Everyone has a war college now: what an incubator is for when advice costs nothing

In June 2026 ChatGPT crossed a billion monthly active users, the fastest any application in history has reached that number, faster than Google Maps, TikTok or Instagram. It now takes in something like two and a half billion prompts a day, and roughly a third of people start a search by asking an AI instead of a search box. Buried inside those numbers is an event almost nobody in our industry has said out loud, because it is inconvenient for most of us: startup advice just went to zero.

Not cheaper. Zero. Every founder alive now carries an infinitely patient, tireless, widely-read advisor in their pocket, one that has ingested every business book, every accelerator's playbook, every teardown and pitch guide ever written, and will talk it through with them at two in the morning for the price of a subscription. The thing that incubators and accelerators and business schools have sold for forty years, guidance from people who have read more than you, is now a free utility, like tap water and a dial tone.

We run an incubator. So we had to sit with the most uncomfortable question our own business has ever faced, and answer it honestly: if advice is free, what is anyone paying us for?

The frame everyone else is arguing in is already dead

Pick up any comparison of incubators and you will find them fighting over the wrong thing. Better mentors. A stronger curriculum. More sessions, sharper frameworks, a fatter alumni handbook. Every one of those is a claim to give you better advice than the incubator down the road. And every one of those claims is now competing, directly, with a free tool that gives more advice, more patiently, at more hours, than any human mentor network can. That is a race to the bottom of a well that has already run dry.

The interesting question is not which incubator has the best advice. It is what survives the moment advice becomes worthless, and the honest answer took us a while to reach, because it required admitting that most of what our own industry sells is already gone.

There is an old idea from software strategy that gets you most of the way there. Smart companies, the argument runs, work to commoditise their complements: make the thing next to your thing cheap, so all the value flows to the thing you actually hold. The AI labs have, without particularly meaning to, commoditised advice, the complement that sat next to building a company. And when one half of a pair goes to zero, the value does not vanish. It rushes into the other half. Advice was always just the complement to the thing that was hard. The hard thing is building. Advice going free does not devalue execution. It prices it, for the first time, at exactly what it is worth, which is everything.

No plan survives contact

In 1871, fresh from taking apart the Austrian and French armies, the Prussian field marshal Helmuth von Moltke wrote the line every operator eventually learns in their body: no plan of operations extends with any certainty beyond the first contact with the enemy's main strength. The plan is the easy part. It is clean, it is written in the quiet before anything is at stake, and it is obsolete the instant the world pushes back.

Advice is the plan. It has never been more abundant and it has never been worth less at the one moment that decides everything, which is contact. The AI in your pocket will write you a flawless plan for your launch, your pricing, your fundraise. It cannot be in the room when the launch goes sideways, when your one enterprise pilot goes quiet for eleven days, when a competitor ships the exact feature you were building and you have four hours to decide whether to change course. Contact is where companies are actually won and lost, and contact is precisely where counsel evaporates.

Here is the part people miss about Moltke, and it is the whole point. His army did not win because its plans were better. Everyone had war colleges and staff manuals by then. It won because of what it did after the plan broke. The Prussian staff drilled judgment at the point of contact, trained officers to improvise toward the intent when the map stopped matching the ground, and put experienced commanders forward with the troops rather than lecturing from the rear. They were good not at planning but at the moment the plan died. That is the only skill that was ever scarce, and it is the only one a language model cannot hand you, because it is not information. It is presence under pressure, next to someone who has been there.

Everyone has a war college now. It is in their pocket and it is free. Almost nobody has someone who has been shot at standing next to them at contact.

So why would you pay us anything?

Let us take our own medicine and ask it flat out. If what an incubator sells is advice, you should not give us a penny. You have the plan generator already, and ours is not better than yours. Any accelerator still fundamentally in the advice business is a wrapper around a free tool, and it is charging you for the wrapper. We think a lot of them know it, and are hoping you do not.

We are not in the advice business. We are in the contact business. What Moonlabs sells is two people who have built and sold companies sitting at the bench next to you for twelve weeks, not to tell you the plan, but to be there when the plan meets the world. When the pilot goes quiet, we have sat through that exact silence and we know which of the eleven days it is safe to wait and which one you break and call. That is not advice. It is a nervous system for the moment advice runs out, and it is the one thing in this business that got more valuable, not less, the day the tool went free.

The bench, not the dais

If you want to see the shape of this, do not look at a business school. Look at a Florentine workshop five hundred years ago.

When a boy of fourteen wanted to become a painter in 1466, he did not enrol on a course and sit beneath a dais while a master lectured about pigment. He was placed in a bottega, a working workshop that took paying commissions, and he learned by grinding the colours and then by painting the actual picture the master had been hired to deliver, at the next bench, while the master painted beside him. Leonardo da Vinci learned that way in the workshop of Andrea del Verrocchio, and the account everyone remembers is that he painted one of the angels in Verrocchio's own commissioned Baptism of Christ, real work, sold work, done alongside the master on the master's own job. Nobody handed him a plan. He was put into contact, on live work, next to someone who could already do it, and he was corrected in the doing.

That is the oldest and most reliable way humans have ever transferred a hard skill, and it survives every technology because it was never about information transfer in the first place. The lecture from the dais is the advice model, one voice, general, one-directional, and it is exactly the part a machine now does better and for free. The bench is the operating model, specific, two-directional, on live work with real stakes, and it is exactly the part a machine cannot touch, because the value is in the presence, not the words.

Moonlabs is a bottega, not a lecture hall. Twelve weeks, a working product shipped, the two of us at the next bench on your actual company while it is being built, not narrating from a stage. We charge for the bench, because the dais is free now, and pretending otherwise is how you sell people something they already have.

What actually survives the collapse

Strip an incubator down to what it really offers and you get three things: advice, network, and money. It is worth being clear-eyed about which of them the last twelve months killed.

Advice is gone. Fully commoditised, and any programme leaning on it is already a museum piece that has not read its own valuation. Network survives, but changes: warm introductions still matter, though a founder who cannot get a first meeting on their own in 2026 has a sharper problem than an incubator can fix. Money survives, obviously, but money is the most commoditised thing of all, an undifferentiated commodity that competes only on price and terms and has never, on its own, taught anyone to build.

What is left standing, the only thing that got scarcer rather than more abundant, is contact. An operator at the bench with you when the plan breaks. Everything else in the incubator model is now either free or fungible. That single thing is neither, and it is the entire reason we structure Moonlabs the way we do: a tiny cohort, founders we can actually stand beside, twelve weeks of live building rather than a term of seminars, and terms that only pay off if what we built together actually works.

The billion people who opened an AI assistant this month did not devalue what we do. They clarified it, brutally, by making the cheap part of our industry free and leaving the expensive part standing alone in the open where everyone can finally see it. Advice was never the thing. It was the plan, and the plan does not survive contact. We sell the person beside you when it doesn't.

Louis O'Connell-Bristow and James Freestone are the co-founders of Moonlabs, the operator-led AI incubator and academy. They previously built the home.co.uk, Homemove and homedata.co.uk stack. Moonlabs backs a small cohort of founders and builds alongside them for twelve weeks. Site: moonlab.ventures.