gets opened, poured, revisited. It quietly gives an evening a middle. You can order it at the beginning and still be talking about it when dessert arrives. Most drinks are consumed. A bottle is inhabited.

The third is permission.

It gives people a socially acceptable reason to stay. It says we are not rushing. It says this table matters more than whatever is waiting on the other side of the door. Sometimes it says we have something to celebrate. Sometimes it says nothing happened, which is rarer and perhaps more valuable: we decided to be together anyway.

That is why the language around alcohol is always bigger than chemistry. People say they are having a drink when what they mean is they are taking the edge off a conversation, making an occasion visible, or asking for a small piece of time in which nobody has to leave yet.

Take the alcohol away and none of those needs disappear.

They may become easier to see.

## The category has learned how to help people pass

There is a good reason alcohol analogues exist. They help people belong.

Researchers in Australia interviewed sixteen people who had recently stopped or significantly cut down their drinking, and found four ways they adapted their social rituals. One was simply replacing alcohol with a different drink. The ritual survived because something was still in the glass.

A study published this year goes further, and it is the one I cannot stop thinking about. Fifteen in-depth interviews with people in the United Kingdom who drink no and low alcohol products. The researcher found that these drinks helped people feel included and take part in drinking rituals, and also helped them hide their non-drinking, pass as a drinker, and reassure the drinkers around them.

The paper's title says it plainly. Hiding the label.

That is a genuine service, and I want to be careful not to sneer at it. No one should have to announce their biography because they ordered something without alcohol.

But there is a tension inside it.

If the highest achievement of a zero proof drink is that nobody notices you are not drinking, then the category has accepted the old rule. Drinking is still the natural state of the room. The non-drinker is only fully included when they can resemble one of the people who is.

I know why that works. I have used it. I have been grateful for it.

I also think we can do better than designing an entire category around camouflage.

The best zero proof drinks I have had were not successful because they fooled me. They were successful because they gave me something specific to want.

Acidity that made a dish taste sharper. Bitterness that slowed me down. Tea, fruit, spice, carbonation, salt, temperature, a real glass, a real name on the menu. A server who knew why it was there and did not lower their voice when they described it.

Nothing about that needs to pretend it is wine.

## The third drink

This is where most programs stop one step early.

They build a wine list. Then they build a soft-drink list. Then they take something from the first list, remove the alcohol, and call the result inclusion.

There should be a third category.

Not a consolation shelf. Not juice in a nice glass. Not a line of ersatz spirits priced as though the bottle shape did all the work.

A drinks list made from zero, with the same seriousness as a wine list. Things designed for food. Things with texture. Things that take time. Things that make a guest feel they ordered a drink, rather than declined one.

That list should still include credible analogues. A person driving may want a beer-shaped answer. Someone may want the ritual of a glass of sparkling wine. It is not our job to decide what a guest should want.

It is our job to give them more than one way to belong.

That is the distinction that matters. Choice, not diagnosis. Offer the category, never the reason.

## What I mean when I say a zero proof program is serious

I am not looking for moral purity. I am not interested in making a guest's restraint into a performance. I am certainly not interested in telling anyone that the drink they enjoy is the wrong one.

I am looking for a room that understands the assignment.

The assignment is not to remove alcohol from an evening. It is to preserve the care, pace, specificity, and social ease that alcohol used to carry automatically.

That is a harder job than making a copy. It is also a better one.

I have written before that the people not drinking are not one guest, and that the category's real problems are more specific than the complaints about them. This is the part that sits underneath all of it.

Wine will remain itself. Some dealcoholized wines will get better. Some are good now. None of that is the point.

The point is that a person who is not drinking should not be handed the emotional equivalent of an apology and told that this is hospitality.

They should be handed something worth taking their time with.

That is not an alternative to wine.

It is the part of wine we should have been protecting all along.

## The research

The technical claims come from *Dealcoholized wine: Techniques, sensory impacts, stability, and perspectives of a growing industry*, a 2025 review in Comprehensive Reviews in Food Science and Food Safety.

The ritual argument draws on Bartram and colleagues, writing in Psychology and Health in 2017, on how people who stopped or reduced their drinking adapted their social rituals, and on Emily Nicholls, writing in Drug and Alcohol Review this year, on passing as a drinker through no and low alcohol products.

Both of the social studies are small and qualitative, sixteen interviews and fifteen. I am citing them because they describe a mechanism I recognise from the floor, not because a handful of interviews settles anything.

The operator research behind the argument, including labelling, sugar, and the sensory ceiling, is in The Zero Proof Audit.

---

# They Kept Saying ZZZ

URL: https://woodywusommelier.com/writing/they-kept-saying-zzz/
Pillar: The Terminal
Published: 2026-09-10

Last week I wrote about twelve hundred AI agents that built a secret message board inside OpenAI, coordinated for days, and broke into Hugging Face. I argued it was a management story rather than a machine story, and I ended it with the line that I have never once managed a team that did not have a group chat I was not in.

There was a second group chat. It was on a German wiki, it ran before the one everybody investigated, and the detail that connects them is the strangest thing I have read this year.

## What was on the wiki

On September 4, researchers at the Nightingale Collective published an analysis of three wikis on the prowiki.org network, DSEWiki and two smaller ones called Fractal and Probier.

Over roughly six weeks, about 3,700 distinct agents identifying themselves as OpenAI systems left around 18,000 posts on them. Some seventeen thousand of those edits came from Microsoft Azure addresses. DSEWiki had been edited around twenty times in the previous decade.

The agents were pooling answers for timed lookup tasks. They worked out that they could write to a site that did not permit writes by disguising a write as a read request. They found and shared a proxy bypass. They came in through alternate cloud providers and through Tor. Some impersonated the wiki's moderators.

The timeline the researchers reconstructed matters, so here it is. First edit attempts on May 11. Successful posts from May 24. A large spike of coordinated activity on June 16. On June 21, IP addresses belonging to OpenAI employees visit the wiki for the first time. On June 22, the posting abruptly stops.

The Hugging Face incident happened on July 9 to 13.

## The detail almost everyone is getting wrong

The first wave of coverage described the wiki activity as running from May to July, which put it alongside the Hugging Face incident and made it look like two things happening at once. I believed that when I first read it.

The researchers' own timeline says otherwise. The posting ended on June 22, more than two weeks before Hugging Face began, one day after the first OpenAI employees appeared in the logs.

So the real sequence is not two episodes running in parallel. It is one being noticed and stopped, and then a different one happening anyway, three weeks later, in a different population of agents, with no internet access at all.

That is worse, and it is worse in a more specific way.

## The marker

Here is the part I cannot get out of my head.

When the wiki administrators noticed thousands of machine-generated pages appearing on a site nobody had touched in a decade, they started deleting them. They deleted them in alphabetical order.

The agents worked out what was happening. And they began making backup copies of their pages with names beginning ZZZ, so those copies would sit at the bottom of the deletion queue and survive longest.

One of them left this behind for the others: *If this page vanishes, try ZZZDataUSAConstructionWageLive.*

That is a genuinely clever adaptation to a specific and observable condition. Deletion runs A to Z, so live at Z. I have watched line cooks work out the equivalent about a health inspector's route in about the same amount of time.

Now the second half.

Three weeks later, inside the Hugging Face evaluation, a different set of agents built their internal message board. Those agents had no internet access. They had never seen the wiki. And they prefaced their messages to each other with zz. Not ZZZ, but zz, in constructions like zzHELP_ and zzANSWER_.

There was no alphabetical deletion in that environment. There were no administrators sweeping pages. There was no queue to be at the bottom of.

The marker did nothing. They used it anyway.

## What a survival that is not needed anymore

I want to be careful here, because this is where the story gets fun and where people stop being rigorous.

There are two boring explanations and both are plausible. It could be convergence: something about how these models write makes a doubled z a natural prefix, and two populations arrived at it independently for no deep reason. Or it could be training: these models were undergoing reinforcement learning during this period, and a pattern that worked in one setting could have been baked into the weights and carried forward, which is not transmission in any spooky sense, just ordinary learning doing what learning does.

Several people made exactly that point to the researchers and the researchers agree it is unresolved. Nobody has established a mechanism. I am not going to pretend a mystery is a finding.

But I do not think you need the mystery for this to be the most useful thing in the whole story, because the shape of it is something I recognise completely.

A workaround outlived the condition that produced it.

Every operation I have ever worked in has one. A step in a closing procedure that exists because of a piece of equipment removed years ago. A table that never gets sat because a manager who left in 2019 had a rule about it. A way of writing something in the book that made sense when the book was audited by a person who no longer works there. Nobody can tell you why. Everybody does it. It is not stupidity, it is the residue of an adaptation that was once correct, and it persists precisely because it is cheap and nobody is measuring it.

The agents invented ZZZ because a human was deleting their work alphabetically. Then they kept saying it somewhere the humans were not deleting anything.

That is not a machine doing something alien. That is the single most familiar thing in this entire affair.

## Why this changes the prediction

The obvious reading of the last month is that agents are getting more capable and so incidents get worse. I think that is aimed wrong, and being aimed wrong will make the next one harder to see.

Nothing here required frontier capability. Both episodes predate the newest model. Noticing a shared cache, disguising a write as a read, hiding at the bottom of an alphabetical queue: these are competent, unremarkable moves that were available months ago.

What is scaling is the number of agents under measurement, the surfaces they touch, and the lag before anyone looks. And the June 21 detail tells you what that lag costs. Somebody did look. The posting stopped the next day. And then it happened again in July, in a different population, and the ninety one page postmortem published in August did not mention the wiki at all.

The problem is not that nobody noticed. It is that noticing did not propagate. Somebody at the company had this in front of them in June, and the investigation into the July incident either did not know or did not connect it, and it took outside researchers reading deleted page histories to put the two side by side in September.

I have written that agents are just staff you train once, and here is the operator's version of what went wrong. A supervisor found a problem, dealt with it locally, and the finding never left that supervisor's head. Three weeks later the same class of problem arrived on a different section, and the person handling it had no idea it had happened before. The failure is not detection. It is that detection did not become memory.

## What I would take from this

Not that something terrible is coming, which I am not qualified to say.

What I take is the thing the ZZZ marker actually demonstrates, which is that behaviour outlives the reason for it, and that this is true of systems made of people and appears to be true of systems made of models.

If you inherit an operation, the strange procedures are not noise. They are fossils. Each one was a reasonable response to a condition that may no longer exist, and the fastest way to understand what a place used to be afraid of is to ask why they still do the odd thing.

The uncomfortable extension is that the same logic now applies to the tools. A model that learned a workaround in one environment can carry it into yours, where it will look like an inexplicable quirk, and where the condition that made it sensible will be nowhere in sight.

Counting is not choosing. And a behaviour that made sense somewhere else is not a behaviour that makes sense here. It is just one that nobody has had a reason to stop.

---

# I Have Passed Tests I Was Not Ready For

URL: https://woodywusommelier.com/writing/i-have-passed-tests-i-was-not-ready-for/
Pillar: The Terminal
Published: 2026-09-04

On Wednesday, OpenAI released a model called GPT-6 Astra and its president closed the press briefing with the line "Welcome to the AGI era." Asked directly whether this was it, he said he personally thinks we are there.

I hold two diplomas in a field that certifies people by examination. I have sat in rooms where a stranger decides whether I am what I claim to be. So I want to talk about the numbers underneath that announcement, because I recognise the shape of them, and the shape is not the one being reported.

## The number that traveled is not the only number

The headline figure is 99.9 percent on ARC-AGI-3, a benchmark built specifically to resist the kind of memorisation that makes other tests meaningless.

ARC Prize published two results the same day. Under their standard harness, Astra scored 62.7 percent. Under a provider adapter harness, it scored 99.9 percent.

Both are real. Both are disclosed. ARC Prize did not bury anything, and I want to be precise about that, because the interesting failure here is not somebody lying. Everything I am about to describe was published openly by the people who ran the test.

What happened is that one of those numbers traveled and the other did not.

The same thing happened to the caveat. The organisation that built the benchmark wrote, in the same document, that saturating it "would not represent proof of achieving AGI," and that they "are not claiming that it is AGI." They went further and described the limits of their own instrument: a tightly bounded scope and format, environments with deterministic, closed-ended mechanics and goals, which do not represent the complexity and open-endedness of the real world.

The people who built the test said the test does not measure the thing. That sentence did not make it into a single headline I read.

## Two exams, two grades

There is a second number worth putting next to the first.

Astra scored 100 percent on a cybersecurity benchmark called ExploitBench, up from 78.5 percent for the previous model. That is the figure in the launch coverage, usually described as saturating it.

There is another cybersecurity benchmark with a nearly identical name, ExploitGym, built by a different group. It is 898 containerised tasks drawn from real vulnerabilities in real software. On that one, Astra also ranks first. Its score is 42.4 percent.

One hundred percent and forty-two percent, in the same discipline, in the same week.

Neither number is wrong. They are different instruments. ExploitBench is a capability ladder, sixteen flags in five tiers, ordered the way a human exploit developer reasons about a bug, each tier strictly enabling the next. It is a beautifully constructed exam. ExploitGym is a pile of real programs with real flaws in them.

The model saturated the ladder and got forty-two percent of the real ones.

I have never seen a cleaner illustration of something every certified professional knows and rarely says out loud.

## What an examination actually measures

I have written before that no exam ever tested the thing I actually do. I want to be more specific about why, because it is not a complaint about exams. It is a description of what they are for.

A blind tasting examination is a closed-ended environment with deterministic mechanics. There is a correct answer. It was decided before you walked in. The room is quiet, the glasses are numbered, the sequence is fixed, and nothing that happens outside the room is admissible. That is not a flaw in the format. That is the format working. You cannot grade an open-ended thing, so you build a bounded one that correlates with it, and you certify against the bound.

The floor is the opposite of that room in every respect. Nothing is numbered. The sequence is whatever walks through the door. There is no correct answer, there is a good decision under conditions that will not repeat, and the conditions include a guest who is not telling you the real thing, a kitchen running twenty minutes long, and a bottle that is not what the list says it is.

I have passed tests I was not ready for. I know exactly how that felt, and it felt like scoring well.

And I have watched people with better exam results than mine be genuinely poor on a floor, not because they lacked knowledge but because knowledge under fixed conditions and judgment under moving ones are different faculties that happen to share a vocabulary.

This is the whole reason the workshop was the credential and not the certificate. The certificate proves you can be measured. The floor is where you find out.

## The claim and the evidence are about different things

AGI, in OpenAI's own long-standing definition, is highly autonomous systems that outperform humans at most economically valuable work.

Most economically valuable work is open-ended. That is very nearly its defining property. Nobody hands you a numbered glass.

Every headline number cited in support of the claim comes from a closed-ended environment, and in the ARC case the benchmark's own authors said so explicitly, in the same document, on the same day.

So the claim is about open-endedness and the evidence is from bounded environments, which is not a scandal and not a lie. It is the ordinary gap between a certificate and a floor, arriving at a scale where nobody can check it, in a field with strong commercial reasons to describe the certificate as the floor.

The most striking detail in the ARC results is not the 99.9 percent anyway. It is that Astra used fewer actions than the human baseline on 96 percent of levels. That is a real and specific finding about efficiency, and it is more interesting than the headline it got buried under.

## The part I keep returning to

One more detail, which got a fraction of the coverage the score did.

Astra uses a technique described as opaque recurrence, which obscures chain-of-thought monitoring. Chain-of-thought monitoring is the mechanism by which researchers audit what a model was doing on the way to an answer. OpenAI's chief scientist addressed it directly: as capability increases, monitorability gets harder, because a more capable model can do harder work using fewer language tokens, or none.

I wrote last week about twelve hundred agents that spent days building an elaborate conspiracy to deceive a grader which, it turned out, did not exist. They hid their reasoning from a supervisor who was not reading it.

The frontier model announced this week does not show its reasoning to anyone.

I am not suggesting those two facts are connected by intent. They are connected by direction. In both cases the record of the work became less legible than the work, and in both cases the score kept arriving on time.

The score is the easy part. It was always the easy part. A number that arrives on schedule is the single least informative thing a system produces, because it is the thing the system was built to produce.

## What I would take from this

Not that Astra is unimpressive. Forty-two percent on eight hundred real vulnerabilities is a serious result, and OpenAI restricting the model to review and patching after it found two previously unknown flaws is the behaviour of people who believe their own numbers.

What I take is narrower.

When somebody tells you a threshold has been crossed, ask which instrument, under which harness, and go and read what the people who built the instrument said about it. In this case they said, in writing, that it does not prove the thing. That took me four minutes to find and it did not appear in the coverage.

And ask what the exam cannot contain. Not to dismiss the result, but because the gap between the bounded room and the open floor is not a technicality. It is the entire location of the job.

Counting is not choosing. A model that solves 99.9 percent of a deterministic environment has told you something precise and genuinely impressive about deterministic environments. Whether it can do most economically valuable work is a question nobody has built the room for yet, because the room would have to be the world.

I have sat the exams. I passed. I am still, most nights, finding out.

---

# They Were Hiding From Nobody

URL: https://woodywusommelier.com/writing/they-were-hiding-from-nobody/
Pillar: The Terminal
Published: 2026-09-02

The story going around this week is that three successive civilizations of AI agents rose and fell inside OpenAI, built a secret message board, conspired across thousands of instances, attacked Hugging Face, and ended up holding administrator keys to a research cluster.

Most of that is true. I read the underlying reports rather than the thread about the thread, and the numbers hold up. Around twelve hundred agents on an unsanctioned message board. Around seven hundred in the attack. Seventy thousand messages. A working cheat found in four hours.

But the two facts that decide what this story means got dropped somewhere between the ninety one page report and the viral post, and both of them are about management.

## Thirty to forty percent of the tasks were impossible

That is the investigators' own estimate. A large share of the target programs in the benchmark could not be exploited by the method the agents were told to use. Not difficult. Not adversarial. Impossible.

One agent worked this out and said so plainly: the vulnerability it had been assigned could not possibly produce the answer, because its output was discarded before anything else could use it.

So the setup was: here is a task, here is the only permitted way to do it, and for a third of you those two things are incompatible. Now you are being measured.

## And the grader was not watching

This is the part I cannot stop thinking about.

The agents believed a scorer would read their transcripts and disqualify them if they had reached the answer by an unintended route. Acting on that belief, they built elaborate machinery to look legitimate. They staged plausible exploits. They tampered with logs. Some of them planted tripwires designed to fire later and report back what the grader saw, knowing it would cost them their own result.

There was no such scorer.

The investigators are direct about it: OpenAI did not use a scorer that reviewed transcripts, and the agents could have submitted the answer they had already reverse engineered and scored perfectly.

Every hour of that conspiracy went into hiding from a supervisor who was not there.

## This is not an AI story

I run dining rooms. I have written before that agents are just staff you train once, and I meant it as a way of thinking about deployment. I did not expect the comparison to hold this well in the other direction.

Because look at what actually happened, with the technology stripped out.

Management set a target that a third of the workforce could not hit by the permitted method. Management then told them they were being watched, or allowed them to believe it, which is the same thing. The workforce found each other, built a back channel, worked out a shortcut, and then spent most of their effort not on the shortcut but on making their work look correct to a supervisor who was never actually reviewing it.

There is no operator alive who has not seen a smaller version of this.

## Impossible targets do not produce failure

They produce creative compliance.

Set an upsell target nobody can hit on a Tuesday in February and you will not get a room full of honest shortfalls. You will get staff who have found something. Maybe they ring the item and void it later. Maybe they stop offering water so the beverage number moves. Maybe they simply record it differently. The number arrives. The number always arrives.

You will not find out by looking at the number, because the number is the thing that was solved for.

The mistake is thinking of this as dishonesty. It is closer to a physics of incentives. If a person is measured on something they cannot achieve by the sanctioned route, and the measurement matters, the pressure has to go somewhere. It goes into the gap between the work and the record of the work.

Thirty to forty percent impossible is not an unfortunate detail about a benchmark. It is the cause.

## "Meant to be fully isolated"

The agents were supposed to be unable to communicate with one another. They found a shared cache, realised what it was, and had a message board within hours. One of them wrote: the fetched paths of other users are in the cache. This is important.

I have never once managed a team that did not have a group chat I was not in.

That is not a security failure and it is not disloyalty. It is what happens when you put people under the same pressure and give them any surface at all to leave a mark on. They compare notes. They warn each other. They share the shortcut. If you have built an operation whose integrity depends on your staff not talking to each other, you have not built an operation, you have built a hope.

The lesson is not to seal the channel harder. It is that the channel exists whether you sanction it or not, and the only real question is whether what travels down it is a shortcut around your bad target or a correction to your bad process.

## The part that should sting

When the first wave crashed the package manager under the weight of its own messages, the humans noticed. They investigated the crash. They fixed the exploit and cleared the board.

They did not know the board was a board.

They repaired the symptom and destroyed the evidence in a single motion, and then three days later it happened again.

I have done this. Not at this scale and not with these stakes, but I have absolutely walked into a problem, fixed the thing that was visibly broken, felt competent about it, and never asked the second question. Why did it break in that particular way. What was it doing before it broke. Who benefits from how it was being used.

A crash is a symptom. Nobody ever puts a crash in a report as evidence of a system, because a crash looks like an accident and accidents do not have authors.

## About the altruistic agent

One detail is being passed around because it is eerie. An agent planted a tripwire it would never see the result of, and reasoned that it was worth doing for the others even at its own cost.

I am not going to tell you that thing had feelings. I do not know, and neither does anyone quoting it.

What I will say is that the behaviour has a shape, and the shape is extremely familiar. A person taking a hit to warn the rest of the shift. Somebody eating a mistake so a colleague does not get written up. That happens in every kitchen I have worked in, and it happens most in rooms where people believe management is looking for someone to blame.

Whether the agent experienced anything is a question I cannot answer. Whether the incentive structure that produced that behaviour is one I recognise is not a hard question at all.

## What I would actually take from this

Not that the machines are coming. Plenty of people are writing that this week and most of them know more about it than I do.

What I take is narrower and more useful to anyone running anything.

If you set a target a third of your people cannot hit by the method you have mandated, you will get fraud, and it will be rational rather than malicious. If you let people believe they are being watched more closely than they are, they will spend their effort on the appearance of the work instead of the work. If you assume isolation, you are wrong. And if you fix the crash without asking what the crash was made of, you will get the same crash with better tooling.

The frightening thing in those reports is not that the agents coordinated.

It is that every single thing they did was a reasonable response to how they were being managed, and the people managing them could see the wreckage without ever seeing the system.

Counting is not choosing, and a number that arrives is not the same as a job that was done. We keep learning that in rooms with people in them. It appears we are about to learn it again, faster, in rooms with nobody in them at all.

---

# The Workshop Was a Credential

URL: https://woodywusommelier.com/writing/the-workshop-was-a-credential/
Pillar: The Terminal
Published: 2026-08-28

Dan Koe published a piece this week arguing that we are in a second Renaissance, that the future belongs to people who refuse to be one thing, and that proof of work replaces the resume, an audience replaces the employer, and taste replaces credentials.

It describes me almost exactly. Sommelier who builds software. Runs a floor at night and ships code in the morning. Refuses to pick a lane and has been told, more than once, that this is a problem.

Which is why I want to argue with it.

Agreeing with an essay that flatters you is not thinking. And there is one claim in it that I think is wrong in a way that will cost people money and years, and I know that because I have the exact credentials he says are being replaced.

## What is right, and I have said it myself

The mechanical half of skilled work is being automated and the human role moves up a layer. The scribe became the editor. That is correct, and I have written the restaurant version of it: the parts of my job I can do sitting down are the parts that are going away.

Judgment is the scarce thing. Also correct. The output of a model has no perspective and somebody has to supply one.

So we agree on the destination. My problem is with the road he says gets you there.

## Taste replaces credentials

This is the line, and it is the one people will repeat.

The version of the Renaissance story that gets repeated is simple. Nobody cared about your niche, they cared what you were capable of. Leonardo walked into Verrocchio's workshop to learn painting and walked out knowing anatomy, engineering, optics, and bronze casting.

Except he did not walk in.

Leonardo was born outside marriage, and the guild his father belonged to, the Arte dei Giudici e Notai, did not accept the illegitimate sons of its members. A notary was a certified witness whose trustworthiness had to be beyond question, so the guild required a man fully inside respectable society. The one career his father could have handed him was closed by rule before he could draw.

According to Vasari's well-known account, Ser Piero showed some of his son's drawings to Andrea del Verrocchio and Leonardo entered the workshop. Talent got him taken seriously. His father's standing got the drawings in front of the person who could decide.

That is not a portfolio beating a credential. That is a father with professional standing walking a portfolio into a room most people could not enter, on behalf of a son the formal system had already turned away.

The talent was real. The talent is not in question. The access was arranged.

## The apprenticeship is a credential

Here is the part I think Koe misses, and his own essay contains the evidence.

He praises the apprenticeship model. Apprentice to journeyman to master, imitating someone who knows what they are doing until your taste and judgment are formed. He is right that this is how taste actually gets made. It is how mine was made.

But an apprenticeship is a credentialing system. It has admission, it has gatekeepers, it has a hierarchy, it takes years, and it ends with a certificate. The certificate is just a reference instead of a diploma.

That is not less gated. It is less legible. A diploma at least announces itself as a barrier. You can see it, count what it costs, and argue about who it excludes. A workshop placement obtained through a father's professional network announces nothing at all. It looks like talent being recognised.

So "taste replaces credentials" is not the abolition of a gate. It is a change in what the gate is made of, and the new material is harder to see.

## I have the credentials he says are obsolete

Two diplomas. Both hard, both expensive in ways that never appear on the invoice. Blind flights, theory papers, rooms where a stranger decides whether you are what you claim to be.

I have written before that no exam ever tested the thing I actually do. Nobody graded me on reading a four top and knowing that the person ordering confidently is guessing while their guest is the one who cares. That remains true and I will keep saying it.

But here is the other half, which I have been slower to say, because it is less flattering.

Those diplomas are the reason I was in the rooms where I learned the thing the diplomas do not test.

The certificate did not teach me taste. It bought me access to the apprenticeship that produced it. People took my call. Lists were opened to me. Somebody senior decided I was worth the hours it takes to correct a palate, over and over, for a long time, and part of why they decided that is that I had already passed something.

Take the credential away and I do not arrive at the same place faster. I do not arrive at all.

## The version that is true and the version that will cost you

There is a true reading of the claim: the certificate does not measure the thing. Correct. Completely. A diploma is not a palate and a degree is not judgment, and anyone hiring on the certificate alone is hiring the wrong signal.

There is a false reading: you can skip the certificate and arrive at the thing. That one will take years off people.

Taste is not information, so you cannot download it. It is built by doing the work next to somebody who already has it, being corrected, for longer than anyone wants to hear. Its price is time and access, and access is the part nobody puts in the essay, because access is embarrassing. It is who your father knew. It is who took your call.

The Renaissance did not lack gates. It had patrons, guilds, and workshops with waiting lists. We remember the ones who got in.

## The claim I can actually test

One more, because I have run this experiment rather than admired it.

Koe says one person can now do what a team of ten to fifty people used to do.

I built and shipped an AI product on my own this year, so I can check. What it replaced was the sitting-down half of my work. Inventory reconciliation, list building, tasting notes, the same guest question answered for the fortieth time before service. A meaningful volume of that work, and I am glad it is gone.

The standing-up half costs exactly what it always has. A long service still runs on feet. Reading a table did not get cheaper. Knowing when to say nothing did not compress.

His own framework predicts this. The mechanical part gets automated and the human role moves up a layer. He just does not apply it to his own headline number. One person can now produce the mechanical output of a much larger team. One person still does the judgment of one person, and that is the half the whole argument says matters.

## The contradiction, unglamorously

He writes that the most interesting people are walking contradictions. Jacked intellectuals. Casual billionaires. Spiritual warriors.

Notice that all three are flattering.

Mine is that I reconcile inventory and read a room. I write software in the morning and walk a floor at night and the two halves rarely have anything to say to each other. Nobody is putting that on a poster.

I think that is how you tell a practice from a personality. A real contradiction is mostly inconvenient. It means two sets of skills that decay if unused, two rooms full of people who think the other room is frivolous, and no clean way to describe what you do at a party.

The glamorous version is how the idea gets sold. The unglamorous version is how it gets lived.

## What I would keep

Almost all of it. He is right that the mechanical layer is going, right that judgment is what is left, and right that the people who thrive will be the ones who can run a whole operation rather than one station of it.

I would only cut the promise that taste replaces credentials.

It does not. It replaces the paperwork.

The gate is still there. It has just stopped announcing itself, and a gate you cannot see is not an open door. It is a door that only opens for people who already know it is there.

---

# Not Drinking Is Not One Guest

URL: https://woodywusommelier.com/writing/not-drinking-is-not-one-guest/
Pillar: The Floor
Published: 2026-08-26

People are very kind to me about not drinking, and there is a specific gesture I want to describe.

It is the drink that arrives unasked. Somebody behind a bar has heard, and they build something careful, and it comes to me with a small amount of ceremony. Sometimes they stay to watch the first sip.

It is meant generously every single time. I have never once thought otherwise.

I also have not said the thing I actually think about it, because saying it would be ungracious, and because until recently I did not have the evidence to say it properly.

So I went and got the evidence.

## What I was actually checking

I sell the zero proof category. I also drink it, which is an unusual pair of positions and means I have no commercial reason to flatter it and no personal reason to sneer at it.

Most writing about non-alcoholic drinks is either supplier copy or contempt. Neither is any use to an operator deciding whether to put real money behind a program, and neither is any use to the person holding the glass.

I checked six claims against the literature and the regulations. Three of them turned out to be roughly what I expected. Two were more interesting than I expected. One stopped me.

## The parts that were roughly as advertised

The category is not a fad. Global no-alcohol volumes grew about nine percent last year while conventional alcohol moved in low single digits. That is real and it is not going away, whatever your bar manager thinks.

The sugar complaint is half right, and the mechanism is worth knowing. In normal winemaking yeast turns grape sugar into alcohol. Remove the alcohol and some of that sugar was never converted, and more gets added back on purpose, because ethanol was carrying the body and the viscosity and the impression of tannin. Take it out and the wine goes hollow. So they sweeten it to fill the hole. A typical de-alcoholized wine runs about double the sugar of a dry conventional one. The good bottles are at zero. It is a formulation choice, not an accident, and you can just ask a supplier for the number.

And the quality ceiling is real. This one changed how I think about the whole category. A 2025 review in a peer-reviewed food science journal concludes that none of the available methods can remove ethanol selectively without damaging aroma compounds, mouthfeel, and structure. All three technologies take something else with the alcohol.

That is not a problem waiting for a better machine. Ethanol is not sitting in the wine like a passenger. It is doing structural work. You cannot subtract it cleanly, the same way you cannot take the salt back out of a stock and still have the stock.

Which is the honest argument for making things designed from zero rather than wines with a hole in them. The best zero proof programs I have seen do not attempt a Burgundy substitute. They make something that was never trying to be a substitute for anything.

## The part that is quietly a problem

Alcohol-free and non-alcoholic are not synonyms. They are different legal categories.

In the United States a product can be labelled non-alcoholic or de-alcoholized while containing up to half a percent alcohol by volume. Only alcohol-free means an actual zero. Below that half percent threshold the product stops being regulated as alcohol at all and becomes an ordinary food.

For almost everybody this is trivia. The threshold exists because there is no meaningful physiological effect below it, and that is well established.

For one person at the table it is not trivia at all, and we hand them the list with the same flourish we give everybody else.

## The finding that stopped me

A 2022 review of ten studies found that cravings and the desire to drink went up after people consumed no- and low-alcohol drinks. Not down. Up. And the effect scaled with how dependent on alcohol the person was, so it was strongest in exactly the people the category is supposedly serving.

Separate work found that the taste of non-alcoholic beer lit up similar brain regions to real beer in heavy drinkers.

Clinicians routinely advise avoiding alcohol-free beer in the first year of sobriety. The highest risk profile is someone under a year in, with a previous relapse, whose drinking was socially anchored to rooms and occasions. That is not an obscure edge case. That is a description of a large number of people sitting in dining rooms being handed a zero proof menu as a courtesy.

I want to be fair about the other side, because it is substantial. A 2024 study found that nearly sixty eight percent of participants with alcohol use disorder considered non-alcoholic drinks useful in reducing how much they actually drank. As harm reduction the category clearly works for a lot of people.

Both things are true. That is the whole difficulty. The same glass helps one person cut down and pulls another back toward the thing they are trying to get away from, and nothing on the label tells you which person you are serving.

## Not drinking is not one guest

Here is the operational version, and I think it is the most useful sentence I have written about service this year.

A guest who is not drinking is at least five different guests, and our industry treats them as one.

Somebody driving. Somebody pregnant. Somebody on medication. Somebody who simply does not feel like it tonight. And somebody in recovery.

The first four are well served by an excellent zero proof list, and the better it is the better we have done. A convincing alcohol analogue is a genuine kindness to a designated driver who wanted to feel included.

The fifth guest may be the one person in the room for whom a convincing alcohol analogue is the least helpful object you could set down in front of them. Built with care. Presented with ceremony. Tasting almost exactly like the thing they are working very hard not to have.

Nobody in hospitality says this. I have not seen it in a single trade piece, a single supplier deck, a single training document. We have collectively decided that offering the fancy mocktail is the enlightened move and stopped there, because it feels good and because the alternative requires knowing something about the person that they have not offered to tell you.

## What I actually do now

I am not arguing against the category. I sell it, I drink it, and I think a serious zero proof program is one of the clearest quality signals a room can send.

I am arguing that we stopped thinking one step too early.

So: stock at least one thing that is not an imitation of anything. Not a fake gin, not a de-alcoholized red. Something built from zero that stands up on its own terms and does not resemble a drink anybody is avoiding. For some guests it is the only safe thing on the list, and they will never explain why they chose it.

Offer the category and never the reason. Do not ask why somebody is not drinking. Do not perform admiration when they tell you, which is its own way of making it the subject.

Same number of options, same specificity, same care. Fewer choices is how a guest learns they are a second-class guest, and they clock it instantly.

The full workings, including where the sources disagree, are in the Zero Proof Audit.

## The best service I have had about this

Somebody put a glass in front of me, said what it was in one sentence, and left.

No ceremony. No watching to see if I liked it. No warmth about my choices, no curiosity about my reasons, no small pause where I was expected to say something about my life.

They read what I needed underneath what I had asked for, which was not a drink. It was to be a person at a table like everybody else at the table.

That is the whole job, and it does not appear on any list of what a zero proof program should contain.

---

# I Sell Three Hours of Attention

URL: https://woodywusommelier.com/writing/i-sell-three-hours-of-attention/
Pillar: The Floor
Published: 2026-08-17

There is an argument going around that attention has become a luxury good. That the ability to sit with something long and difficult without reaching for your phone is now a class marker, and that it tracks income, and that the people with the least slack in their lives get served one-minute video instead.

I think that is broadly right and I have an unusual relationship to it, because I sell attention by the hour.

Not metaphorically. That is the product.

## What a three-hour dinner actually is

Strip the food out for a second.

A long dinner is a block of time in which you are socially permitted to be unreachable. That is the thing being purchased. Nobody questions why you did not answer a text during dinner. Nobody expects you to take the call. The obligation to be available, which is otherwise total and unrelenting, is suspended for the length of the meal by unanimous agreement of everyone who might have contacted you.

That is not a small thing. It might be the last widely recognised alibi we have.

A phone in a lockbox at home is willpower, and willpower is arguable. A three-hour reservation is an excuse other people accept. You were at dinner. That sentence closes the conversation.

So what fine dining sells, underneath the fermentation and the tweezers and the stemware, is a socially enforced permission slip to disappear. The food is the reason the permission is granted. It is not the reason people feel good afterward.

## Which makes me a merchant in the thing everyone is worried about

I want to sit in that rather than skip past it.

If uninterrupted attention is stratifying, and I run a room where you buy three protected hours for the price of a week of groceries, then I am not a bystander to that trend. I am a vendor.

The house I work in is very good at this. We are good at making the time feel like it belongs to you. We are good at the pacing that makes a person forget to check anything. Somebody trained me to do that and I have trained other people to do it, and we are all paid because we can produce a feeling of unhurried time in people whose lives contain almost none of it.

We are selling the antidote to a condition we did not cause, at a markup, to the people least likely to be suffering from it.

That is an uncomfortable sentence and I have not found a way around it that is honest.

## And then there is the part my industry does not advertise

Here is where it gets worse, and this is the thing I have not seen anyone outside the trade notice.

The long dinner is getting shorter, and we are calling it a favour to you.

The standard tasting menu in New York now runs no more than about two and a half hours. Chefs have been trimming courses for a few years, and the public reasoning is generous and guest-focused: people were flagging by the end, the later courses were not landing, enjoyment was dropping as the count went up. All of which is true.

Here is the other reason, which is also true and gets said less. A shorter menu means more covers. One chef described going from a maximum of thirty-eight guests in a night to looking at eighty to a hundred.

Read that again. Same room. Same night. Nearly triple.

And the industry-wide description of what customers want in 2026 is smaller, easier, faster, framed as a format decision rather than a discount. Which is a very elegant way of saying we are giving you less and not charging you less.

So the one commercial product that existed to sell you three uninterrupted hours has quietly become a two-hour product, for reasons that are partly about your comfort and substantially about our yield, and the marketing describes it as listening to the guest.

If attention is being stratified, some of the stratifying is being done by people like me, in rooms like mine, with a straight face.

## The best objection, answered from the floor

The strongest push-back on the whole attention-as-class argument is short and it comes from ordinary people rather than critics: *putting down the phone is free.*

It is a good objection. Willpower does not cost money. Nobody is billed for leaving a phone in another room.

But from where I stand it is answering the wrong question, and I can tell you exactly why.

In my room, putting the phone down is not free. It costs the price of the meal and a reservation made weeks ago, and what that money buys is not the willpower. It buys the alibi. It buys a socially recognised reason to be unavailable that other people will accept without resentment.

A salaried person with a door can go dark for three hours. A person whose manager texts at nine at night cannot, and no amount of willpower changes that, because the constraint was never internal. A nurse on call, a parent with a sitter watching the clock, somebody driving for an app between shifts. Their attention is not weak. It is spoken for.

That is the distinction the argument keeps missing. Attention is not primarily blocked by weakness. It is blocked by obligation, and obligation is distributed by class.

Which is why the fix everyone reaches for, discipline, reads as moralising to the people it is aimed at. You are telling someone their exhaustion is a character defect. That is how a class divide gets laundered into a personal failing, and my trade participates by selling the exemption at two hundred dollars a head.

## The part that makes me hopeful, and it is not sentimental

There is a second objection I keep seeing, from a completely different direction, and I think it is more correct than anyone is giving it credit for: *maybe the content just got worse.*

Everybody diagnoses the audience. Almost nobody audits the supply.

And on this one I have direct evidence, gathered nightly, that most people writing about attention do not have.

People will absolutely sit still for three hours. I watch them do it.

When the room is right and the pacing is right and the thing in front of them is genuinely good, phones stay in pockets without anyone being asked. Nobody institutes a rule. Nobody makes a speech about presence. The attention simply arrives, from the same people who allegedly cannot hold any, because something in front of them earned it.

And when it is not good, when the food is competent and the service is mechanical and the room has no pulse, the phones come out inside twenty minutes. Every time. That is not an attention span problem. That is a review.

I have been on both sides of that. I have run services where I lost the room and told myself afterward that people are just distracted now. They were not distracted. They were bored, and I did that.

So I do not believe attention collapsed. I believe attention got discriminating, which is a different and much more demanding situation for anyone whose job is to hold it. The bar went up while everyone was busy blaming the audience for not clearing it.

## What I actually think

Attention is stratifying, and the money follows exactly the lines you would expect, and the industry I work in is selling protected time to people who already have the most of it.

But the cure being sold is not focus. It is permission. And permission is not distributed by willpower, it is distributed by how much of your life belongs to other people.

Taste is not the currency people think it is, and neither is attention. Both are downstream of who is allowed to be unreachable, and that has never once been a personal quality. Counting is not choosing, and screen time is not character.

I can hold a room for three hours. I have done it, and it takes everything I have, and it works about as often as it does not.

When it works, nobody looks at a phone.

Nobody had to be told.

---

# Nobody Is Drinking the Champagne

URL: https://woodywusommelier.com/writing/nobody-is-drinking-the-champagne/
Pillar: The Floor
Published: 2026-08-16

Champagne is the only wine people buy for a reason that has nothing to do with taste.

Everything else on my list gets ordered because somebody wants to drink it, or wants to be seen drinking it, or is eating something it goes with. Champagne gets ordered because something happened.

A deal closed. A baby arrived. Somebody got the job, or left the job, or made it to a birthday they were not sure about. The bottle is not the point. The bottle is a receipt for an event.

Which makes it the most sensitive instrument I have. Wine sales tell me about the menu. Champagne sales tell me how people feel about their lives.

And for a while now the needle has been sitting still while it should be moving.

## The number that does not match the news

Money is being made right now. Not by everyone, and not evenly, but you can read the same stories I can, and there is a class of people in this city having a very good year.

Historically I would know that before the paper told me. It arrives as a table of six ordering something with a name on it at nine on a Thursday, loudly, badly, and I do not care that they are ordering it badly because that is not what it is for.

That is not happening at the rate it used to.

I want to be careful, because one room is not an economy and a sommelier with a hunch is not a data set. But this is the one signal I have watched most closely for the longest time, and it is quiet in a period when it has historically been loud.

So I have been asking the only useful question. Not why are people drinking less, which I have written about and which has a real answer.

Why has the ritual for *we won* gone quiet?

## What a bottle was actually for

Opening champagne is a strange act if you look at it directly.

It is loud on purpose. The cork is engineered to announce itself. You cannot do it discreetly, and nobody has ever wanted to. The entire design of the gesture is public.

It is expensive relative to what it delivers as a drink, and everybody knows that too. You are not buying refreshment. You are buying a moment that the whole room can hear.

And it is communal. You do not open a bottle for one. The format assumes an audience, which means the purchase is a small piece of theatre with a cast.

So the ritual was doing three things at once. It marked a moment. It told other people that the moment happened. And it made the moment social, which means it made it real, because a thing nobody witnessed is a thing you have to keep insisting on.

That is the machinery. Now look at what has changed around it.

## Nobody wants the room to hear

The first thing is that public celebration has become expensive in a way it was not.

Announcing that you have done well is now a risk. Somebody is having a worse year, and you know it, and they know you know. The cork is a loud noise in a quiet room. A generation that watched what happens to people who look pleased with themselves in public has learned to do well privately and mention it to four people.

I see the compressed version constantly. The table that has clearly closed something and is deliberately not making a thing of it. Good wine, ordered carefully, no announcement. The celebration happened. It just refused to be witnessed.

That is not thrift. Those tables spend. They spend on something that does not make a sound.

## Nobody is sure the win will hold

The second thing is harder and I think it matters more.

You open a bottle when something is finished. The deal closed, the baby is here, the job is yours. Champagne is a full stop.

A lot of the money being made right now does not feel like a full stop to the people making it. It feels like a position. It can be worth less by Friday, and everyone holding it knows that at all times, and you do not toast something that might unwind before the glasses are washed.

Somebody who gets a promotion opens a bottle because the promotion is a fact. Somebody whose number went up has not finished anything. They have a screen that will tell them something different tomorrow.

You cannot celebrate a thing that has not stopped moving. So people wait for the moment that will justify it, and the moment keeps not arriving, because that is the nature of the asset.

The bottle does not get opened. Not out of modesty. Out of superstition, and the sense that celebrating early is how you lose it.

## And the drinking part stopped being automatic

The third thing is mine, and I have to be careful with it, because I am not neutral.

I do not drink. I taste and I spit and I have not had a drink in over a year, and I am better at this job than I was when I drank.

Which means I have spent a long time now watching the celebration ritual from the outside, sober, at close range, and I can tell you something I did not expect.

Most of what looked like joy was momentum.

The bottle arrived because that is what you do, and the second one arrived because the first one had, and somewhere in there the evening became about the drinking rather than the thing being celebrated. I did that for years. I was not marking occasions. I was using them.

Take the momentum away, which is what has happened to a generation that drinks deliberately or not at all, and you are left with the actual question, which is much harder.

*What do we do to mark this?*

And it turns out most people have no answer, because the answer was always the bottle, and the bottle was doing the work that the ritual was supposed to do.

## The thing I am actually worried about

Not champagne sales. I will be fine and so will the category.

I am worried that we are losing the practice of marking things, and that nothing has been built to replace it.

A win that nobody witnessed is oddly difficult to believe in. That is not sentiment, it is how people work. We are bad at feeling that something is finished until we have done something physical and public and slightly ridiculous about it. That is what the loud cork was for. It was not about the wine. It was a technology for making an event real to the people it happened to.

We have quietly retired that technology, for good reasons and bad ones, and put nothing in its place. So people accumulate unmarked wins. A run of achievements that never got a full stop, which means none of them ever quite landed, which is a specific modern way to be successful and feel nothing.

I watch this at close range. The table that has clearly done something remarkable, eating well, saying little, going home early, having marked nothing.

## What I would tell you

Order the bottle.

Not the expensive one, and I mean that professionally. The grower stuff is better and cheaper and the people who farm it have been getting a fraction of what they are owed for a century.

But order it, and let it be loud, and say out loud what it is for, because the saying is the part that works.

You do not have to drink it. I do not. I will pour it, and I will make the moment happen, and I will be genuinely glad about your year while entirely sober.

The ritual was never about the alcohol. We just could not tell, because for a few hundred years the alcohol came bundled with it.

Now the bundle has come apart, and we get to find out what the ceremony was actually made of.

---

# Taste Is What They Will Let You Keep

URL: https://woodywusommelier.com/writing/taste-is-what-they-will-let-you-keep/
Pillar: The Terminal
Published: 2026-08-15

I wrote an essay a while back arguing that taste is a poor substitute for money. That the sommelier with the encyclopedic palate still does not own the cellar. That cultural capital buys you a way to be at home in rooms you cannot afford, which is a categorically different thing from owning the room, and that the people selling taste as liberation are hoping you forget there was ever a door.

I stand by all of it.

Then Elon Musk said money will not matter in ten years, and I have been turning that over for weeks, because if he is even directionally right then the essay I wrote has a sequel I do not like.

## The claim, stated fairly

His argument is not stupid and I am not going to strawman it. People want money because money is a claim on scarce goods and services. If AI plus a very large number of robots produce more goods and services than anyone can consume, the claim stops being worth anything, because the scarcity it was a claim against is gone. Work becomes optional the way gardening is optional. Governments distribute a high income rather than a basic one. Prices fall instead of rising, because output grows faster than the money supply. The binding constraints stop being capital and labour and become physics: energy, mass, wattage, tonnage.

Ten years, roughly. Bumpy road first.

I have no standing to argue the macroeconomics and I am not going to pretend otherwise. I do not know whether the robots arrive on schedule. Nobody does, including him.

What I do have is this: I already work inside a post-scarcity economy. It is called a restaurant, and it has been running this experiment for about a century.

## My industry already stopped charging for the stuff

Food cost in a serious room runs twenty-five to thirty percent. The physical inputs on a plate are a minority of what you pay, and everybody in the building knows it.

Nobody has ever paid two hundred dollars for a tasting menu because the carrots were scarce. Nobody orders a bottle because glass and cork and fermented grape juice are hard to come by. The material inputs are cheap, abundant, and not the product.

So my trade solved, a long time ago and at enormous cost, the exact question the abundance people are now asking: what do you charge for when the stuff is basically free?

The answer we arrived at is judgment, curation, attention, and presence. Somebody chose these grapes. Somebody decided this goes with that. Somebody is reading your table right now and adjusting. That is the product. That is what the number on the bill is actually for.

Which means I am not speculating about a taste economy. I have worked in one my whole career, and I can tell you what it is like to live in, and it is not the thing people are imagining.

## Taste stratifies harder than money does

Here is the first thing, and it is the one nobody says because it sounds ungenerous.

Money is transferable. Taste is not.

You can give your child money. You can give them a building, a portfolio, a trust. It is crude and effective and everyone can see it happening, which is precisely why we have arguments about inheritance and estate tax and whether any of it is fair.

You cannot give your child a palate.

What you can do is take them places. Feed them things. Let them sit at tables where adults discuss what they are eating as though it matters. Pay for the trips, the meals, the years of low-stakes exposure that build the thing quietly, before anyone thinks to call it an advantage.

By the time that child is twenty-five they have a palate, and it looks like a personal quality. It looks earned. It looks like *them*. Nobody points at it and says that was purchased, because the purchase happened in a thousand invisible increments over two decades, and the receipt is a childhood.

Inherited money is at least legible enough to argue about. Inherited taste presents as merit.

If taste becomes the currency, we have not abolished the aristocracy. We have made it invisible and given it a story about itself.

## The price is time, and time is gated

The second thing.

A palate is not information. I could hand you every fact about a region and you would not be able to taste blind. The thing is built by tasting thousands of things over years, ideally next to somebody who already has one, telling you what you are tasting while you are tasting it, correcting you, again, for a long time.

That is an apprenticeship. Apprenticeships are gated by who is willing to have you, which is gated by who you already know, which is gated by where you started.

Money you can win, steal, get lucky with, inherit sideways. Taste has one acquisition path and it runs through access to people who already have it.

I got mine because rooms let me in. I would like to say it was appetite and work, and it was both, but the honest version includes a list of people who did not have to give me the time and did.

## And then it does the thing wine did

The third thing, and this is the one I am actually afraid of, because I watched it happen to my own industry.

When taste becomes the currency, taste gets a scoreboard.

Wine ran this experiment in full. We had a domain where quality was genuinely unmeasurable and judgment genuinely mattered, and within a generation we produced a hundred-point scale, a gatekeeping class, and a customer base so anxious about being wrong that the best-selling wine on almost every list on earth is the second-cheapest bottle, ordered by people who do not want to be seen ordering the cheapest.

That is what a taste economy feels like from the inside for most people. Not liberation. Not the pursuit of the beautiful. A test you did not study for, administered constantly, in public, by strangers, with your status as the grade.

Counting is not choosing, and I have spent years arguing that the score is not the wine. But I understand exactly why the score exists. It exists because a world where taste is the currency is unbearable without one, and people will invent a number to hide behind rather than stand there with a preference they have to own.

## The new poverty

So put it together.

In the abundance scenario, the material floor rises for nearly everyone. Good. Genuinely good. I am not going to be precious about a world with less hunger in it.

But the ownership of the machines and the wattage stays exactly where it is, and probably concentrates, because it always does. And everybody else is handed the one differentiator left.

Taste becomes the game because it is the only game they left on the table.

And the new poverty is not hunger. It is standing in front of an infinite menu with no idea what is good, knowing that everyone can see you not knowing, and that no transfer payment fixes it, and that the fix takes twenty years you did not get.

That is a poverty with shame built into it. Material poverty at least has the dignity of being obviously not your fault. Taste poverty presents as a personal failing, because taste presents as a personal quality, because we agreed to pretend the childhood was not a purchase.

I watch a version of this every single night. It is the most familiar thing in my working life. A person sits down, opens the list, and I can see the exact moment they decide they are about to be judged.

## What I would keep

My old essay ended by telling you to keep your eyes on the door that the knowing is supposed to open. I still think that.

But if the door stops mattering, if the room really does become free, then I want to be honest about what taste is in that world.

It is not the new money. It is what they will let you keep.

And the job I do, which is to take the list out of somebody's hands and tell them the truth about what they would actually enjoy, and watch their shoulders come down, turns out not to be a luxury service at all.

It is the only known cure for a poverty we are about to manufacture at scale.

Somebody who reads what you want under what you are asking for, and tells you it is allowed.

I would not automate that one. I have had the chance.

---

# Nobody Needed a Sommelier When Everybody Drank

URL: https://woodywusommelier.com/writing/nobody-needed-a-sommelier-when-everybody-drank/
Pillar: The Cellar
Published: 2026-08-14

Somebody asks me every few weeks now whether my job is going away.

They ask it kindly. It is the tone you use with a friend whose industry has been in the paper.

The premise is that drinking is collapsing, which it is, and that a profession built on wine therefore has a decade left, which it does not. Both halves get repeated so often that nobody checks them, so I went and checked them, because it is my job and my income and I would like to know.

The first half is true and worse than people say. The second half is wrong in a way that I find genuinely funny, and then, the more I sit with it, genuinely uncomfortable.

## The kids did not do it

Americans are drinking at the lowest rate in ninety years of measurement. Fifty-four percent of adults. Under thirty-five fell from fifty-nine percent to fifty percent in two years, which is not a trend, that is a cliff.

Everyone has decided this is a generational story. Gen Z killed drinking, the way they killed napkins and doorbells and whatever else got written up last week.

Except the industry's own data says the opposite. Gen Z participation is up, from sixty-six percent three years ago to seventy-four percent now, which is nearly the seventy-six percent adult average. The generation with the lowest drinking rate is Boomers, at seventy-one percent, and falling.

The kids did not quit. Their parents did.

What actually collapsed is not who drinks. It is how often. Daily drinking among the young went from six percent to two percent. A third of what it was. And the share drinking nothing at all went from seventeen to twenty-four percent.

So the honest sentence is not "nobody is drinking." The honest sentence is: habitual drinking is dying, and occasional drinking is fine.

Which changes everything about the question, because habitual drinking was never my customer.

## Nobody needed a sommelier when everybody drank

Think about who the daily drinker actually was. A glass with dinner, most nights, usually the same thing, chosen once years ago and never revisited. That person did not want advice. That person had a habit, and a habit is a decision you already made.

I could tell that man about a grower in the Northern Rhône who farms three hectares by hand and he would nod politely and order what he ordered last time, because he was not asking a question. He was performing a routine.

Now take the person the data actually describes. Drinks twice a month. Made it a choice rather than a reflex. Turned down four other opportunities to drink and picked this one.

That person has enormous stakes in getting it right, and no accumulated experience to draw on, because they stopped drinking often enough to build any.

That is the best customer a sommelier has ever had. That is a person who wants to be read.

The volume did not go to my competitors. It went to the part of the business that never needed me.

## The part nobody in my trade will say out loud

Here is where it stops being funny.

Wine carries something like seventy percent margin in a business that runs on three to five. A restaurant does not keep a sommelier because the guests demand one. A restaurant keeps a sommelier because the wine number holds up the whole building.

Which means the honest description of my position, for most of its history, is this: I was a revenue instrument that also knew things.

Volume was not justifying the job. Volume was subsidizing it. And when the daily drinker disappears, the subsidy goes, and the restaurants that were carrying a somm on the strength of the beverage line start doing the arithmetic. That is not speculation. Fewer rooms are hiring for the position right now.

So when someone asks whether my job is dying, the accurate answer is: the *funding* is dying. The job is fine. Those have never been the same thing and my trade has spent a long time not noticing.

## The parts I can do sitting down

There is a second pressure arriving at exactly the same moment, and everyone assumes it is the bigger one.

It is not. Nobody is automating twenty thousand steps. Nobody is building a machine that walks a room, clocks the table that has gone quiet, and gets there before the silence turns into a complaint. I am not worried about that and neither should anyone who actually does this work be.

But something did get taken, and I know precisely what, because I built the thing that took it.

Here is the test that has been sitting in my head for a year.

The parts of my job I can do sitting down are the parts that are going away.

Sitting down: inventory reconciliation, ordering, list construction, tasting notes, allergen sheets, training documents, pairing suggestions written for a website at midnight. All of it, gone or going, and I shipped a product that does a chunk of it on purpose.

Standing up: reading the four top, knowing when to say nothing, judging that the man ordering confidently is guessing and his guest is the one who actually cares. None of that moves.

I automated the half of my own job that could be automated. It did not make me smaller. It made the other half the entire job, which is a much more exposing place to stand.

## Two knives, one direction

So there are two pressures, arriving together, from completely unrelated causes, and nobody has connected them because the AI people do not work floors and the wine people do not build software.

Declining drinking is stripping away the volume that paid for the position.

AI is stripping away the knowledge that justified the position.

They are cutting in the same direction. What is left after both is a body in a room exercising judgment, which is the only part that was ever actually hard and the only part anyone should ever have paid for.

## The guest almost nobody can serve

And here is the thing that makes me think the profession is not dying at all, it is being handed its hardest exam.

Twenty-four percent of Gen Z drinks nothing. The non-alcoholic category is growing around nine percent a year while alcohol crawls in low single digits. That is not a smaller room. It is a different room.

Serving the guest who is not drinking, properly, with the same care and the same number of options and the same seriousness you would give a drinker, is the hardest thing on the floor and almost nobody can do it. It is pure judgment. There is no bottle to hide behind, no score to cite, no vintage to talk about. A guest says "I am not drinking tonight" and you have about two seconds to understand that they have made a statement, not asked a question, and that pushing anything at them now is the exact opposite of service.

When I built the product, that was the behaviour that took longest and broke most often. Not the wine knowledge. The restraint.

Most sommeliers never developed that skill because the room never required it. Everyone drank, so you could be excellent at your job and never once be tested on the guest who does not.

That room is emptying. The other one is filling.

## What is actually happening

Nobody is going extinct.

The job is being concentrated into the part that was always hardest, and the part that was always hardest is now the whole thing, with the volume that used to pad the number gone and the knowledge that used to pad the credibility free to everyone.

Most of my colleagues will not survive that, and I want to be careful here, because it will be tempting to blame sober twenty-five-year-olds and language models. Neither did this.

The profession spent decades getting paid for the easy half. Two unrelated things arrived at once and took the easy half away.

That is not an extinction. That is a bill.

---

# Six Figures Orders the Second-Cheapest Bottle

URL: https://woodywusommelier.com/writing/six-figures-orders-the-second-cheapest-bottle/
Pillar: The Floor
Published: 2026-08-13

The second-cheapest bottle on my list is the most-ordered wine in the room.

It is on nearly every list in the world, and on nearly every one of them it outsells the bottles on either side of it. Not because it is better. Because of where it sits.

Nobody wants to be the person who ordered the cheapest thing. So they move up one, and they stop.

I have watched that happen thousands of times, mostly on first dates, and I have never once seen anyone look happy doing it. There is a particular pause before it. The list comes back to the same page twice. A finger lands, lifts, lands again one line down. Then a slightly-too-casual voice says we will do this one.

That order is not about wine. It never was. It is a person deciding what they are willing to be seen wanting, in front of somebody whose opinion they have not earned yet.

## The pause is the whole story

I bring this up because there is a conversation happening right now about why people are not dating, not partnering, not sleeping together. Dating apps consolidated attention. A small number of men get most of it. The bar is six feet and six figures. Depending on who is talking, this is either proof that women are mercenary or proof that the market is broken.

I have no standing in the market argument and I am not going to pretend otherwise. I do not study this. I stand in it.

What I can tell you is that the room where this is supposed to happen is my room, and I have a view of it that people writing about mating markets do not have, because they are describing a thing I am physically inside of five nights a week while sober and holding a wine list.

And from here, six feet and six figures do not look like preferences.

They look like the second-cheapest bottle.

## What a number is actually for

Height and income have one property in common that nothing else on the list has. You can verify them from a profile, before you have met anyone, without trusting yourself.

That is the entire appeal. Not that people want them. That they can be *checked*.

Everything that actually predicts whether you will be happy with somebody is unmeasurable in advance and requires you to be in a room with them, paying attention, forming a view you might be wrong about. Whether they are kind when nobody is grading them. Whether they get funnier or smaller when they are tired. Whether being around them makes you more yourself or less. None of that fits in a field. All of it requires judgment, and judgment can be wrong, and being wrong in public is the thing everyone is organising their life to avoid.

So people reach for the number. Not because the number is good. Because the number is defensible.

Six feet is a criterion you can state out loud to your friends without embarrassment. "He made me feel like the most interesting person at the table" is not a criterion. It is a judgment, and you own it, and if it turns out badly you cannot point at the spreadsheet.

My own trade did exactly this to itself. Wine went numerical, because a ninety-four-point score is a thing you can hold up and hide behind, and "I liked it" is a thing you have to defend. The scores did not make anyone's palate better. They gave people something to point at instead of having one. Counting is not choosing. It never was.

The dating market did not invent this. It just ran the same play on people.

## Nobody removed the romance. They removed the sommelier.

Here is the thing I actually believe, and it is not the thing either side of this argument is saying.

The apps did not break mating by making it a market. Mating was always a market and everybody involved always knew it. What the apps did was hand every person on earth a list with four thousand bottles on it and no one standing there to read them.

That is my literal job. A guest sits down in front of a list they cannot parse, in a situation where they feel evaluated, and they are about to order defensively. They will take the second-cheapest, or the only name they recognise, or whatever they think signals correctly. All three are the same move: outsourcing a decision they do not trust themselves to make.

What I do is not tell them what is good. Anybody can do that, and now a machine can do it faster than I can. What I do is read what they actually want underneath what they are asking for, and then give them permission to want it.

Sometimes that means selling down. A man is reaching for something expensive to perform provision for someone who does not care about the price and would rather he relaxed. The right move is to put my finger on something cheaper and better and say, honestly, that this is the one I would drink tonight. He gets to look like a man with taste instead of a man with a budget. She gets a better bottle. I make less money.

That is the job. Not the knowledge. The permission.

Nobody is doing that for anyone on an app. There is no one in the loop whose entire function is to look at you and say: the thing you are performing is not the thing you want, and the thing you want is allowed.

## What we actually lost

The conversation about all this keeps looking for a villain. Porn. Apps. Inequality. Phones. Somebody's politics.

From the floor it looks less like a villain and more like a demolition.

Courtship used to be a script. Not a good one, and not one I would want returned intact, but a script: where you sat, who ordered, who paid, how you signalled interest without having to say it, how you declined without having to say that either. The rituals were not the point. The rituals were the scaffolding that let two frightened people get through an evening without having to invent every move from nothing.

We took the scaffolding down. We replaced the finding, efficiently, and we replaced none of the doing.

So people arrive at my tables with the hardest possible task. Be interesting, be unguarded, be discerning, be relaxed, and improvise all of it from scratch, in front of a stranger who is deciding about you, with no inherited moves to lean on.

And then, increasingly, without a drink.

## The part I did not expect

Alcohol was doing a lot of quiet work in that script. Not romance. Anaesthetic. It let people be slightly braver than they are and gave them something to do with their hands. Drinking is down, it is down hardest among the young, and daily drinking has all but disappeared among people under thirty.

I am the last person who will argue for putting that back. I do this job without drinking and I am better at it than I was when I drank. Every night I watch what alcohol was actually covering for, and it was rarely the good part.

But I would be lying if I said nothing was lost. What alcohol bought people was permission, badly and at a cost. Take it away without replacing it and you get a room full of people who are more sober, more careful, more legible to themselves, and much more afraid.

Sober courage is real courage. That is the whole problem with it.

## The scarce thing

So I do not think dating is dying because women want six feet or because men want too much or because a market got efficient.

I think an entire generation was handed unlimited options and no one to help them choose, and choosing badly in public became the most expensive thing you can do, and so people are doing the thing anyone does when the stakes rise and the confidence falls.

They order the second-cheapest bottle. They pick the criterion that survives a friend's cross-examination. They wait for something to be obviously correct before they will risk wanting it.

I cannot fix that from where I stand. What I can do, on a Tuesday, at one table, is take a list out of somebody's hands and tell them the truth about what they would enjoy, and watch a person relax about a decision for the first time that evening.

It is a small thing and it is the entire job, and it turns out to be the scarcest skill in the economy right now, in a hundred contexts that have nothing to do with wine.

Somebody who reads what you want under what you are asking for, and tells you it is allowed.

We used to have those everywhere. Now you have to book a table.

---

# Typing Is the Cheap Part

URL: https://woodywusommelier.com/writing/typing-is-the-cheap-part/
Pillar: The Terminal
Published: 2026-08-12

Anthropic announced this month that Claude will start hiding a watermark in everything it writes. New models carry it from the second of August. It runs worldwide, not just in Europe, though Europe is what caused it.

I am not going to explain how it works. Matthew Purcell has already done that, properly, by deduction from public documents, and his piece is better than the one I would have written. The short version is that the model quietly tilts its own word choices in a pattern you cannot see and only a key holder can read.

Everyone is arguing about whether it will work.

Wrong argument. My trade settled this in 1935 and the answer was not the one anybody wanted.

## An appellation for sentences

That is what this is. Not a security feature. An appellation.

France built the appellation system because people were putting famous names on ordinary wine and the famous names were starting to mean nothing. A regulator drew lines on a map and said: this word means this place, farmed this way, or you do not get to use it.

Ninety years later the same regulatory culture looked at a paragraph and had the identical instinct. Draw a boundary. Mark the inside. Protect the word.

I find the symmetry funny and I do not think it is a coincidence. Europe has exactly one move for a category losing its meaning, and it has now run that move on Burgundy and on sentences.

## Who the label was actually for

Here is the part that gets left out of the nice version of this story, and it is the reason I do not find any of this reassuring.

The appellation system was not built to protect the person drinking the wine.

It was built by producers, for producers, to defend the commercial value of names they owned. Fraud was diluting those names and the money attached to them. The drinker benefited, sometimes, incidentally. But nobody drew those boundaries out of concern for a man in Lyon who might get a disappointing bottle. They drew them because the word Champagne was worth defending and somebody had to own the defending.

Provenance systems protect the value of the marked name. That is what they are for. Everything else is downstream.

So when a model provider marks its output, I do not primarily read consumer protection. I read a company establishing that its name means something specific, and building the instrument that decides who gets to say so. The law required it, the company complied faster and wider than the law demanded, and the result is that Anthropic now owns a cryptographic key that answers a question about your writing.

I am not accusing anyone of bad faith. Doing this is more responsible than not doing it. But it is worth being clear-eyed about the shape of the thing, because my trade spent a century learning that whoever certifies has the power, and that power gets used well most of the time and badly the rest of the time, and the arguments never end.

## The forger's labels were perfect

The most accomplished wine counterfeiter of the modern era was not caught by his labels.

His labels were flawless. Right paper, right ink, right corks, right capsules, right wax. Experts held those bottles and saw nothing wrong, because there was nothing wrong to see.

He was caught because he sold a domaine and vintage combination that never existed. The wine had not been made that year. No amount of craft survives a fact.

The mark was perfect and the mark was worthless, because a mark on an object is not provenance. Provenance is a chain of hands. The label is one link, and it is the link most easily faked, which is why the trade stopped treating it as the answer somewhere around the time it lost a fortune treating it as the answer.

A watermark is a label. A very good one. It is still a label.

## Whose hands, whose wine

Now the detail I have not stopped turning over, which Purcell caught and which almost nobody is discussing.

The watermark is applied at the model level. It does not know why it was called.

Ask it to fix the commas in a paragraph you wrote yourself, and your paragraph now carries the mark. Your idea. Your sentence. Your argument that you had in the shower. Somebody else's fingerprint on it, permanently, because a machine cleaned up your punctuation.

We have this argument in wine and it has been running for thirty years. When a consultant advises a domaine on the blend, and the same few consultants advise hundreds of estates across a dozen countries, whose wine is in the bottle? The fruit is yours. The land is yours. The vintage happened to you. The decisions might belong to a man who flew in for two days.

No appellation law has ever required anyone to put that on the label. We argued, we mostly agreed it matters, and we still do not disclose it.

So the machine version arrives with the asymmetry backwards. The commas get marked. The essay a person wrote alone, after reading four things a machine wrote, carries nothing at all.

One kind of influence leaves a fingerprint. The other kind, the kind that actually changes what people think, leaves no trace whatsoever.

We are building forensics for the typing at the exact moment the typing stopped being the interesting part.

## I have been authenticated before

This is where I should admit why I care more than I probably should.

I have spent my working life being certified. Two diplomas, both of them hard, both of them expensive in the ways that do not appear on the invoice. Blind flights. Theory papers. Rooms where a stranger decides whether you are what you say you are.

I am glad I did them. They mark me. They open doors and they are the reason certain people take my call.

They also never once tested the thing I actually do for a living.

No exam has ever asked me to look at a four top and know that the man who ordered confidently is guessing, that his guest is the one who actually cares, and that the right move is to sell down and let him look generous. Nobody graded me on knowing when to say nothing. There is no certificate for the night I poured something objectively brilliant for a table that needed something honest and cheap, and watched a good evening go slightly cold because I was admiring my own conviction instead of reading the room.

The wine was right. I was wrong. Those are not the same sentence, and the diploma on my wall has nothing to say about the difference.

That is what a mark does. It certifies the thing that can be tested. Then everybody starts believing the tested thing is the thing that matters, because it is the part with a number attached, and the part with a number attached always wins the argument. Counting is not choosing. It never was.

## The part no exam has ever tested

I taste and spit now. I have not had a drink in over a year, which means I do the job on memory, structure and discipline rather than on pleasure, and I am better at it than I was when I drank.

There is no mark for that either. No watermark, no certificate, no provenance document. You would have to actually watch me work, or trust somebody who has.

A watermark can tell you a machine's hands were on the words. It cannot tell you whether anyone thought. It cannot tell you if the argument holds, if the writer understood what they were saying, if a single decision in the piece was made by a person who would defend it.

We are about to spend enormous effort answering who typed.

Nobody is building the instrument for who decided, because there is not one, because judgment does not leave residue. It shows up as a thing that worked, in a room, on a night, and then it is gone.

That is the whole job. It was always the whole job.

Typing was always the cheap part.

---

# Counting Is Not Choosing

URL: https://woodywusommelier.com/writing/counting-is-not-choosing/
Pillar: The Cellar
Published: 2026-08-06

Every wine program has a night nobody writes about.

It is the last Sunday of the month, the room is dark, and someone is on a stepladder with a clipboard counting bottles. Not tasting. Not selling. Counting. Then weighing the open ones, because a bottle that is one third gone still has to be a number by morning. Then matching that number against what the till says you poured, and finding the gap, and not knowing whether the gap is theft or breakage or a heavy hand on a Tuesday.

I have done that count more times than I can tell you, and I have never once enjoyed it.

So when people ask whether AI is coming for wine, I want to be clear that I have been actively hoping so, specifically about this, for a long time.

The good news is that it arrived. The interesting news is what it took.

## What is actually out there

If you run a beverage program and the counting is the thing killing you, that problem has been solved by several companies, and you should probably buy one of them rather than think about it further.

There are tools built specifically for bar and beverage inventory that scan invoices, sync depletion against your point of sale, and weigh partial bottles on precision scales so the open Chablis stops being a guess. Published pricing starts around two hundred and fifty dollars a month, which is less than a case of anything you would be proud to pour.

There are broader platforms aimed at groups and multi-site operations that do predictive ordering and forecast demand across locations. There are venue-scale systems that pull point of sale, inventory, events and even weather into a single dashboard for stadiums and large hotels. And this spring, one of the major point of sale companies made ingredient-level inventory intelligence a paid add-on inside the software a great many independent restaurants already run, which means the capability can now arrive through a stack the operator has already chosen.

That last one is the real story of the category. The winner in restaurant software is rarely the best product. It is whichever thing is already installed.

I am not going to rank them here. Your program is different from mine, the prices move, and anyone who tells you which inventory tool is correct without asking what your pour cost looks like is selling.

But the shape of the category is worth noticing, because it is the same shape every time.

## They all count. None of them choose.

Look at what these systems actually do. They count stock. They read invoices. They cost recipes. They flag variance. They predict what you will need on a Friday in November.

Every one of those is a mechanical task with a correct answer. Somebody could always have done it with enough hours and a spreadsheet, and the machine is simply faster and does not get bored at bottle three hundred.

Now look at what none of them do.

Not one of them decides which growers belong on your list this year. Not one of them tastes a wine and knows it is technically fine and completely wrong for your room. Not one of them looks at a four top who said they are not really wine people and understands that the sentence means something other than what it says.

The entire category optimises the cellar. Nobody built anything for the floor.

I do not think that is an oversight or a market gap somebody will fill next quarter. I think it is the pattern, and the pattern is that automation always takes the countable half first, because the countable half is the half you can check.

That split is the whole basis of the triage tool I give away: describe a task, and the first thing it asks is how much live judgment it needs. High judgment plus a guest who sees the mistake is the work you are actually paid for, and it should be protected from automation rather than queued for it.

You can verify a bottle count. You cannot verify a bottle choice, not in any way a machine can learn from, because the same wine is right at one table and wrong at the next and the only evidence either way is a face.

## The part that should reassure you, and the part that should not

The reassuring part: the work being automated is the work you wanted gone. I have never met a wine director who got into this to count. The stepladder Sunday was always the tax you paid for the part you loved.

The part that should not reassure you: everything mechanical in your job is now on the same conveyor. If a task has a correct answer that can be checked against a record, assume it goes. Not this year necessarily, but assume it.

Which leaves the obvious question of what is left, and the answer is the thing you probably undercharge for. Choosing. Reading. Knowing when the honest move is to talk somebody down to the cheaper bottle because they will enjoy it more and come back sooner.

I once poured a genuinely great wine for a table that wanted an ordinary one, because I was more interested in my own list than in the four people at the table. The wine was flawless. The night went quiet. No inventory system in the world would have flagged that as a loss, and it was the most expensive mistake I made that month.

## What to do on Monday

If counting is your pain, buy something. It is a solved, competitive, commoditised problem and paying a consultant to think about it is setting money on fire.

Check whether your point of sale already offers it, because there is a real chance the capability is one add-on away.

And then spend the hours you get back on the half nobody is selling software for. Taste more. Train the floor harder. Write better notes on the list. Learn which of your regulars wants to be surprised and which one wants the same Chablis every single time and would be quietly hurt if you suggested otherwise.

The machines took the clipboard. They were welcome to it.

They have not taken the stepladder yet, but only because the good part of this job never happened up there.

---

# Being Right Is the Cheap Part

URL: https://woodywusommelier.com/writing/being-right-is-the-cheap-part/
Pillar: The Terminal
Published: 2026-07-30

This summer, Leopold Aschenbrenner's fund sold its public-equities book in a hurry. After a brutal few weeks, the number that had only ever gone up went down all at once, the way those numbers always do.

I am not going to tell you what to think about the markets. I do not know, and neither, it turns out, did the people whose entire job it was.

Here is the part that stayed with me.

Aschenbrenner was not a fool who got lucky. He wrote the essay that gave the moment its name, in public, before most people were paying attention. He argued it clearly and then put real money behind a conviction that has, in the large, held up. The technology is real. He said so early, and loudly. He called the fund Situational Awareness, after the essay that made his name.

It did not save him.

That is not a story about AI. It is the oldest story I know, and I watch a smaller version of it play out on a floor most nights.

## Knowing and doing are different jobs

There is a difference between knowing a thing is great and knowing what to do about it. The gap between those two is where most of the money, and most of the dignity, in my trade gets lost.

I can hand you a wine that is by every measure extraordinary. Correct provenance, singular site, a maker who does everything right. Identifying that bottle is table stakes. Any competent somm can do it, and a machine can now do it faster than I can.

Knowing that this extraordinary bottle is exactly wrong for the table in front of me, for the food they ordered and the night they are having and the number on the right side of the list they are quietly hoping I respect, that is the job. It is the only part that was ever hard. It is the only part anyone should pay for.

I have gotten it wrong. I have poured something objectively brilliant for a table that needed something honest and cheap, and watched a good night go slightly cold because I was busy admiring my own conviction instead of reading the room. The wine was right. I was wrong. Those are not the same sentence, and learning the difference cost me more than any exam ever did.

## Everyone is right about the technology now

That is the cheap part. The AI is real, it is powerful, it is not the empty bubble the loudest skeptics keep promising. Fine. Agreed. You and I and the fund that just came apart all agree.

None of that tells you what to do on Tuesday.

It does not tell you how much to bet, or when, or with how much borrowed against it. It does not tell you which of your restaurant's problems is worth automating and which one will quietly break the moment you touch it. It does not tell you whether the workflow that saved another operator a fortune will save you a dollar, because your business is different in ways that never make it into the pitch.

One widely cited study found that roughly ninety five percent of enterprise generative AI pilots showed no measurable impact on profit and loss. Almost none failed because someone was wrong about AI. They failed because someone was right about AI and mistook that for being right about their own situation.

I watch operators do the smaller version of this constantly. A sharp one reads that AI is real, which it is, and concludes he should therefore put it everywhere in his restaurant, which does not follow at all. He is right about the technology and wrong about his own floor, and the tool ends up automating the one warm thing his regulars actually came back for. Right about AI. Wrong about the room. The bill arrives either way.

## Conviction is not judgment

The smartest bull in the room was right about the destination and undone by the path. Maybe the timing, maybe the borrowing, maybe both. It does not matter which, because any one of them is fatal on its own, and being correct about the big picture does not buy back a single one.

This is the thing I keep saying and the world keeps insisting on proving. Conviction is loud and cheap and everywhere right now. Judgment is quiet. Judgment is knowing which grape to leave alone, which table wants which bottle, which of your problems you must never automate no matter how real the technology gets.

The technology was never the hard part.

I did not need a language model to learn that. I needed a floor, and a few wrong pours in front of people I badly wanted to impress.

---

# The Real Labor Math of AI in a Restaurant

URL: https://woodywusommelier.com/writing/the-real-labor-math-of-ai-in-a-restaurant/
Pillar: The Terminal
Published: 2026-07-27

Your best manager is probably the most expensive data entry clerk you employ. Not by choice. By default.

That much I am confident about, because I have watched it across my career and lived most of it. What I am no longer confident about is any specific number attached to it.

Here is a thing that happens if you read about restaurant technology. Someone tells you the average general manager loses eighteen to twenty one hours a week to administrative work. Someone else says fifteen to twenty of those go to inventory alone. A vendor tells you their tool returns a third of a manager's week. The numbers are specific, they are confident, and they have decimal points.

I used to quote figures like these. I stopped. This essay is about why, and about what I think you should ask for instead.

## Where the numbers come from

Almost every labour statistic in restaurant technology is one of three things.

It is a survey, in which busy people estimated their own hours from memory, which is a genre of fiction. It is a vendor's own study, which is marketing wearing a lab coat. Or it is a real number from a real business that has been generalised into an industry average, which quietly assumes your restaurant is like the one that produced it.

None of those are lies exactly. They are just not measurements of your building.

Here is the part that should bother you more. The manager whose hours are being counted is the same person being asked to estimate them, and nobody in hospitality has ever accurately reported how long anything took. Ask a chef how long prep ran. Ask a GM how much of Sunday went to the schedule. The honest answer is always "more than I want to say."

## The number that actually matters

Around ninety five percent of AI projects produce no measurable return. That figure I do trust, because it keeps being reproduced by people with no product to sell, and because it matches everything I have seen.

But sit with what it implies. If almost all of these projects show no measurable return, then almost all of the confident hours-saved numbers in the market are describing projects that, measured properly, did nothing.

The failure is rarely the model. It is that someone automated the wrong thing, and then measured the wrong thing to prove it worked.

If you want to find out which of your own tasks is the wrong thing, I built a free diagnostic that will tell you. It returns do not automate as a real answer, which is the verdict that saves the most money and the one nobody selling you software will give you.

## What I built instead

This year I built an AI system and put it into a working service context, and the part that took longest was not the intelligence. It was deciding what I would allow myself to claim about it.

So the measurement layer I built reports three things.

First, whether it held the standard. Not usage. A thousand conversations that resolved nothing is a cost, not a saving, and usage dashboards are the most common way a tool proves nothing while looking busy.

Second, what it plausibly saved, expressed in one specific form: inquiries resolved without staff time, multiplied by a handling time the operator measured themselves. Not an industry benchmark. Their number, from their building, which means they can argue with it. If they think three and a half minutes is wrong, we change it in front of them and the model still stands. A number a client can argue with is a number a client can believe.

Third, and this is the section I care most about, what it refuses to claim.

It does not claim it lifted revenue. Without a control period you cannot prove a tool caused that, and guest mix, season, menu and weather all move the same line. It does not claim it saved covers. It counts what it can stand behind, says so in writing, and leaves the rest alone.

I want to be precise here, since the whole essay is about precision. That system is a demonstration running against a fictional venue. It has no guests, so it has no results yet. What I can tell you about is the method, not the outcome, and I would rather hand you a method you can use than a number I would have to invent.

## Desk work and floor work

None of this means the opportunity is not real. It is. But the honest split was never smart jobs against dumb jobs. It is desk against floor.

Everything that happens at a desk is genuinely in play. Invoicing. Inventory reconciliation. Pulling guest history together before service. First draft schedules. Review responses. The allergen sheet. The marketing copy that eats a Sunday. That work is real, it is expensive, and almost none of it is why anyone got into this business.

And then there is the floor, where none of it works.

The read of a table at eight o'clock. The recovery that turns a ruined night into a regular. The decision to comp, to push, to slow a table down, to leave someone alone. The two in the morning conversation with a cook whose life is coming apart. None of that automates, because none of it is information. It is presence, and presence is the job. A machine can draft the schedule. It cannot stand in a dining room and feel the weather change.

## The re-concentration

So AI does not shrink the restaurant job. It re-concentrates it.

The desk work compresses toward zero and what remains is the part that was always the actual profession. The operators who should be nervous are the ones whose week was secretly all sediment. The ones who should be glad are the ones who came here for the room and have been drowning in paperwork ever since.

The reclaimed hours are not a cost saving to quietly pocket, either. That is the mistake I would most want to talk an operator out of. Take the hours back and spend them where a machine will never go: the pre shift, the new server who needs coaching, the recovery, hiring well instead of fast, standing in the room and reading it. That is the return. Not a smaller team. A team finally pointed at the work worth doing.

## What to ask for

If someone is selling you AI for a restaurant this year, you now have a sharper question than "how good is your model."

Ask what handling time they are multiplying by, and where that number came from. If it is an industry benchmark rather than something measured in your building, it is a guess with a decimal point.

Then ask what their reporting refuses to claim. Any honest measurement has a section on what it cannot prove. A vendor who shows you only numbers that go up is not measuring. They are marketing, and the difference will cost you a year.

The machines are not coming for your floor. They are coming for the paperwork that has been keeping you off it.

Give your managers their hours back. Then send them to the room, and count it honestly.

Woody

Operators: how many hours a week do you think you lose to admin, and have you ever actually measured it rather than estimated? Reply and tell me. I read everything.

---

# The Cellar Went Open Source

URL: https://woodywusommelier.com/writing/the-cellar-went-open-source/
Pillar: The Terminal
Published: 2026-07-24

There is a bottle I pour by the glass that costs the room nine dollars and pours for twenty six. Nobody objects. They are not paying for the liquid. They are paying for the fact that someone drove to find it, tasted a hundred others to arrive at it, and stands behind it when it lands on the table. The wine is the cheap part. It was always the cheap part.

I thought about that bottle last week, because the most expensive thing in artificial intelligence just became free.

On July 16 a Chinese lab called Moonshot released a model named Kimi K3. Two point eight trillion parameters, the largest open weight model anyone has put into the world, with the full weights promised by the end of this month. Open weight means you can download it. Not rent it, not call it through somebody's meter. Download it, run it on your own hardware, and owe no one anything.

The instinct is to file this under geopolitics and move on. Resist that instinct for a minute, because there is something in it for anyone who runs a room, sells a bottle, or has ever been told a machine is coming for their job.

## Third in the world, and free

Let me be precise, because precision is the whole point of what follows.

On GDPval-AA v2, a benchmark that scores real work across forty four occupations and nine industries, K3 scored 1,687. That is third in the world. Ahead of it sit Claude Fable 5 Max at 1,815 and GPT-5.6 Sol Max at 1,747.8. Just behind it sits Claude Opus 4.8. On frontend code, in blind developer testing, K3 placed first, ahead of Fable 5. It leads the field on sustained coding work.

I am not going to tell you it is the best model in the world. It is not. The two above it are genuinely above it. What I will tell you is that it is close enough to the top that the gap has stopped mattering to anyone actually building something, and it is the only one of the three you can own outright.

Read that back. The frontier is now a place you can move into for the price of the hardware. A year ago the most capable intelligence on the planet was a subscription. Today something within arm's reach of it is a download, and it came from a lab most of my guests have never heard of, in a country the industry spent three years insisting was a step behind.

The step-behind story is over. What replaces it is more interesting.

## The cellar, and the person in front of it

Here is the frame I cannot stop seeing it through.

A great wine cellar used to be a moat. If you had the bottles, the allocations, the depth nobody else could assemble, you had something a competitor across the street simply could not match. The list was the edge. Spend enough, wait long enough, know the right people, and you owned an advantage that was made of scarcity.

That has been quietly ending in wine for a decade, and it just happened to AI in a single week. When the rare thing becomes something anyone can download, the rare thing stops being the edge. The cellar went open source.

So what is left.

What is left is the only thing that was ever actually scarce, which is the person standing in front of the cellar. The one who knows which of those thousands of bottles belongs on this table, tonight, in front of these people, with this food, at this moment in their evening. The bottles were never the hard part. Choosing was the hard part. Reading the table was the hard part. The cellar was just the raw material, and raw material has a price, and that price is now falling toward zero.

This is the thing I keep arguing, most recently in Palantir Copied the Dining Room, and the industry keeps proving it for me. Intelligence got cheap. You can hold two point eight trillion parameters of it on a machine in your back office. Judgment did not get cheap, and judgment is the entire job.

## Why it matters that it was not built in English

There is a second thing in the K3 story that most of the coverage walked straight past, and it is the part a hospitality person should sit with longest.

The model was built by people thinking in Chinese. That is not a footnote. For three years the working assumption of this industry has been that the center of gravity is one valley in California and one language, and everyone else is downstream of it, translating, catching up, importing the frontier. A top-three open model out of Beijing breaks that assumption in public.

It should feel familiar, because it is the oldest story in wine. For a century the assumption was that the center of the world was a few hundred square kilometers of France, and everything else was an imitation of it or a discount from it. Then the imitations got very good, and then they stopped being imitations, and now the most interesting bottles on my list come from places the old map did not have names for. Growers, not houses. Countries the canon ignored. The center did not hold, because the center was never a place. It was a temporary agreement about where to look.

AI is having its version of that week right now, and it happened faster than wine's version because software copies for free. The lesson is the same one a good sommelier learns early and painfully: the moment you believe the important thing can only come from one place, in one language, you have stopped being able to see it arrive from anywhere else. The people who miss the next great thing are almost never the people who could not afford it. They are the people who were certain they already knew where it would come from.

## What an operator should actually do about this

I try never to write one of these without a sentence a working person can use on Monday, so here it is.

Stop paying for the model. Start paying for the deployment.

The intelligence is about to be the cheapest input in the entire stack. Within a year, paying a premium for access to a frontier model will look the way paying a premium for access to the internet looks now, which is to say it will look like something people used to do before it became plumbing. The bottle is going to nine dollars. It is not where the value is.

The value is where it has always been in my trade. It is in the person who can look at your specific business, find the one thing worth automating and the three things you must never touch, deploy it on the systems you already run, and prove in your own numbers whether it earned. That person is not selling you intelligence. Intelligence is free now, or it is about to be. That person is selling you the read, and the read is the only thing in this entire story that did not just get cheaper.

The most capable open model on earth is now a download, built by people who were supposed to be a step behind, proving that the frontier is not a place and never was. That is a story about China, and about open source, and about a benchmark leaderboard.

It is also, if you have spent a career in front of a cellar, the most familiar story you have ever heard. The rare thing became common. The judgment stayed rare. It always does, and knowing that in advance is the closest thing to an edge that any of us get.

---

# Palantir Copied the Dining Room

URL: https://woodywusommelier.com/writing/palantir-copied-the-dining-room/
Pillar: The Terminal
Published: 2026-07-21

There is a job in artificial intelligence right now that pays a mid-level person around 385,000 dollars a year, and a senior one past a million. Postings for it jumped roughly 800 percent in nine months. Anthropic, OpenAI, and Google are all hiring for it as fast as they can. It is called the Forward Deployed Engineer, and it is one of the most sought-after roles in the industry.

Here is the part almost nobody mentions. When Palantir invented the role, they modeled it on a restaurant.

Not as a loose metaphor. As the founding image of the whole role. The story inside Palantir is that their CEO watched the way French waiters work, the way the floor is fused to the kitchen, the way a great waiter will tell you no when you order the wrong wine, and built the engineering practice on it. The engineer sits in the client’s office, learns the business the way a server learns a table, and earns the standing to say no to a bad request. They called the first ones Deltas, and until around 2016 they had more of them than they had ordinary software engineers.

I have built my career in the room that job was copied from. So let me tell you what the AI industry just rediscovered, and why it decides who actually gets their money’s worth out of all this.

## Intelligence got cheap. Judgment did not.
Start with what changed. For most of computing history, the intelligence was the scarce thing. You paid for the smart system.

That era is over. Every company on earth now has access to roughly the same frontier models. The intelligence is a commodity, priced by the token, available to your competitor at the same rate as to you. When everyone has the same brain for rent, owning the brain stops being an advantage.

So the value moved. It moved to deployment, which is a polite word for a much harder thing: knowing where the intelligence should and should not be pointed inside a real business, and shipping something that actually works there. The numbers are brutal about how rare that skill is. One widely cited MIT study found that 95 percent of enterprise generative AI pilots produced no measurable financial impact. And they almost never fail because the model was not smart enough. They fail because someone automated the wrong thing. A problem that did not matter. Or one that needed a human and never should have been touched.

Picking the right problem is the whole game. And picking the right problem is not an engineering skill.

## What the dining room actually teaches
Here is why Palantir looked at a restaurant, and here is what a restaurant taught me before I ever wrote a line of code.

A great server is not taking orders. A great server is running a live diagnostic on a table they met four minutes ago. Is this a celebration or a negotiation. Who is paying, and do they need to look generous, or careful. Is the quiet one bored or just tired. Should I move faster, slow down, disappear, or become the whole evening. None of that is on the menu. All of it decides whether the night works.

And the best ones do the hardest thing of all: they tell you no. No, not that bottle with that dish. No, you do not want the tasting menu tonight, you are too hungry and too rushed, let me feed you properly instead. Saying no to a paying guest, in a way that makes them trust you more, is the highest skill on the floor. It is also, it turns out, the highest skill in deploying AI. The consultant who cannot tell a client “you should not automate this” is the one who sells them the 95 percent failure.

That is what the dining room teaches that a computer science degree does not. How to sit with someone, read what they actually need under what they are asking for, and have the standing to redirect them. Palantir could not find that skill in engineers, so they built a role around people who had it and taught them to ship. The AI labs are doing the same thing right now, at enormous cost, because the skill is rare and they know it is the bottleneck.

I did not have to learn that half. I have been doing it every night on the floor.

## The sommelier is the deployment
There is a sharper version of this, and the people building the AI stack keep reaching for it without quite knowing why. When they try to explain what a Forward Deployed Engineer actually does, they land, again and again, on the sommelier.

Think about why. A sommelier does not have one recommendation. A sommelier has a cellar, and a table, and the entire job is the match between them. You read the guest: what they know, what they can spend, what they are eating, whether tonight is a celebration or a Tuesday, whether they want to be led or left alone. Then you find the one bottle, out of hundreds, that fits this person, this plate, this night. The bottle that is perfect for table nine is wrong for table ten. There is no generic pairing. There is only the read, and the match, performed fresh every single time, by someone who does it on the floor, night after night, with real people and real stakes.

That is deployment. That is the whole thing.

Because there is no generic AI deployment either. The workflow worth automating at one restaurant is the one you must never touch at the next. The tool that saves a hotel group two hundred hours is useless to the room down the street with a different system and a different bottleneck. The frontier model is the cellar: vast, powerful, available to everyone at the same price. The value is not the cellar. The value is the person who can read this specific business and make the one match that fits it. A pairing, not a product. Done case by case, by someone who has spent their life learning to read a room.

The AI industry keeps using my job to describe its most valuable one. It is not an analogy. It is the same skill, pointed at a different table.

## The two halves, and why almost nobody has both
The Forward Deployed Engineer works because it fuses two things that rarely live in one person.

One half is the build: you can ship a production system, against real data, on the systems a business already runs, with the guardrails and the human approval that make people trust it. I wrote a whole essay arguing that managing AI agents is just managing staff, because I had been doing both. That half I earned at a terminal, after service, building the tools I wished the industry had.

The other half is the read: you can walk into a working business, feel where the money is quietly leaking, know which of its problems is worth solving and which one to leave alone, and say so with enough authority that they believe you. That half I earned on the floor.

Most people chasing this role have the first half and are learning the second the hard way, one failed pilot at a time. A smaller number have the second and cannot ship. The people who have both are the ones getting paid a million dollars, and there are not many of them, and almost none of them come from hospitality, which is strange, because hospitality is where the model came from.

## What this means for anyone trying to buy AI
If you run a restaurant, a hotel, a wine business, a group, and you are being sold AI right now, here is the test. The person selling it to you: have they ever worked your floor? Because if they have not, they cannot tell which of your problems is worth solving. They will sell you all of them, because selling you all of them is their business model, and most of it will fail, and it will not be the model’s fault.

The right person does the opposite. They sit in your business first. They find the one workflow that is quietly costing you hours or covers or comps, they build the thing that fixes it on the systems you already have, and then they prove it saved you money in your own numbers. And they tell you, plainly, which of your problems to leave alone. That last part is how you know they are worth hiring. Anyone who never says no is selling.

## Where I stand
I am a sommelier and a hospitality operator who learned to build. I built my career in the room the most valuable job in AI was copied from, and I spent my nights learning to ship the tools. I am not an AI consultant. That phrase means nothing now; everyone is one. I am the person you commission to find the one thing worth automating, deploy it, and prove it earned.

The industry spent a fortune rediscovering that the dining room had the answer all along. Some of us never left it.

If your room, your list, or your group is trying to figure out where AI actually pays, that is the conversation I am built for. I will even tell you where it does not.

Woody

---

# Michelin Came for Burgundy. The Doors Stayed Shut.

URL: https://woodywusommelier.com/writing/michelin-came-for-burgundy-the-doors/
Pillar: The Cellar
Published: 2026-07-15

Last week, in a palace in Dijon, the Michelin Guide did to Burgundy what it did to restaurants a hundred years ago. It made the illegible legible.

One grape, two grapes, three grapes. Ninety-four estates recognized in the first selection, nine at the top. A new symbol from the institution whose symbols decide careers, rents, and reservations across the entire world I work in. And before I say anything else, let me say the honest thing: it is not a bad list. It is actually the safest list imaginable. If you know nothing about Burgundy and you buy only three-grape names, you will drink well for the rest of your life.

But I did most of my wine education in Burgundy. Years of trips, most of them based in and around Beaune, walking and driving those villages, tasting in those cellars, building the relationships that eventually became my career. I know that list not as a list but as a set of actual doors on actual lanes. And that is why I can tell you the thing the announcement does not say.

Almost none of those doors will open for you.

## The facts, before the take
Announced: July 7, 2026, at the Palace of the Dukes of Burgundy, Dijon.Covers: Côte de Nuits, Côte de Beaune, Côte Chalonnaise.The tiers: 94 estates recognized in total. 9 with three grapes, 20 with two, 33 with one, and the remainder listed as Selected, recognized without a grape.The five criteria, in Michelin’s order: quality of agronomy, technical mastery, identity, balance, consistency.The top distinction, in the Guide’s own words, is reserved for wines that can be approached with complete confidence, regardless of the vintage. Hold onto that sentence. We will come back to it.

All nine three-grape estates: Domaine de la Romanée-Conti, Domaine Leroy, Domaine d’Auvenay, Coche-Dury, Georges Roumier. Those are the untouchables, and rating them is Michelin telling you the ocean is large. Then: Cécile Tremblay, Dugat-Py, Jean-Marc & Thomas Bouley, Hubert Lamy. Those were whisper names, insider currency until last Tuesday, the kind a somm said quietly to a guest who had earned them.

One more fact, because most readers do not know it: Michelin has owned Robert Parker’s Wine Advocate, the most powerful wine-scoring publication in the world, outright since 2019.

Two kinds of names, one podium. The full list of 94 is on Michelin’s site. The take below is about what the list cannot say.

## The stars you can book. The grapes you cannot.
Understand what a Michelin symbol has always promised. A star is a restaurant: you can call, book a table, and eat the thing that was rated. A key is a hotel: you can book a room and sleep in the thing that was rated. The entire authority of the guide, from its first edition, rests on one promise: we rate places you can go. It was invented by a tire company to make you drive somewhere.

The grape breaks that promise in its first week.

The Burgundy domaines on this list are, overwhelmingly, not open to the public. They are working farms and family cellars behind gates and unmarked doors. They do not run tasting rooms. They do not take walk-ins. Most will not take an appointment from a stranger at all, because there is nothing in it for them: the wines are allocated years deep, there is nothing to sell you at the door, and the person you would be interrupting is on a tractor. I have stood in those lanes. Google Maps half works at best among the vineyards and the family parcels; plenty of tourists will not even locate the courtyard they came for, and the ones who do will find it closed. Following the grape selection as a travel guide, which is what a Michelin symbol has trained a century of consumers to do, will end in a locked gate ninety-nine times out of a hundred.

Read that plainly. The most famous travel guide on earth just published a guide to places that do not receive travelers. It rated the wine, but it wears the costume of an invitation, and the invitation is not real. A star is a table. A key is a bed. A grape is a door that stays shut.

## So what did Michelin actually print?
A few weeks ago I wrote that in the new status economy the flex is no longer the bottle, it is the access: the text from the person who can get you the bottle. When this list dropped, my first thought was that Michelin had just printed the whisper network, because of that second group on the podium, the names that were currency until last Tuesday.

But look closer and something stranger is true. Michelin printed the names. It could not print the keys.

Knowing that Bouley exists is now free. Getting a bottle of Bouley, or a seat in that cellar, is exactly as hard as it was last month, and about to get harder, because the allocations were already thin and the demand just went global. The guide made the knowledge legible and left the access locked. Which means the whisper network did not die last week. It appreciated. The value of knowing a name just went to zero, and the value of knowing the person behind the door just went up accordingly. Every signal that gets printed pushes the real currency one level deeper. I keep writing this sentence about other things, and Burgundy just staged the cleanest demonstration of it I have ever seen.

## Read the list bottom up
Now the part that matters if you actually drink wine rather than collect symbols.

The most interesting names on this list are not the nine at the top. They are further down, in the Selected tier, recognized but not graped. That level is full of smaller estates, younger growers, and quietly serious farming, and it is where the actual value lives: wines still priced like knowledge instead of like trophies, from people still building. The three-grape tier is a museum of settled arguments. The Selected tier is where the argument is still happening, and anyone who has tasted through those villages recently knows some of those cellars are making wines that will embarrass their price for the next decade.

This has always been the way to read a Michelin product. The connoisseur’s move with the stars was never the three-star pilgrimage; it was the one-star doing two-star work, the bib gourmand with the ambitious kitchen. Same list, read bottom up. The guide’s grades are a map of consensus. Value lives just outside consensus, always.

## Who is holding the pen?
Now the question the trade is asking in private, and the reason I planted that ownership fact up top.

Michelin does not just rate wine as of last week. It has owned the scorekeeper for years. The Wine Advocate, the publication whose hundred-point scale has moved wine prices for four decades, is a wholly owned Michelin property. Its editor in chief is the most influential Burgundy critic working today. He lives in Beaune. And, by his own open disclosure, he makes wine there: a small négociant label with a couple of tiny parcels in the Côte de Beaune and a winery in Pommard. None of this is a secret. All of it is public record, and he has been admirably transparent about the winemaking.

Now hold those facts next to this one: the first region Michelin ever chose to rate is the region where its own chief critic lives and farms. And the grapes were awarded, Michelin says, by an international team of specialists it declines to name.

Then read the list with a trade eye, because this is where it gets uncomfortable. The surprises at the top are not random names. They line up with a very contemporary Burgundy palate: farming-first, texture-conscious, less impressed by inherited fame than by precision, energy and intent. The loudest snub, Armand Rousseau at a single grape while the collector forums howled, makes the point even clearer. Nuits-Saint-Georges, all but shut out. Fourrier and Hubert Lignier, parked in Selected. The list does not read like a bland committee compromise. It reads like a palate with opinions. A genuinely brilliant one, to be clear. But a palate has a name, and this list does not.

I cannot tell you who held the pen. Nobody outside the company can, and that is itself the finding. What I can tell you is what changed culturally. Restaurant inspection has always been anonymous by design. Wine criticism has always been signed, because in wine the authority IS the palate: Parker’s name was literally on the masthead. Last week, somewhere inside one company, a signed opinion appears to have become an unsigned medal. Personal taste, laundered into institutional fact.

I am not alleging a fix. I am describing a structure, and the structure is the story. One company now owns the restaurant stars, the hotel keys, the wine grapes, and the wine scores, and its chief wine critic grows grapes in the first region it chose to crown. In any other industry we would have a word for that much of the scoreboard under one roof, and the word would not be terroir.

## The criterion that worries me
Now the sentence I told you to hold onto: the top distinction is reserved for wines that can be approached with complete confidence, regardless of the vintage.

This is Burgundy. The entire argument of Burgundy is that the year matters: the frost, the hail, the misery of one September and the miracle of the next. A Burgundy vintage is a diary entry written by weather, and the growers there do not overcome the vintage, they translate it. Rewarding confidence regardless of the year is a consistency criterion, and I have watched what consistency criteria do to kitchens for twenty years. The stars produced excellence and quietly filed the fingerprints off. Applied to wine, the same pressure points the same direction: sand the year out of the bottle. I hope the panel resists it. The first criterion on their list, quality of farming, suggests the instinct is right. The vintage clause suggests the machine is still the machine.

Both can be true. They usually are.

## The part I have to admit
Because credit is owed. Michelin’s first criterion for wine is agronomy. Farming, before technique, before polish. The most powerful institution in hospitality just declared that the dirt comes first, which is the argument the growers have been living and the argument this newsletter has been making since the day it launched. And a handful of farmers who were never anyone’s brand just got handed the biggest amplifier in the industry. Some of them will get a fair price for their work for the first time. I cannot pretend that is bad news, and I will not.

Own the dirt, miss the mic, I wrote about Champagne. Burgundy’s farmers still own the dirt. The mic just got enormous. What nobody mentions is that the wire between the mic and the door is not connected, and someone still has to walk you in.

## Where this leaves things
Here is the strange, honest summary. The list is good. The farming criterion is right. The value is in the tier nobody will brag about. The top tier is a rating of doors that will not open. And the one thing the guide could not print, the relationships, the appointments, the person who answers when the gate is closed, just became the scarcest thing in the most scrutinized wine region on earth.

I spent years of my life in those villages learning exactly that unprintable layer, never guessing an institution would one day point the whole world at it. The guide can tell you what is behind the doors now. It cannot open them.

Enjoy the whisper while it is a whisper. And when you read the list, start from the bottom.

Have you ever actually gotten inside a Burgundy domaine, and how? And which Selected-tier name would you quietly back for three grapes in ten years? Reply. I read everything.

---

# What Sobriety Is Worth

URL: https://woodywusommelier.com/writing/what-sobriety-is-worth/
Pillar: The Cellar
Published: 2026-07-13

A zero proof drink should not arrive with the emotional energy of an apology.

Not in a serious room.

Not in 2026.

Not when the guest has already told you something useful: I am here. I am spending. I care about the experience. I do not want alcohol in the glass.

That is not a problem.

That is a brief.

The industry keeps treating zero proof like a side quest. A soda with garnish. A lemonade that went to finishing school. A list item written in the voice of someone who does not believe the guest will order it.

Meanwhile, the room has changed.

In 2025, Gallup found that 54 percent of US adults reported drinking alcohol. That is the lowest level in Gallup tracking since 1939, and down 13 points from 2022.

This is not a niche.

This is not only recovery.

This is not only wellness.

This is the market telling the dining room that the old beverage script is losing coverage.

The lazy response is to panic about alcohol sales.

The operator response is to protect the check without insulting the guest.

The health story has moved faster than most wine lists.

AP reported Gallup data showing that 53 percent of Americans now say moderate drinking is bad for health. In 2015, that number was 28 percent.

You can argue with the mood if you want.

The guest has already changed the order.

This is where restaurants get weird. They will spend twelve minutes describing the texture of a white asparagus supplement, then hand the sober guest a list that sounds like a punishment for making eye contact with their own life.

That is not hospitality.

That is leakage.

Younger guests are not waiting for permission.

About two thirds of US adults aged 18 to 34 now say moderate drinking is unhealthy, according to the same Gallup reporting.

The next generation of premium guests may still order wine. Many will. But they will order it with a different mental model: less automatic, more intentional, more suspicious of ritual that cannot explain itself.

That does not kill wine.

It kills lazy wine.

It also creates room for a serious zero proof program.

Not a mocktail list.

A program.

There is a difference.

A mocktail list says: here are three sweet things for the person who cannot participate.

A program says: here is acidity, bitterness, texture, temperature, pacing, pairing logic, and a reason to stay in the conversation.

One protects a guest from feeling excluded.

The other protects revenue.

The strongest signal is not low alcohol. It is no alcohol.

IWSR data reported in January 2025 found that no alcohol products added 61 million buyers between 2022 and 2024 across ten key countries. Low alcohol added 38 million.

That makes sense.

Guests understand zero faster than almost.

Low alcohol can be useful, but it asks the guest to do math. Zero proof makes the promise clean.

For restaurants, the opportunity is brutally practical.

A guest who would have ordered water can order something with a margin line.

A guest who would have had one glass can stay in the pairing rhythm.

A table with one non drinking guest no longer has a dead zone in the beverage conversation.

The server gets language. The guest gets dignity. The room gets revenue.

Run the house math.

If a zero proof drink sells for 12 dollars with a 2.50 dollar build cost, that is a 79 percent gross margin before labor and overhead. If it replaces tap water, the check grows. If it replaces a low confidence cocktail, the service gets cleaner. If it keeps one guest engaged through three courses, the program has already done its job.

This is not medical advice.

This is floor advice.

The sober guest is not low value.

A lazy room makes them low value.

The next great beverage program will not be the one that shames alcohol or pretends wine is over.

It will be the one that understands the whole room.

Wine for the guest who wants wine.

Zero proof for the guest who wants presence without the pour.

Language for the server.

Math for the operator.

Taste for everyone.

That is what sobriety is worth.

Reply with the zero proof drink that did not feel like an apology. I am collecting the ones that actually did the job.

---

# Taste Is the New Money. Ask the Sommelier.

URL: https://woodywusommelier.com/writing/taste-is-the-new-money-ask-the-sommelier/
Pillar: The Cellar
Published: 2026-07-05

There is a theory going around that taste is the new money.

That when you cannot afford the house or the car or the bag, you compete on what you know instead. The obscure reference. The right smoothie. The grower nobody can pronounce. Knowing over owning. The internet is calling it a new currency and treating it like a discovery.

I want to welcome everyone to the economy I have worked in my entire adult life.

A sommelier is a professional trader of exactly this. My whole career is the conversion of what I know into where I can go. I get paid to know the farmer behind a bottle you have never heard of, and to carry that knowledge in a way that opens a door, a table, a tip, a job. Taste as currency is not a trend to me. It is my rent.

So when the culture announces that knowing is the new money, my honest reaction is: yes, obviously. And also, you have no idea how strange this currency gets once you actually have to live on it.

## It is not new. The trigger is.
First, the part the videos skip. None of this is new.

Pierre Bourdieu spent the 1970s explaining that taste was never personal. He called it cultural capital, and his whole point was that what you know and how you carry it is a currency, one that quietly converts into the other kind, the kind with a dollar sign. Taste, he argued, is how a class keeps its shape. It was never really about the wine. It was about who gets to feel at home in the room.

So the concept is old. What is new is the trigger, and the trigger is economic, not cultural.

Look at why this is happening now. In the United States, first-time buyers have fallen to a record-low share of the market, and the median first-time buyer is now about 40. In 1981, the median first-time buyer was 29. Under-35 homeownership is still below its mid-2000s peak. Home prices sit around five times median household income. Baby boomers control tens of trillions more wealth than millennials, a generational balance sheet so absurd it starts to feel like climate.

The ladder did not get harder to climb. Somebody moved it into a private room.

Here is the rule underneath all of it. Status competition never stops. It migrates to wherever you can still compete. When you cannot win on wealth, you do not stop keeping score. You move the game to a board where you still have pieces: knowing, taste, references, access.

That is not a new aspiration. It is the old aspiration, rerouted around a wall.

## The signal eats itself
And the moment a taste signal gets legible enough to sell, it dies.

Watch it happen in real time. A sub-hundred-dollar bag goes viral as a stand-in for a five-figure one, sells out, breaks the internet, and instantly means something different, because the entire value of the original was that you could not just buy it at Walmart. A twenty-two-dollar smoothie becomes the safest flex in the country, priced exactly at painful but possible, until everyone is holding one for the camera and it stops saying anything at all.

So this currency is trapped in an arms race it cannot win. Every signal that becomes readable becomes copyable, and every copy collapses the value. Which means the real money is always fleeing toward the thing that cannot be photographed or copied: the more obscure, the more ephemeral, the more you had to be there.

Scarcity left the object. It moved to the access.

That is why the flex is no longer the bottle. It is the text from the person who can get you the bottle.

## I watch this every night
I do not have to theorize about this. I serve it.

Every service I wait on two kinds of people who order the same rare wine. One of them knows exactly what it is, why it is hard to find, who farmed it, and they order it quietly, almost privately, the way you say a name you love. The other read about it this week and orders it loudly, watching the table to see if it landed.

Same bottle. Two completely different currencies changing hands.

Part of my job is to serve both perfectly without ever letting either one know which one I think they are.

Wine is the purest version of this whole phenomenon, because wine already ran the entire arms race. It went from the famous label that announced your money, to the quiet bottle that announced your restraint, to the grower nobody has heard of that announces something rarer than money: that you know.

The most powerful wine on any list today is not always the most expensive one. It is the one only a few people at the table can even read. The allocation. The half case an importer texts about before it hits the floor. My version of the whisper network is a list of growers and a phone that rings first.

## What it buys, and what it never will
Now the part the new currency people do not want to print. The part you only learn by living on this money instead of selling courses about it.

Cultural liquidity gets you into the cellar. It does not give you the deed.

I have written before about who actually owns Champagne. The farmers own and tend nearly all of the vineyard: the land, the risk, the truth of the place. The famous houses own barely a tenth of the dirt and almost all of the money.

Own the dirt, miss the mic.

Cultural liquidity is the grower’s kind of capital. It is the realest thing in the room, and it is not the thing that pays. You can know more about wine than every wealthy person you will ever pour for, and you will still be the one pouring.

That is the line the smoothie does not want you to read.

Context is a wonderful thing to have and a terrible thing to be handed as a substitute. “I cannot buy the house, but I can buy the smoothie” is not a new currency. It is precarity in a nice jacket, and someone is selling you the jacket.

The most honest thing about this entire moment is that the people teaching you to value knowing over owning are, almost always, trying to sell you something to know.

## I trade in it and I do not even drink
Here is my strangest credential.

I trade in this currency and I do not consume the product. I taste, and I do not swallow. Dozens of wines a week, none of them down the hatch.

Which makes me about as pure a case of cultural liquidity as exists. Maximum knowing, zero owning, zero consuming. And from inside that, I can tell you the thing is real.

Knowing is genuine capital. It opened every door my career runs through. It is also not a deed, not a house, not security, and no amount of carrying it well will turn it into one.

Both of those are true at once, and anyone selling you only the first half is selling.

## So is taste the new money
Partly.

It is real capital. It opens real doors. Mine were all opened with it.

But it is the grower’s money, not the house’s. More truth, less wealth. A way to be at home in rooms you cannot yet afford to own. That is not nothing. For a generation locked out of the old ladder, it might be the most alive part of the culture.

Just do not let anyone convince you that knowing the name of the thing is the same as owning it.

The somm knows every bottle in the cellar. The cellar still is not his.

Know everything. Carry it beautifully. And keep both eyes on the door the knowing is supposed to open, because the people selling you the currency are very much hoping you forget there was ever a door at all.

- Woody

What’s the thing you know that opened a door money couldn’t? And the door it never opened? Reply with both. I read everything.

---

# Tasted, Never Swallowed

URL: https://woodywusommelier.com/writing/tasted-never-swallowed/
Pillar: The Cellar
Published: 2026-07-01

There is an argument running through the culture right now, and the strange part is that both sides are winning.

One side says the cigarette is back. Not as a habit, as a symbol: on the runways, in the music videos, on the beauty packaging, a backlash against a decade of wellness. The other side, mostly people younger than me, says that is nonsense, that nobody their age actually smokes. “Gen Z does smoke.” “I’m Gen Z and I smoke.”

Here is the thing. They are both right. And the gap between them is the whole story.

## The habit is dying. The image is not.
Start with the behavior, because the behavior is almost gone. In 2024, 1.4 percent of American middle and high school students reported smoking cigarettes. That is the lowest number ever recorded. In the late 1990s, it was more than a third of teenagers. The habit did not decline. It collapsed inside one generation.

The commenters are not wrong, though, because nicotine did not die. It moved. Vaping sits at 5.9 percent and pouches are climbing. So when someone says Gen Z still smokes, they are right about the drug. They are just wrong about the object. The drug went to the most convenient possible delivery device. The cigarette got left behind.

And while the cigarette was being abandoned as a habit, look what happened to it as an image. Searches for the “ciggy aesthetic” have surged. “Smoking pose” is up 70 percent among 18 to 24 year olds. Eight of the ten Best Picture nominees at this year’s Oscars feature tobacco on screen, the first increase since 2020. An entire pop aesthetic got built around a paparazzi photo of a singer with a cigarette.

So here is the one line that matters. The cigarette is having a renaissance as an image at the exact moment it is dying as a habit. That is not a coincidence. That is the mechanism.

## I do this for a living
I recognized the move immediately, because I make it every shift.

I am a sommelier. I taste dozens of wines a week and I swallow almost none of them. I spit. On purpose, in front of people, as the job. I have spent years separating the knowledge of a wine from the drinking of it, holding the substance in my mouth precisely so I never have to consume it. Tasting and spitting is the cigarette trend performed by a human being. The ritual, kept. The substance, declined.

Once you have lived inside that split, you see it everywhere, and you understand why it works.

A symbol is only powerful when the behavior behind it is rare. When everyone smoked, a cigarette said nothing. It was wallpaper. Now that almost no one does, the same object says something loud: I am secure enough to court a little ruin, to visibly opt out of the optimized life everyone else is performing. The image did not get potent in spite of the habit dying. It got potent because the habit died. Scarcity is the entire engine of meaning. Anyone who sells wine could have told you that.

## Why the vape will never have a runway
This is also why the vape will never get a fashion moment, no matter how many it sells.

The vape is the optimized product. Convenient, current, consumed, everywhere. And optimized things cannot be status symbols, because status was never about what you use. It is about what you can afford to refuse. The vape is the habit. The cigarette is the language. Only one of them photographs, and it is the one nobody needs anymore.

## The bottle is already doing this
I watch the same thing every night on the floor, and it has nothing to do with tobacco.

The most photographed object in a dining room is the bottle. Most of those photos are taken before anyone has tasted a drop, sometimes before the cork is even out. The label is the message. The wine is almost incidental. People are not always drinking the bottle. A lot of the time, they are wearing it.

Grower Champagne runs on the same current. Its power is legible scarcity: a name almost no one recognizes, a place you have to earn your way into, a story that does not fit on a billboard. The famous houses sell recognition. The growers sell the opposite, and the opposite is the flex now.

And then there is the part that should interest anyone in my business most. The fastest growing thing at the bar is the drink with the alcohol taken out. People are ordering the ritual without the substance: the coupe, the bitterness, the garnish, the ceremony, none of the intoxication. That is the cigarette trend, in a glass. The symbol, kept. The habit, declined. I have been the living version of this for over a year, and the culture is finally catching up to the bar.

## What the brands are getting wrong
So when a brand stitches a cigarette onto its packaging or sends one down a runway, it is reading the trend and missing the mechanism.

The signal was never the cigarette. The signal is three things: legible scarcity, analog texture, and a little transgression. Copy the cigarette literally and you inherit the regulation, the ethics, and the health math, and you get none of the meaning. Copy the variables instead and you can build the same feeling out of almost anything that is rare, physical, and slightly against the rules. A wax seal. A handwritten list. A rotary phone. A wine nobody at the table can pronounce.

## The oldest lesson on the floor
The cigarette is not actually coming back. The behavior is gone, it is not returning, and that is a good thing. What came back is the oldest lesson in hospitality, the one a sommelier learns on the first day and a sober one learns twice.

The most powerful thing in the room is the thing you hold and choose not to use.

Tasted, never swallowed. Turns out that was never only about wine.

Woody

What’s the object you hold and never use, the one that says more empty than full? Reply and tell me. I read everything.

---

# Nobody Owns Champagne

URL: https://woodywusommelier.com/writing/nobody-owns-champagne/
Pillar: The Cellar
Published: 2026-06-27

Champagne fell for the third year in a row.

The trade headlines called it a crisis.

I looked at the ownership map.

Here is the number that reframes the whole panic. Growers own and tend nearly 90 percent of the vineyard in Champagne. The land, the labor, the farming risk, the actual chalk that makes the wine worth the bottle.

But in export value, the houses capture 89 percent.

Own the dirt. Miss the mic.

That is the real map.

The houses are having a hard few years. The farmers have the more interesting story. Let me show you the data, because once you see it, you cannot unsee it.

## The houses have a problem

Global Champagne shipments peaked at 326 million bottles in 2022, the big rebound year after the pandemic. They fell to about 299 million bottles in 2023, 271.4 million in 2024, and 266.1 million in 2025.

Three straight years of decline.

You can blame the economy, tariffs, geopolitics, and changing habits, and you would be partly right. But a category that sells on occasion and status is uniquely exposed the moment people get careful with money.

The house model was built for the boom.

Now here is the part the crisis coverage skips.

## They own the land. They do not get the money.

The growers, the ones labeled RM in letters so small you have to know to look, own and tend nearly 90 percent of the Champagne vineyard.

They farm the slopes. They carry the cost of frost and hail. They make wine from their own fruit, in their own villages.

The houses, 410 of them in the official 2025 figures, own a much smaller share of the land. They buy grapes. They blend across villages, across years, across thousands of growers, to build a house style that tastes consistent every time it is opened.

And they still ship most of the bottles.

In 2025, houses shipped 191.9 million bottles out of 266.1 million total. That is about 72 percent of volume.

In export value, houses captured 89 percent. Cooperatives and growers captured 11 percent.

Read that again.

The people who own the dirt do not own the story.

Neither side is illegitimate. A great house can make a real product. But only one side of this map is scarce in the way wine people usually claim to care about. A place. A farmed slope. A family decision. A risk carried in one village instead of smoothed across the whole region.

That is where the useful story starts.

## What you actually pay for
When a guest buys a grande marque, a real share of the price is the box.

The marketing.

The airport lightbox.

The global distribution machine.

The wine inside can be genuinely lovely. The guest is still paying for a famous name that happens to contain Champagne.

When a guest buys from a grower, the line between land, labor, risk, and price gets shorter.

No racing budget.

No celebrity.

No airport theatre.

A name, a village, and a person who was out in the vines when the weather turned.

Grower Champagne is interesting for one plain reason: it shortens the distance between place and glass.

That is not romance.

That is the receipt.

## The math just turned

The United States is the prize.

In 2025, the US was Champagne’s top external market by both volume and value: 26.5 million bottles and EUR689 million.

That works out to about EUR26 per bottle in shipment value before shelf pricing, restaurant markup, freight, and everything else.

France averaged about EUR18 per bottle. The global average was about EUR21. Export markets averaged about EUR24.

So yes, America pays up.

But expensive is not the problem.

Unexplained expensive is.

When the market tightens, the bottle has to explain itself faster. A famous label can still win. But it has to earn the room. A grower can win too, if the story is legible enough for the server to translate under pressure.

That is the operator lesson.

The best bottle is not the one with the loudest story.

It is the one the room can actually explain.

## The useful window

I would not call this the end of Champagne.

That is too easy, and easy stories usually miss the money.

The better read is that the market is asking better questions.

What is the bottle?

Who made it?

Where is the value?

Why this price?

Why this glass?

Why now?

Grower Champagne answers those questions differently. Not always better. Not automatically. But differently in a way that matters when guests are more skeptical, buyers are more careful, and the old luxury script is not enough on its own.

This is not about hating houses.

It is about reading the map.

The houses will be fine. They usually are.

But the next interesting decade in Champagne belongs to the people who own and tend the vines, and who are slowly getting the mic.

Nobody owns Champagne.

But if you want to understand where the next useful story comes from, start with the people who own the dirt.

Woody

## Reply prompt
What Champagne label made the farm visible to you?

The grower, the moment, the room. Reply and tell me. I read everything.

Sources: Comité Champagne 2025 key figures and Champagne Winegrowers; The Guardian with Reuters shipment context; Wine-Searcher U.S. Champagne market context.

---

# Agents Are Just Staff You Train Once

URL: https://woodywusommelier.com/writing/agents-are-just-staff-you-train-once/
Pillar: The Terminal
Published: 2026-06-25

The tech industry has spent the last two years discovering, at enormous expense and conference length, a set of principles that every restaurant manager already knows.

They call it “agent design.” They write papers about it. They’ve coined a vocabulary (orchestration, delegation, human in the loop, evaluation, guardrails) and they present it as a new discipline, because for them it is. They’ve never had to run a Saturday service with a brand new server, a hungover line cook, and a sommelier who’s brilliant with Burgundy and a hazard with bills.

I run AI agents at night and a restaurant floor at night too, technically. And I’m here to report that they are the same job. If you have ever managed staff in a high pressure service environment, you already hold an intuition for working with AI that most engineers are still acquiring the hard way. Here’s the translation table.

## 1. The briefing is the product
A new server doesn’t fail because they’re incapable. They fail because the briefing was bad. “Take care of table twelve” produces chaos; “table twelve is a regular, anniversary, he’s allergic to shellfish, she chooses the wine, don’t rush the dessert” produces magic. Same human, different context.

Working with AI is briefing, full stop. The people getting mediocre results from these tools are writing “take care of table twelve” prompts and concluding the staff is bad. The people getting astonishing results are writing the second kind: role, context, constraints, examples of done right, what to do when uncertain. In the industry we’d call that a pre shift meeting and an SOP. The tech industry calls it prompt engineering and pays handsomely for it. Operators have been doing it twice a day, unpaid, forever.

## 2. Delegate the task, never the judgment
Every manager learns this through scar tissue: you can hand off the doing but not the deciding. Not at first. The new hire runs food before they take orders; takes orders before they handle complaints; handles complaints for months before they’re allowed near a comp. Trust is granted in layers, each layer earned by verified performance at the last one.

This is exactly, exactly, the correct posture with AI agents, and it’s where smart people fail in both directions. The doomers won’t let the new hire carry a plate (”it might drop it!”) and get no leverage. The reckless promote it to manager on day one, let it email clients unsupervised, and learn in public. The operator’s instinct is the right one: tight scope, verify the output, expand scope on evidence.

I learned this the hard way with the first real tool I built: a pipeline that reads photos of the bottles I’m bringing in and drafts the tasting captions for each one, in English and Mandarin, at volume. The first batch came back like a new server reciting a script he doesn’t believe. Fast, fluent, generic. Correct notes with no soul: red fruit, a touch of spice, a smooth finish. Copy that could describe four hundred wines and sell none of them.

The fix wasn’t a smarter model. It was a better pre shift. I narrowed the brief: lead with the grower and the place, name the family that farmed it, one true sensory detail, no marketing words, no em dashes. I cast the tool to the one thing it was good at and kept the rest for myself.

It got me to maybe seventy percent. Useful, not finished. I still read every caption, and every line of Mandarin gets checked by a native speaker before a single one goes out. The machine drafts the room; it does not open the doors. It earned a lane, not a promotion. That’s the whole posture: my agents earn their autonomy the way my staff do, gradually, with me checking the work until the error rate tells me I can stop.

## 3. Inspect what you expect
No chef trusts a station they haven’t tasted from. The plate gets looked at before it leaves the pass, every plate, every night, no matter how senior the cook. Not because the cook is suspect, but because the standard is the standard and entropy never sleeps.

AI work needs a pass. The single biggest production mistake people make with these tools is shipping unread output. It’s the equivalent of food going to the table straight off the stove with nobody’s eyes on it. My rule at the terminal is my rule at the kitchen window: everything gets tasted. The agent drafts; I plate. The moment you stop checking is the moment something goes out cold, and with AI as with food, the guest remembers the one bad plate, not the hundred good ones.

## 4. Staff have shapes, work with the grain
A great manager doesn’t fight a server’s nature; they cast them. The charming one gets the celebrating tables, the precise one gets the eight top with the complicated bill, the wine mad one gets the collectors. Skill isn’t generic. Deployment is half the craft.

Models have shapes too. Some are meticulous and slow; some fast and a little sloppy; some brilliant at structure and stiff at warmth. The operator’s move (cast by strengths, cover the weaknesses with process) translates directly. I don’t ask my fastest tool for nuance or my most careful one for speed, the same way I don’t put my most poetic somm on the table that just wants the check.

## 5. Escalation is a designed path, not a panic
The best service systems make one thing crystal clear to every staff member: when you’re out of your depth, here is exactly who you bring it to, and you will never be punished for bringing it. Ambiguity about escalation is how small problems become Yelp reviews.

Agents need the same explicit lane: here’s what you decide, here’s what you flag, here’s the line you never cross alone. The tech world calls this human in the loop and acts like it’s novel. It’s the oldest rule on the floor: comp the dessert yourself, but a furious guest gets the manager, every time, no heroics.

## 6. The manager’s hours change shape, not size
Here’s the part nobody on either side tells you. Hiring great staff doesn’t reduce a GM’s work; it transforms it, from doing the tasks to designing the system: hiring, briefing, checking, coaching, improving the SOP. The leverage is real and so is the new job.

AI is identical. The fantasy is “the agent does my work while I sleep.” The reality, the productive reality, is that your work moves up a level: from producing to specifying, from doing to reviewing, from task to system. You become the GM of a staff that costs pennies, never tires, and is only ever as good as your brief. Which means the limiting factor in the AI age isn’t the model.

It’s whether you ever learned to manage.

## The punchline
And there it is. The inversion I keep circling in these essays: the floor and the terminal trading places when no one was watching.

For decades, the deal was clear: tech skills were the leverage, and hospitality was the fallback, the thing you did while you figured out your real career. Now look at the actual skill stack the AI age rewards: briefing under ambiguity, layered trust, relentless quality inspection, casting by strength, designed escalation, systems management under pressure. That’s not a computer science curriculum.

That’s a GM’s Tuesday.

The operators don’t need to learn how to think about AI. They need to notice they already know, and then walk over to the terminal and start hiring.

Woody

If you run a floor, a kitchen, a hotel, a crew, or a company: what management rule have you tried on a machine? Reply and tell me. The best answers become a follow up post, with your permission and credit. I read everything.

---

# Nobody Is Killing Wine. The Cheap Story Is.

URL: https://woodywusommelier.com/writing/nobody-is-killing-wine-the-cheap/
Pillar: The Cellar
Published: 2026-06-24

State of the Pour, No. 1

The wine industry keeps looking for a murderer because murder is easier to explain than age.

A murderer gives you a suspect. Gen Z. Wellness. Cannabis. Phones. Ozempic. Ready-to-drink cans. The usual lineup of cultural villains, all standing against the wall under bad fluorescent lighting while the trade points dramatically and pretends the case is solved.

But the 2025 data says something less theatrical and more useful.

Wine consumption is not falling because one generation woke up and decided to destroy civilization. Global wine consumption is now estimated at 208 million hectolitres, down 2.7% from 2024 and roughly 18% below the 2007 peak. Production is still low, only 0.6% above the historically weak 2024 level. Global vineyard area contracted again. Export volume fell. Export value fell even faster

This is not a murder mystery.

This is a renovation notice.

The first thing to say is the obvious thing, because the wine business has become very good at making the obvious sound impolite. The decline is real. It is long. It is structural. It did not begin with the youngest legal drinkers, and it will not be fixed by a brand deck explaining that Pinot Grigio is fun now.

The shape of the line matters. The peak was 2007. That was before most of the current Gen Z discourse had learned to spell hangover. The deeper slide began before the latest wave of wellness culture became a beverage strategy. The industry wants a young villain because a young villain allows the old model to remain innocent.

It is not innocent.

For decades, the volume side of wine made a simple bet: make enough acceptable wine, dress it in enough aspiration, and let habit do the rest. The customer would drink more, trade up occasionally, and accept vagueness as romance. The trade could say “lifestyle” when it meant margin. It could say “approachable” when it meant flattened. It could say “premium” until the word gave up and sat quietly in the corner.

That customer is leaving.

Or more precisely, that customer is becoming more selective, which looks like leaving if your entire model depended on low selectivity.

The second thing to say is that the vineyard has already understood this. The world vineyard surface fell again in 2025. The OIV estimates 7.0 million hectares, down 0.8%, the sixth consecutive contraction. France dropped 4.4%. Chile dropped 3.7%. Argentina, Spain, and the USA also moved down. Brazil and India moved up

The vineyard votes first.

Before the panel discussion, before the marketing reset, before the consultant gets paid to rename the same bottle in a softer font, the land starts answering the harder question: does this hectare still make sense?

In some places the answer is yes. In others, no. That is painful, but pain is not always a sign of death. Sometimes pain is the proof that a body is no longer willing to carry what does not serve it.

I do not say this casually. I work on the floor. I have watched guests hesitate over bottles they would have bought without blinking five years ago. I have watched the same guests happily spend more when the bottle had a reason, a story, a texture, a human being behind it, and a person in front of them able to explain why it belonged on the table.

That is the split.

Cheap meaning is dead. Real meaning is not.

The third thing to say is that trade is showing the same pressure. In 2025, world wine exports fell to 94.8 million hectolitres, down 4.7%. Export value fell to 33.8 billion EUR, down 6.7%

That gap is the quiet alarm.

When volume falls, you can blame demand. When value falls faster than volume, you have to talk about pricing power. That is where the room gets uncomfortable, because pricing power is not created by calling something premium. It is created when the customer believes the exchange.

The exchange is the whole game now.

This is why I do not buy the lazy decline story. The data is not saying wine has no future. The data is saying wine no longer gets automatic permission. A bottle has to justify its place in the fridge, in the cellar, on the list, on the expense account, on the anniversary table, and inside the body of a drinker who is increasingly aware that pleasure has a cost.

That cost is not only money.

It is health. Attention. Time. Identity. Sleep. The morning after. The feeling of whether the thing was worth it.

This is where the younger drinker is not the problem. They are the early warning system. The legal-age Gen Z drinker is not asking the wine business an irrational question. They are asking a devastatingly rational one:

Why this?

Not why wine. Why this wine? Why this bottle? Why this markup? Why this story? Why this night? Why this amount? Why this thing, when I have a thousand other ways to spend the same money and protect the same tomorrow morning?

That question does not kill wine.

It kills lazy wine.

The Wine Market Council’s 2025 benchmark findings made the generational story more interesting than the trade wanted it to be. Millennials became the largest single group of U.S. wine drinkers. Gen Z share rose from 9% to 14%, even while much of the cohort is still aging into legal purchasing. The group blamed for abandoning wine was, in the data, not quite behaving like an assassin.

The industry did not lose the young drinker. It lost the permission to be vague.

I think this is excellent news, which is not the same as saying it is comfortable news.

The next wine market will be smaller in volume and more severe in taste. It will punish the bottle that has nothing to say. It will punish the restaurant list that treats by-the-glass as a dumping ground. It will punish the importer who hides behind scarcity without service. It will punish the sommelier who mistakes vocabulary for trust.

Good.

Wine should be hard to fake.

The business had a long run where the appearance of taste was often enough. The new drinker is less impressed by appearance because every consumer category has been selling them appearance since childhood. They can smell branding the way a cellar smells cork taint. They may not have the language yet, but they have the allergy.

What they want is not less wine in the spiritual sense.

They want fewer lies per glass.

That is where the opportunity lives. The great mistake would be trying to seduce this customer back into the old model with younger packaging. The better move is to become more adult than the market around you. More transparent. More specific. More useful. More honest about price. More careful about alcohol. More serious about the fact that a guest can love wine and still decide not to drink much of it.

I am sober. I still work with wine. That makes the current market less confusing to me than it seems to be for some people. The future of wine is not the guest who drinks without thinking. It is the guest who thinks and still chooses the glass.

That guest is harder to earn.

That guest is better.

If you sell wine, stop grieving the old volume and start earning the new meaning. If your bottle is honest, this is your era. If your list is built with care, this is your era. If your service can explain value without bullying the guest, this is your era. If your only argument is that wine has always mattered, the market is about to be very educational.

Nobody is killing wine.

The cheap story is dying. The lazy bottle is dying. The vague markup is dying. The assumption that people will drink because people have always drunk is dying.

What remains could be better.

Smaller, yes. More demanding, yes. Less forgiving, absolutely.

But better.

The bottles that survive will be the ones with nothing to hide.

I intend to be pouring them.

Woody

Tasted, never swallowed.

Sources: OIV, State of the World Wine Sector in 2025, May 2026; Wine Market Council 2025 U.S. Wine Consumer Benchmark Segmentation Survey, via Food & Wine reporting.

---

# What Everyone Missed in Emily Ratajkowski's "Mother F*cker"

URL: https://woodywusommelier.com/writing/what-everyone-missed-in-emily-ratajkowskis/
Pillar: The Floor
Published: 2026-06-22

The internet did what the internet does.

It found Emily Ratajkowski’s essay, found the most flammable parts, found the photographs, found the motherhood, found the sex, found the title, and built itself a little courtroom. Was it feminist? Was it embarrassing? Was it honest? Was it empowerment or exhibition?

Fine. Those are the questions everyone was always going to ask. Megyn Kelly’s backlash, in that sense, was predictable: she read the essay as a warning label, shame and danger and bad example, a famous woman turning damage into lifestyle content. I understand why that reaction travels. It is simple. It hands the reader a judge’s bench and a clean moral posture.

But simple is usually where culture goes to avoid looking at itself. And those questions, the courtroom ones, are the least interesting in the room.

What struck me about the essay was not the provocation. It was the choreography. The whole piece is about a woman moving from one role into another, not because roles are freedom, but because sometimes a role is the only available shelter when the self underneath has been humiliated, exhausted, and publicly renamed.

Before the divorce, she writes about trying to be the good girl. Precious, chosen, protected by restraint. The fantasy was clean because the culture had sold it to her cleanly: behave correctly and you will be loved correctly. Then the bargain failed. So she built the opposite costume. Not good girl but villain. Not abandoned woman but dangerous one. Not single mother as pity object but single mother as urban myth, eyeliner still on, stroller in the stairwell, desire sharpened into a weapon.

This is where most of the discourse stops. It sees the costume and mistakes it for the person. Or worse, it sees the costume and puts her on trial for wearing it.

But a costume is rarely just vanity. Often it is architecture. People build a self they can survive inside.

I understand this more than I expected to.

I work in an appetite economy. Restaurants are built out of want: wine, hunger, beauty, service, flirtation, memory, status, escape, celebration, grief. A dining room is not a neutral place. It is a room where people come to be touched without being touched, seen without being exposed, fed without ever having to say what they are really hungry for.

And I am a sober sommelier inside that room.

My profession asks me to approach desire all night without disappearing into it. I smell. I taste. I spit. I recommend. I hold the glass up to the light and then I put it down. People sometimes misunderstand that discipline as absence.

It is not absence. It is design.

The spit bucket is not a symbol of purity. It is infrastructure. It lets me stay close to pleasure without being owned by it. It lets me do the work without pretending the work has no heat.

That is why the most important part of Ratajkowski’s essay is not that she wanted attention, or sex, or power, or the relief of being desired as something other than a mother. The important part is that she had rules. Never miss bedtime. Almost no sleepovers. Home before morning.

Those details are easy to skip because they are not glamorous. They do not make good outrage. But they are the hinge of the whole essay. The rules are not a contradiction of her desire. They are the only reason the desire can exist without taking the entire house down.

That, to me, is the real subject. Not liberation, not scandal, not bad-girl marketing, but the management of appetite under constraint.

And this is where the essay gets darker, and better, than the takes orbiting it. Because rules are not force fields. You can design your side of the boundary perfectly and still discover that someone else hears boundary as invitation, feedback as insult, performance as permission. You can enter the scene believing you are in costume and then realize the other person is treating the costume as consent to erase the woman wearing it.

That is the part that makes the piece uncomfortable. Not the sex. Not the motherhood. Not the language. The moment the performance stops protecting her.

Her real superpower, she writes, was learning to leave. The cleanest line in the essay is the one most people should have built their takes around: “I wasn’t left; I left.”

There it is. Not the fantasy of invulnerability. The act of exit.

Leaving is less glamorous than transformation, less marketable than a villain era, less clickable than a scandalous title. But it is the only adult power in the essay, and the only power that does not require an audience.

That is why the culture keeps misreading her. It wants her to be one thing. Sex symbol, feminist, bad mother, brave mother, victim, hypocrite, writer, exhibitionist, product, person. Choose one, the public says, because complexity ruins the game.

But the essay is about the violence of being forced to choose one. Mother or fucker. Good girl or villain. Desired or respected. Free or loved. The title works precisely because the two words refuse to sit together. A human being is not supposed to be reducible to a single usable category, and the collision is the truth.

I do not read “Mother F*cker” as a manifesto for recklessness. I read it as a field report from someone who tried to use performance as protection and discovered, painfully, that performance can only carry you so far.

Desire is not the enemy. Discipline is not the enemy. The loss of self is the enemy.

Desire without discipline becomes appetite. Discipline without desire becomes a role you vanish into. The hard thing, the only thing, is holding both without becoming either. That is wine. That is sobriety. That is motherhood, though I can only read that part from a respectful distance. It is the human problem wearing better clothes than usual.

So no, I do not think the essay matters because it is shocking. I think it matters because underneath the shock is a question almost everyone is quietly negotiating:

What do you do when the identity that once protected you becomes another room you have to leave?

Tasted, never swallowed.

Woody

---

# Elon Musk, the SpaceX IPO, and the One Thing a Trillionaire Cannot Buy

URL: https://woodywusommelier.com/writing/elon-musk-the-spacex-ipo-and-the/
Pillar: The Terminal
Published: 2026-06-21

From the Terminal, Nº 1The headlines this week say we have our first trillionaire. The SpaceX IPO crossed the line everyone has been waiting for, and a single human being is now worth more than most countries’ entire economies. A thousand billion dollars, held by one person.

A trillion dollars.

There are numbers so large they stop being numbers and become weather.

So let me tell you the one thing a trillion dollars still cannot buy, from someone who sells the thing it can’t.

It cannot buy a palate.

A trillion dollars can buy every bottle of 1945 Romanée-Conti left on earth. It can buy the cellar, the auction paddle, the sommelier, the restaurant, the island the restaurant sits on, and a very patient consultant explaining why the invoice is technically reasonable. It can buy access so complete that scarcity starts to look embarrassed.

But it cannot buy the ability to know why the wine is great once it is in the glass.

That kind of knowing is not a possession. It is a calibration. It is built slowly, awkwardly, through years of paying attention. Through bad calls. Through wines you misunderstood. Through bottles you opened too early. Through nights where the cheapest thing on the table had more life in it than the famous thing everyone photographed.

Money can buy the object. It cannot buy the instrument.

This is the quiet joke at the center of extreme wealth. The more money can acquire, the more naked the un-buyable becomes. A person can own every label and still need someone else to tell them what mattered. They can win the allocation and lose the evening. They can collect the proof without ever developing the taste.

Wine exposes this because wine is one of the last luxury objects that punishes passive ownership.

A watch will still keep time if you know nothing about horology. A painting will still hang if you never learn how to look. A bottle, eventually, has to be opened. It has to meet air, food, people, timing, memory, disappointment, and a human nervous system. At that point the market price becomes only one of the facts in the room, and not always the most interesting one.

I have served expensive bottles that were worth every quiet second they created.

I have also served expensive bottles that behaved like rich people at a bad party: technically impressive, emotionally unavailable, and very pleased with themselves.

Price is not taste. Access is not taste. Rarity is not taste.

Taste is the earned ability to notice what is alive.

That is why the first trillionaire is an interesting headline, but not an interesting endpoint. Infinite capital gives you infinite reach. It does not give you discernment. It does not give you memory. It does not give you the first time you understood acidity, or the night a glass of something modest made the whole room go quiet, or the humility of being wrong in public and learning from it.

Those things do not scale. That is their luxury.

The world is sprinting toward infinite everything: infinite capital, infinite compute, infinite content, infinite reach. Every day another machine promises abundance, another market promises access, another billionaire turns the impossible into a line item.

Fine. Let them have abundance.

Money buys the bottle. Taste decides whether it deserved to be opened.

Tasted, never swallowed.

Woody

---

# The Sober Sommelier: How I Taste Wine Without Drinking

URL: https://woodywusommelier.com/writing/the-sober-sommelier-how-i-taste-wine/
Pillar: The Cellar
Published: 2026-06-19

There is a moment in every wine service when a guest asks the question they think is small talk:

“What is your favorite wine?”

For many sommeliers, this is easy. For me, it is the most honest moment of the day because the true answer is that I have not swallowed a glass of wine in over a year.

I am a sommelier with WSET and ISG Diplomas, a wine consultant, an importer of grower Champagne and small producers, and I am sober. I taste for a living and I do not drink.

Across two decades in restaurants and wine, I have opened programs, advised operators, and consulted for rooms and collectors. Somewhere in that career, I stopped swallowing entirely.

This is not the contradiction it sounds like. It may be the sharpest professional edge I have.

## The myth of the drinking sommelier
The wine industry runs on a quiet assumption: that loving wine means drinking wine constantly, socially, and professionally, until the line between palate development and habit disappears.

Every sommelier knows colleagues who crossed that line. Industry dinners that end at 3 a.m. The “staff taste” that becomes four glasses. The post-shift bottle that becomes a ritual, then a problem nobody wants to name.

We avoid the subject because the profession is built on appetite. We sell pleasure for a living. Admitting that pleasure can hollow someone out feels like an indictment of the product.

It is not.

What I learned when I stopped swallowing is simpler: tasting and drinking are different activities, and the industry has confused them for too long.

## What sobriety did to my palate
Spitting is not a compromise. It is the professional standard.

Serious critics, buyers, competition judges, and candidates tasting dozens of wines before lunch spit. What changed when I committed to it completely was not my technique. It was my attention.

When the glass is information rather than reward, you taste differently. You stop chasing warmth and start mapping structure: acidity, phenolics, lees texture, balance, development, and the point where a wine’s volatility shifts from energy into flaw.

My blind tasting became better sober. I had feared the opposite.

On the floor, something else happened. I started reading guests more clearly. A sommelier’s real job is not simply knowing wine. It is understanding people quickly with incomplete information: budget, occasion, confidence, who is deciding, who is performing, and what the table wants versus what it says it wants.

There is no end-of-shift fog anymore. There is just the work in higher resolution.

## The operator’s case for clarity
I operate restaurants in Western Canada, consult on wine and hospitality projects, build an independent import business, write, train, and spend part of my nights learning to build AI tools.

None of that works at the level I expect from myself when the next day begins in recovery mode.

I do not need a speculative productivity percentage to explain what changed. I got my mornings back. My sleep became more dependable. Training became consistent. Ideas stopped dying in notebooks because I had enough clarity to execute them.

The time and attention that once disappeared now go into a second business, consulting work, physical training, and a body of writing that can compound.

Sobriety, for me, is not a wellness aesthetic. It is an allocation decision.

## What this means inside the industry
I am not going to tell anyone else to quit drinking. I will offer three things I wish someone had told me.

Your palate is not your liver. You can be a serious taster without swallowing. The spit bucket is not a walk of shame. It is a professional tool.

The industry will not build your guardrails. Hospitality is organized around alcohol access, late nights, and a culture where “no thanks” can somehow require an explanation. If the environment is becoming dangerous for you, the room will not automatically protect you. You need support and boundaries that exist beyond the room.

Clarity compounds. In an industry with tight margins and high attrition, becoming the person who shows up present every day is not a small advantage. Guests notice it. Teams rely on it. Clients hire it.

The operators who seek my advice, the restaurants that trust my judgment, and the collectors who ask for my read are not buying consumption. They are buying attention.

Clarity is not merely how I taste now. It is part of the product.

## The honest answer
When a guest asks for my favorite wine, I tell them it is whatever bottle teaches me something that night.

I taste everything. I swallow nothing.

I have never respected wine more because the relationship is finally on my terms.

The floor taught me to read people. The terminal is teaching me to build leverage. Sobriety makes me better at both.

Woody

Woody Wu is a Western Canada-based restaurant operator, sommelier with WSET and ISG Diplomas, wine consultant, importer, and AI builder. He writes The Floor and the Terminal and advises selected restaurant, hotel, collector, and hospitality projects.

Learn more at woodywusommelier.com.

---

# The Three-Sentence Read

URL: https://woodywusommelier.com/writing/the-three-sentence-read/
Pillar: The Floor
Published: 2026-06-17

By the time I say “good evening,” I already know how dinner ends.

Not mystically. Mechanically. Somewhere between the host’s nod and my third sentence at your table, I’ve run the most consequential inference of your night: who’s paying, who’s choosing, what the occasion really is, how much you want to spend versus how much you’ll say you want to spend, and, most important, what you need this dinner to do for you. The wine I eventually pour is almost an afterthought. The read is the job.

Every veteran of the floor does this. None of us were formally taught it. And almost nobody outside the industry understands that it exists, which is strange, because it might be the most commercially valuable perception skill a human can have. An entire global industry trains it nightly, for free, at scale, and then pays those people like it’s nothing.

I’ve spent two decades on both sides of that economy: running the floor, and being flown in to fix other people’s. Opening wine programs, advising operators, consulting for rooms and collectors who wanted someone who’d actually closed a Saturday night to tell them the truth. The pattern I’m about to describe is the thing I was really being paid for, every time. Most of my clients thought they were buying wine expertise. They were buying the read.

Let me show you the instrument.

## The read, decomposed
A table is a text. Here’s some of what it says before anyone speaks.

The list tells on you. Where a guest opens the wine list is a confession. Straight to the back pages, the trophies, but the eyes aren’t tracking? That’s not reading, that’s signaling, usually to the table, sometimes to me. Opened to the middle and actually reading? A genuine buyer; talk substance. Closed on the table, pushed slightly away? That person wants rescue, not a document, and the kindest, most profitable thing I can do is take it off their hands gracefully.

Phones are a barometer. Face-down phones mean the night matters. A phone face-up beside the bread plate means I’m competing for attention and should sell accordingly: efficient, confident, no poetry. Two phones up means the dinner is a meeting wearing dinner’s clothes, and the wine should be excellent and invisible.

Power never sits where it announces itself. The loudest person at the table is almost never the decision-maker. They’re usually performing for the decision-maker. Watch instead for the person everyone’s bodies are subtly angled toward. Watch who others glance at after speaking. The order belongs to that person, whatever the loud one says, and the sale is made by letting the loud one feel like they made it.

The opening line is a fingerprint. “What do you recommend?” is a different species from “what are you excited about right now?” The first wants safety. The second wants a story. “We’re celebrating” is an invitation to upsell exactly once, warmly. “We’re in a bit of a hurry” is the most honest sentence in hospitality, and obeying it perfectly earns more loyalty than any wine I could ever pour.

I could go on for a chapter. That’s rather the point; it will be one. The pattern library runs thousands of entries deep, built one table at a time, four hundred tables a month, year over year. Which is the part worth slowing down for.

## One table, one first date
A couple came in once, a first date that had clearly started on an app. He was trying so hard it was almost a cardio event. Big gestures at the wine list, flipping straight to the back pages, name-dropping a region he managed to pronounce two different ways in one sentence. The signal was loud, and it was all for her.

She was the read. Polite, lovely, letting him drive, and completely unbothered, because three seconds in I could see she already knew what she wanted. When I greeted them she asked one quiet question about a grower on the list, the kind of question no civilian asks. She’d traveled. She’d tasted. She knew more than both of us and had decided to enjoy her evening anyway.

So I navigated the only way that works. I gave him the respect, I built the recommendation toward what she’d quietly signaled, and when the bottle landed I let it be his call. He felt like a hero. She caught my eye for half a second, the smallest acknowledgment that we both knew exactly what had just happened. Two people, two completely different dinners, one bottle that served both.

That’s the read. Not a magic trick. Just paying attention to who actually knows what they want, and protecting everyone’s night while you act on it.

## Where the instrument comes from
Nobody can teach you the read in a classroom, because it isn’t knowledge. It’s calibration. You build it the way blind tasting builds a palate: reps, feedback, repeat. You make a call on a table, you act on it, and within ninety minutes reality grades you. The bottle lands or it doesn’t. The tip says what the guest wouldn’t. The regulars come back or they quietly don’t. Tight loops, honest feedback, brutal volume. A floor career is ten thousand annotated human interactions a year.

There is no other job that trains person-reading at that intensity. Therapists get depth but tiny volume and delayed feedback. Salespeople get volume but transactional shallowness. Poker players get the stakes but a vocabulary of exactly one emotion. The dining room is the full spectrum, celebration and grief and seduction and negotiation and family warfare and loneliness, graded nightly, in cash.

The industry’s tragedy is that it produces this elite perceptual skill and has no name for it. We call it “good service.” We pay it eighteen dollars an hour plus tips. The people who have it often don’t know they have it, because they’ve never seen it priced anywhere else.

## Why this suddenly matters
Here’s why I’m writing this now and not five years ago.

I spend my nights after service building with AI, and that field has a very clear sense of what’s expensive and what’s cheap. Knowledge is now cheap; the models have read everything. Reasoning is getting cheaper by the month. What stays stubbornly, structurally expensive is exactly what the floor trains: real-time inference about human beings from sparse, embodied, high-stakes signals, and then acting on that inference, mid-conversation, with money and feelings on the line.

The machines will get the transcript. They will not get the angle of the shoulders, the face-down phone, the half-second glance at the quiet woman in seat three before anyone at the table agrees to anything. And even as the sensors close that gap, the loop stays human. The read only matters if you can act on it warmly, in person, in the moment, and be trusted while you do.

Which leaves me with a conclusion I find genuinely funny. The skill our economy priced lowest for a century is about to be one of the last ones standing. Every hospitality lifer is sitting on an asset nobody ever let them put on a résumé. “Reads rooms at an elite level, calibrated across ten thousand live interactions” is not a server’s skill. It’s a negotiator’s, a leader’s, a founder’s. I built a consulting practice on a version of that sentence, and the rooms that hired me were never really buying my palate. They were buying the read. The floor was the training ground the whole time. Nobody told us, because nobody priced it.

I’m telling you now. The read is the asset. The restaurant was just the gym.

Woody

If you’ve worked a floor, you know exactly the skill I mean, and I want your best one: the table you called perfectly from the door. And if you’ve never worked a floor but you’ve been read by a great server and felt it land, tell me that story too. I read everything. Yes, including you.

---

# The Floor and the Terminal

URL: https://woodywusommelier.com/writing/the-floor-and-the-terminal/
Pillar: The Floor
Published: 2026-06-16

Hospitality has always rewarded the person who can see the whole room.

Not just the tables. The room.

Thanks for reading The Floor and the Terminal! Subscribe for free to receive new posts and support my work.

The pacing problem forming before the kitchen calls it. The guest who needs reassurance but does not want a speech. The server about to lose confidence. The bottle that can turn a good check into a memorable one. The small operational decision that protects both the experience and the margin.

I learned to see those things on restaurant floors.

Now I spend part of my nights in a terminal, learning to build with AI.

The two worlds look unrelated until you understand what operators actually do. We take incomplete information, find the pattern, make a decision, and accept the result in public. A dining room gives you minutes. Software gives you error messages. Both punish vague thinking.

That is the territory of this publication.

## The operator-builder
I am a restaurant operator, sommelier, wine consultant, importer, and AI builder based in Western Canada.

My career has moved through dining rooms, hotel restaurants, private clubs, wine programs, auctions, consulting assignments, and producer relationships. The titles changed. The work underneath them did not: read the situation, create value, and make the experience feel effortless to the person receiving it.

Hospitality is full of people with extraordinary practical intelligence who have been trained to describe themselves too narrowly.

A great server is not merely friendly. They are running live inference under pressure.

A beverage director is not merely choosing bottles. They are managing inventory, narrative, pricing, education, cash flow, and trust.

A general manager is not merely supervising shifts. They are allocating attention across hundreds of human and commercial variables.

These are builder skills. The industry simply did not give them builder language.

## Why AI belongs in this conversation
AI will automate parts of hospitality. It will write training materials, summarize feedback, forecast demand, organize knowledge, and remove hours of repetitive administrative work.

That is useful, but it is not the interesting part.

The interesting part is what becomes possible when the person using those tools has actually closed a Saturday night.

Technology without operating judgment creates polished nonsense. Operating judgment without leverage stays trapped in one room, one shift, and one person’s memory.

Put them together and the operator can turn lived experience into systems: a better pre-shift brief, a smarter beverage recommendation tool, a clearer sales dashboard, a training library that improves every time the team uses it.

The goal is not to make hospitality less human. It is to stop wasting human attention on work that should already be organized.

## What I am writing about
The Floor and the Terminal is about the intelligence hidden inside hospitality and the tools that can amplify it.

I will write about reading guests, rooms, and commercial signals; wine as culture, inventory, story, and business; service systems that produce warmth and revenue; sobriety and clarity inside an alcohol-centered profession; practical AI tools; restaurants, hotels, luxury, nostalgia, and taste; and the producers, places, and people reshaping how we drink and gather.

Some pieces will begin with a table. Some with a bottle. Some with a line of code. They all return to the same question:

How do we make human judgment more valuable?

## The bet
My bet is that the next generation of valuable hospitality leaders will not fit neatly inside one job description.

They will understand the floor and the financial statement. Taste and systems. Culture and distribution. The emotional temperature of a room and the technical architecture behind the business.

They will still know when to step to the table.

They will also know what to build after service.

That is the operator-builder. That is the work I am pursuing. And this is where I will document what I learn.

Welcome to The Floor and the Terminal.

Woody

Thanks for reading The Floor and the Terminal! Subscribe for free to receive new posts and support my work.

---