Michel Maddalena

01Who is writing

I am not an engineer. I have no degree, no lab, no team, and no funding.

What I have is a particular kind of stubbornness — the kind that does not stop when something does not work, but asks why it does not work, and measures the answer.

That produced a 193-page technical document, a complete simulation campaign, and a public repository anyone can verify. When the numbers were wrong I corrected them and said so — nine published figures were retracted over the course of the campaign, each with the measurement that killed it written next to it.

Almost none of them were arithmetic errors. They were correct numbers measuring the wrong quantity — a window that included the robot standing still, a contact counted as a step, a coefficient produced by reading the wrong actuator. That failure mode is more dangerous than a bug, because it looks like a result.

Fourteen years of classical piano taught me that precision is not the enemy of expression — it is the condition for it. Years of running a language school, and teaching at a US Air Force base, taught me that the gap between what you intend and what the other person receives is where most things fail.

I am from Friuli-Venezia Giulia, in north-eastern Italy. I work alone, with an AI team I treat as colleagues and push back against when they are wrong.

02Before this

A linguistic background, not a technical one — and a decade of running something small and making it work.

03How it is built

Every module prints the mass of the model it loaded, with an assert. Measured forces are reported with their closure check. RMS and duty-cycle figures are computed over the steady-state window and the window is declared. Results are marked [C] when computed in simulation and [A] when assumed and pending hardware.

None of that is method learned in a classroom. Each of those conventions exists because something went wrong first — a batch that ran at the wrong mass, a force that did not balance, a figure computed over a window that included the robot standing still.

04What is public, and what is not

The simulation campaign is open under MIT licence for one reason: a claim nobody can check is worth nothing. It is the evidence, and evidence has to be inspectable. That is why it carries the locomotion work and nothing else.

The technical design document, the product architecture and two subsystems central to the product are not public and will not be. They are reserved for a conversation under agreement.

05What I am looking for

A manufacturer with a body, a supply chain and customers — and no legs. The engineering that comes next is CAD, prototype and industrialisation: the part a production company already does every day, and the part I cannot do alone.

What is on the table is a licence for a defined field of use, a design-authority role during development, and the platform architecture that turns one machine into two markets and a decade of revenue.