Why twelve, and why twelve weeks: the constraints behind the Academy

People assume the shape of the Moonlabs Academy is arbitrary. Twelve students, twelve weeks, a product shipped every week, taught by the two of us in Derby. It sounds like the kind of thing you land on because it reads well on a page. It is the opposite. Every one of those numbers is a constraint we chose on purpose, and most of them cost us money or scale to hold. This is the reasoning behind each, because the reasoning is the actual product.

Why twelve students, not thirty

The obvious business move is to run bigger cohorts. Thirty students at £6,000 is a far better spreadsheet than twelve, and every education business we looked at when we were designing this had drifted upward in class size for exactly that reason. We capped at twelve and we intend to keep capping at twelve, and the reason is the thing we are actually selling.

What makes the Academy work is that the two of us are in the room, on your specific project, giving you feedback while it is still warm. That is not a lecture. It is closer to an apprenticeship, and apprenticeship does not scale the way a lecture does. Past about a dozen people, we stop being able to hold every student's project in our heads at once. We stop knowing, on a Tuesday, that you shipped something shaky on Friday and need a nudge before you compound the mistake. The teaching degrades into content delivery, and content delivery is precisely the thing that models already do better than we do.

Twelve is the largest number at which both founders can still carry every student's context in their heads for twelve weeks. Thirteen would be easier to sell and worse to be inside. So we hold the line at twelve, and we treat the smallness as the feature it is.

Why twelve weeks, not a year

The duration cuts the other way. Twelve weeks is short for a serious education, and the instinct of most institutions is to stretch. A year lets you cover more, charge more, and feel more substantial. We deliberately did not.

A twelve-week clock does something a year cannot: it forces every week to matter. There is no room to coast through a soft middle, no term where nothing ships, no time to hide. The compression is the pedagogy. It mirrors the actual condition of building a company, where you never have the luxury of a slow year and the constraint is always time. A student who learns to produce real output under a tight clock has learned the thing that transfers. A student who had a comfortable year has learned something that does not.

There is a harder reason too. Twelve weeks is roughly how long we can give this level of intensity before it stops being sustainable for us as operators who also run companies. We would rather do twelve weeks at full commitment, three times a year, than a diluted year where the attention thins out. The bound is honest, and honesty about it makes the programme better.

Why a product every week, not a final project

Most courses build toward a capstone. One big project at the end, assessed once, forgotten shortly after. We ship something every single week instead, and this is the design decision we defend most fiercely.

A capstone teaches you to plan a project. Weekly shipping teaches you to finish, over and over, until finishing is a reflex rather than an event. The difference matters more than it sounds. The single most common failure mode we see in technical people is not that they cannot build. It is that they cannot close, cannot cut scope, cannot get the imperfect thing in front of a real person and move on. Twelve consecutive weeks of "it has to be live and in front of someone by Friday" rewires that. By week six the fear of shipping is gone, and once that fear is gone a person is a different kind of operator for the rest of their life.

Weekly cadence also gives us twelve honest data points on each student instead of one. We see how you respond when something breaks on a Thursday night. We see whether you cut scope intelligently or thrash. You cannot fake twelve weeks of shipping the way you can polish a single capstone, and the students learn far more from twelve small real deadlines than from one large artificial one.

Why operators teach it, not educators

We teach the Academy ourselves. Not guest lecturers, not a hired faculty, the two people who built and run the companies. This is expensive in the most literal sense: it is our time, and our time has a high alternative cost. We do it anyway because the alternative produces graduates who know the theory and cannot do the thing.

The closest working model is medicine. Doctors do not learn medicine from people who read about medicine. They learn on rotation, on real cases, next to someone who has treated a thousand patients and can tell them in the moment what they are about to get wrong. The three pillars we teach, Coding, Commercials and Investment, are all skills where the gap between the textbook and the room is enormous. You can read every article ever written about a first sales call and still fumble your first one. What closes the gap is someone who has made that call a hundred times sitting next to you while you make yours. That person has to be an operator, because only an operator carries the tacit knowledge that never makes it into the article.

Why fit is the filter, not fee

The last constraint is the one people find most counterintuitive. The Academy costs £6,000, and yet the price is not how we decide who gets in. We select on motivation and fit, and there are funding routes, including sponsorship, for people who are right for the programme but cannot write the cheque. We turn away people who can pay and admit people who cannot, and we do it on purpose.

A cohort is a shared organism for twelve weeks. One wrong person, someone disengaged, or there only because they could afford it, drags the energy of the entire room down. Twelve motivated people pull each other up; eleven motivated people and one passenger is a worse experience for everyone, including us. If we let ability to pay do the selecting, we would be optimising for the wrong variable and degrading the one thing the small cohort exists to protect. So fee funds the programme, and fit fills it, and we keep those two jobs separate on purpose.

The constraints are the curriculum

Put these together and a pattern emerges. Small enough that we know every project. Short enough that every week counts. Frequent enough that finishing becomes a habit. Taught by people who have actually done it. Filled by fit rather than fee. None of these is the cheapest or most scalable option available to us. Each is a deliberate choice to protect the thing that makes the programme work, at the cost of the thing that would make it bigger.

We are occasionally asked when we will scale the Academy, run larger cohorts, franchise the model, add a self-paced tier. The honest answer is that most of those moves would quietly destroy the specific value we built, because the value lives in the constraints, not despite them. We would rather run something small that changes twelve trajectories a cohort than something large that changes none of them very much. The numbers are not arbitrary. They are the argument.

Louis O'Connell-Bristow and James Freestone are the co-founders of Moonlabs, the operator-led AI incubator and academy, and previously built the home.co.uk, Homemove and homedata.co.uk stack. The Academy runs twelve students per cohort over twelve weeks. Application is on fit, not fee. Site: moonlab.ventures.