Operators writing it down.
Essays from the Moonlabs operating tables. AI-native building, commercial leverage, fundraising, what is actually working on the ground. No hot takes, no link-bait. Just the things we wish someone had handed us two years ago.
The curriculum with no textbook: teaching a field that reinvents itself every quarter
Any fixed syllabus for AI engineering is obsolete before the cohort graduates. The tool we teach in week one is often superseded by week twelve. So the Moonlabs Academy has no textbook. Here is what we teach instead.
The candle and the token: what happens when the price of thinking collapses
In 2023 a million tokens from the best model on earth cost thirty dollars. Today a better model costs under fifty cents. The price of machine cognition is falling roughly tenfold a year, and almost everyone is reading the wrong history.
The first week in the Incubator, hour by hour
Most incubators waste their first week on onboarding and vision workshops. We ship something live by Friday. Here is what the opening five days of a Moonlabs Incubator actually look like, hour by hour.
The walls of China are going up: three scenarios for UK AI sovereignty
This week Beijing began fencing its open models. Washington fenced its closed ones weeks ago. Most people are reading this as a trade story. It is an arms-control story.
Hand Claude Code a red test suite: the bisect, fix and PR loop
A regression turned twelve tests red and nobody knew which commit did it. So we handed the failing suite to Claude Code with a tight brief: reproduce, bisect to the guilty commit, fix the cause not the symptom, get the suite green, open the PR.
Why twelve, and why twelve weeks: the constraints behind the Academy
Every number in the Moonlabs Academy is a deliberate constraint. Twelve students, not thirty. Twelve weeks, not a year. A shipped product every week, not a final project. This is why we chose each one, and what breaks when you get them wrong.
The compute fence: Pax Silica, cloud export controls, and where UK founders now sit
Two weeks ago the story was that a government could switch off a model. This week it is bigger. With the second Pax Silica summit and a US bill that treats renting a cloud GPU as an export, the fence has moved down a layer to the compute itself.
The competitor teardown: turning a rival's YouTube channel into a feature-gap spec
A competitor had forty product videos on YouTube and no public docs. So we had Claude watch all of them, sample a frame every five seconds, and hand us back a structured spec of every feature they shipped and every one we were missing.
Open weights are the only model they can't ban
A government can switch off an API in ninety minutes. It cannot recall a weight that has already been downloaded to ten thousand servers. After the Fable shutdown, open weights stopped being an ideological choice and became basic business continuity.
A world without frontier AI: the moral arithmetic of the closed lab
For three days in June it was possible to imagine that the best AI you can buy might simply stop being available to you. This is a thought piece on how to live and build in that world.
The kill switch was always there: what the Fable shutdown tells the UK
On 12 June a US export-control order took Anthropic's two best models offline worldwide in ninety minutes. If you build a British company on frontier AI you do not control, you have just watched your future get switched off from another country.
Twelve deliverables — the anatomy of a Moonlabs cohort portfolio
Every Moonlabs Academy graduate leaves with the same twelve artefacts. This essay walks through what each one looks like and why we chose it over the obvious alternatives.
Why sponsoring an engineer through the Academy beats hiring one
For most UK companies in 2026, the cheapest way to get an AI-native engineer onto the team is not to hire one. It is to send someone you already have through the Moonlabs Academy. Here is the maths, the precedent, and the case where it does not hold.
Why the Incubator charges a £999 commitment fee
We added a £999 commitment fee to the Incubator agreement in 2026. It is credited back in full against the success fee at round close — so if you raise, it costs you nothing on net. This post explains why we did it.
From idea to investor in ten weeks — a cohort case study
One of our recent Academy students went from "I have an idea" in week one to "I have a term sheet" in week ten. The story is unusual but it is not magic. Here is the actual week-by-week.
What investors actually see in a Moonlabs deck
Most pre-seed decks are theatre. The deck we coach Moonlabs cohort founders towards is the inverse. This essay walks through the eight slides, what each one contains, and what investors take away from it.
Why most great ideas never ship
Most of the great ideas we hear in founder calls are not new. They have been parked for years. This essay walks through the five reasons we have seen ideas stay parked, and the moves that unpark them.
The eval discipline that keeps AI products honest
Most AI products in 2026 are shipping without proper evals. The teams that have figured it out are quietly running circles around the ones still relying on "looks good to me." Here is the discipline that actually works.
How to hire the AI-native operator
The job market for AI engineering hires in 2026 is broken in specific, fixable ways. This is the hiring playbook we run when companies in our network ask us to help them screen candidates.
A day in the life of a Moonlabs cohort student
What twelve weeks at the Moonlabs Academy actually looks like — a typical day from 9am Monday to the Thursday evening review. Written for anyone considering applying and trying to picture the reality.
What the Academy actually teaches, week by week
We are asked weekly what the Moonlabs Academy curriculum looks like up close. This is the full twelve-week breakdown, the books on the shelf, the projects shipped, and the moments we keep watching change students.
Inside a Moonlabs demo day
A walkthrough of what actually happens on Moonlabs Academy demo day — the programme, what each graduate presents, the audience, and the outcomes we expect.
Moonlabs Academy vs jointakeoff vs YC Startup School
We get asked the comparison weekly. Here is the honest version of how the three options early-career AI builders actually consider stack up, with what each one is good at and where it falls down.
How to write a Moonlabs application that gets a yes
We read every application personally. We say no to about nine in ten. The yeses share specific patterns and the nos share specific failures. Here is the honest list, so you can write yours sharper.
Hiring the graduates who can actually ship
We are now hiring out of the Academy at salaries that look senior on paper. The reason is not that we are overpaying juniors. The reason is that the work has changed, and the graduates who can do the new work are worth what we pay them.
Postgres + pgvector vs a real vector database
The default architectural decision for an AI-flavoured product in 2026 is between extending Postgres with pgvector or running a specialised vector database. The answer for the kind of company we build is almost always Postgres.
Why we say no
We turn down roughly nine in ten applications. Most accelerators would call that a problem. We think it is the most important number we report. Here is the actual breakdown.
Cursor, Claude Code, and the ten-times operator
The 10x engineer of the 2010s is now the 10x operator of the 2020s. The tools that produce them are not secret. They are sitting on every laptop in our office. The interesting question is which ones are actually load-bearing.
The forty-thousand-pound MVP is a myth in 2026
A founder asked me last week what a "fair" budget was for an MVP. The honest 2026 answer is much lower than the number everyone is quoting. Here is the actual breakdown.
Our prompt engineering folder structure
Most teams treat prompts like config files — flat, stringly-typed, lost to history within months. We treat them like code. Here is the actual folder structure we use across our companies.
How we raised five million without warm intros (mostly)
The standard fundraising advice — "warm intros only, never cold-email an investor" — was written by people who already had the network. We did not have the network. Here is what worked.
Inside the Moonlabs data room
The single most common reason a UK seed round falls over in diligence is a half-built data room. Here is the exact structure we now use, with the files that matter and the ones investors quietly skip.
Why Derby is the quiet bet for AI builders
London is loud, expensive, and full of consultants. Derby is quiet, cheap, and full of operators. We picked Derby on purpose. Here is the argument.
What we look for in a co-founder pair
Co-founder pairs that fail mostly fail in predictable ways. The strong ones share a small set of properties that separate them durably from the weak ones. Here is the list.
Coding is not the bottleneck anymore. Commercials are.
We have spent six months teaching student cohorts and the pattern is now unmissable. Every twenty-year-old in the room can ship a working SaaS in a weekend. Almost none of them can sell it, price it, or raise on it.
The quiet death of the dev shop
There are roughly twelve thousand digital agencies in the UK. Most of them have a margin model that depends on the cost of producing software holding steady. It has not. The reckoning is two years out.
The compounding stack — how home.co.uk, Homemove, and homedata work as one company
We run three businesses across the same vertical. Most people see them as separate. We do not. This essay is the architecture of the compounding stack and why we built it this way.
Why we run an incubator, not an agency
Every dev shop in the country is about to discover their billable hour is worth half what it was last year. The smart ones will stop selling hours and start selling outcomes — and the very smartest will stop selling outcomes and start owning them.
Your next chapter starts here.
Tell us about the company you want to build. If we’re a fit, we’ll get back within a week.
© 2026 Moonlabs Incubator. All rights reserved.