Humans and machines learn and grow together

Humans and machines learn and grow together here: when a piece of work teaches us something, the lesson changes the skills, the positions or what we teach, and the next project runs on the change. A lesson that is written down and changes nothing went nowhere, exactly like an action nobody took.

How a lesson becomes a change

Every piece of work teaches something, and we capture it at the moment rather than in a meeting weeks later. Every note names where it goes.

The three destinations are the point. A lesson that belongs in the skills becomes a rule that fires on its own, so nobody has to remember it. A lesson that belongs in a position changes what we argue, and sometimes what it reveals is a gap, which is a stronger finding than a fix. A lesson that belongs in the learning programme is the admission that no system would have helped and a person simply needed to know more.

Naming the destination costs a few seconds and is the whole difference between a log and a loop.

Two limits

Routing is not applying. Somebody still has to sweep the notes and make the change, and a note nobody acts on went nowhere, exactly like an action nobody took. That step is deliberately human, because deciding what is a real pattern rather than one bad afternoon is a judgement call, and an agent rewriting our conventions on the strength of a single run is the failure we are most careful to avoid.

Capture is the step most often skipped, because the moment something is fixed the urge is to move on. That is precisely when the note is most valuable and least likely to be written.

It already runs on everything

The skill that does this is mxs-learn, and it is a studio skill rather than a delivery one on purpose. A research round, a pitch or a workshop teaches as much as a build, and used to have nowhere to put it. The mechanism runs on any work, whether or not it touches a build, which makes this the one principle already in place across the whole studio.