Skip to main content
BriX Consulting

The Best Tool Still Needs a Black Belt

Why we won't hand you our SW Factory, and what a judo teacher has to do with it.

Michele Brissoni
The Best Tool Still Needs a Black Belt

[Trenčín, October 2026] On Friday afternoon the phones went into the backpacks. They stayed there for two days.

Six hours a day on the mat, no-gi grappling, in a hall full of sixteen-year-olds, twenty-somethings and thirty-somethings. And two people who were out of the ordinary: Ivica and me.

Ivica is a Serbian friend, a judo teacher with a 6th dan, and a doctor. Medicine is where he spent his working days. The martial arts are where he spent everything else, practising and teaching, and he holds more black belts than I can list. For him, as for me, those belts are not status symbols. Our lives have gone into studying and improving ourselves, and the belt is a small ornament that reminds us why we are here: to learn, every day.

We drilled and we rolled, and the young ones came at us hard. Our hearts reach the anaerobic threshold sooner than a twenty-year-old’s, and we were gassed out long before they were. We kept going anyway, like our trainer Leszek, an old-school grappler, now over forty. Between rounds we talked for hours about philosophy, spirituality and religion, and we recognised ourselves in a book about grappling after forty, Old Grappler Handbook by Rick Ellis. At our age, 50 plus, you feel every session in the skin. But it is used skin. It has seen training, bruises, cuts and fractures, and like my tattoos it carries a story. At home I am not only the oldest in my dojo: most of the people I train with have worn a uniform, and they are young and well trained. Nobody is joking. We go down hard, the bruises are real, and getting up the next morning is always an epic undertaking.

Here it was different. Standing among people my children’s age, working on the art I have spent years teaching, I understood that all that teaching had borne fruit. Outside the hall, most of us walk with our heads bent over a screen, at least some of the time. Inside it, for two days, the screens stayed in the backpacks. Everybody was there to learn and practice.



The number, and the reason behind it

On October 2, Anthropic announced the Claude Frontier Academy, backed by a commitment of $100 million to train 10,000 engineers by the end of 2027. They are called Frontier Deployed Engineers, and their job is to take Claude from an idea to a system in production inside a real company.

Read it slowly. The company that makes one of the best tools in the world is spending its own money to train people who do not work for it. And it is a target, not a result: the first badges are expected in early 2027, and the announcement gives no breakdown of the spend.

The number is not what held my attention. The reason behind it did.


What the announcement actually says

The details matter more than the figure.

The engineers are not Anthropic’s hires. Their employers nominate them. The first cohorts include people from Accenture, Bain, Capgemini, Commonwealth Bank of Australia, Deloitte, McKinsey, Morgan Stanley and Novo Nordisk, and each arrives with a named Claude project to lead once back at work.

The program, Anthropic says, borrows its structure from medical training: learn beside practitioners, work on realistic cases, be assessed before you operate alone. Ivica, a doctor, knows that model from the inside. For the engineers it means a multi-day course in person, a simulated enterprise deployment from the choice of the use case through security review to handover, and a graded practical. Those who pass spend 12 weeks leading a real use case in their own organisation, supported by Anthropic engineers, and are assessed again.

Two details stayed with me. The first is who is asked to apply: hands-on software engineers with strong fundamentals. The second is Anthropic’s own diagnosis.

The people who can make AI work inside a real business, it writes, have become “the hardest talent to find.”


Anthropic says it. And it pays for it.

Anthropic does say that a handful of deeply skilled people account for a disproportionate share of what AI delivers in a company. It then does something more telling: it puts $100 million behind them. My reading, and it is mine, not theirs:

when the maker of one of the best tools in the world pays to grow the people around it, those people are part of the result.

I had been looking at the same truth for decades in my Dojo, and for two days on a mat. Between the potential and the real, the only difference is training with competent people who guide you on the field: a sensei in a dojo who helps you become a black belt, in judo, in jiu-jitsu or in any craft.

What Anthropic is doing is what I have been doing for years in the martial arts, with friends like Ivica, Maros and Martin, and with my senseis in Italy, like Sergio. Sergio is past eighty. Until a few years ago he was still rolling on the tatami, and even now he keeps training, inside the constraints of his own body.

Back to the first detail. What is really remarkable is that Anthropic asks for strong fundamentals at the door. The announcement does not say where those come from, or why it matters so much. DORA’s 2025 research offers a clue: AI amplifies what an organisation already is, its strengths and its dysfunctions alike. And an organisation is, first of all, its people. The stronger they are, the greater the result.


Why we won’t hand you our SW Factory

This is the promise I made at the end of the last article, so here is the answer.

A factory is made of your people, your processes, your gates, your standards and your regulators. If we handed you ours, you would be running our habits, shaped by another company’s identity. That is why we chose not to sell our SW Factory as a SaaS. The glue between your people and your systems has to be yours, owned by you. A recipe cannot be borrowed.

So we do something different. We build your own factory with you, on your processes, and we train the people who will run it. Anthropic’s Academy teaches how to build with Claude, which is what a vendor’s academy should do. We know what the fundamentals are: I have practised them for two decades in The SW Craftsmanship Dojo®, and we bring all your people up to a high level of them. Only then do we teach the role itself. Frontier Deployed Engineer is Anthropic’s own credential, and we do not issue it. We teach the craft underneath: the engineer who owns AI in your company from the first use case to daily operation, whichever LLM provider you choose. Fundamentals do not come in a box, and they do not belong to one provider.

Many agile and DevOps transformations followed the book to the letter and still missed the practice. Now AI arrives on top of that gap, and your people are asked to learn the practice of agile, DevOps and AI in one go: a triple somersault with no mat underneath. That is why we do it differently. People who have bled on the field know where the traps are, and can walk your people through them one step at a time, unharmed.

This is our promise. We do not hand you our SW Factory. We grow your people, so that you end up with your own SW Factory and with people who know how to drive it, without needing a seat in a program like Anthropic’s Academy, where engineers are nominated and not everyone gets in. What we want is what the mat taught me:

people in your company who become elite performers, and who stay that way.


The goulash and the tea

On Saturday evening the hall went to dinner. Everyone celebrated: goulash, rivers of beer, long tables. Ivica and I took a plate of vegetables and a Japanese green tea. We are both vegetarian, neither of us drinks, and both of us have pointed our lives at a place where the pleasure is watching others flourish. Our own fruit has already been given. We talked about the best ending we can imagine: one last roll on the tatami.

The young ones had spent two days learning from people who had paid for the lesson with their own skin. Anthropic has just attached a number to that kind of learning, and the number is large. Its Academy will train 10,000 engineers on Claude, each nominated by an employer. It is a serious investment, and I respect it. But a program with nominated seats will always leave most of the people who need the learning outside the room.

We believe it should work the other way round. The learning should come to your people instead of waiting for them to be chosen: on your own processes, inside your company, next to someone who has been on the field and stays beside them until they can drive alone, on fundamentals that hold whichever provider you pick. A dojo does not ask who nominated you. It asks whether you showed up to train.

Who in your company is training next to someone who has already paid for the lesson?

→ Tell me who your people train with


Next article:

a conversation with Jeff Bailey, principal engineer, on how an engineer chooses the way he works with AI.


Sources

  1. Anthropic, Claude Frontier Academy (2 Oct 2026)
  2. BriX, #106 Your AI Is Going Hybrid. Who Holds It Together?
  3. DORA 2025, State of AI-assisted Software Development
  4. Rick Ellis, Old Grappler Handbook (2026), Audible listing