capslock is an ai systems lab that builds from inside real companies. these are the notes we keep while doing it — what we make, what breaks, and what surprises us, published as we go. not announcements; just the work, with the door open.
we built a machine to stop us from believing our own agent's code works. the rule underneath it: a model can never be the judge of whether the model succeeded.
the honest version of working with an AI on visual things: drop one thing, look, adjust — and why that beats writing a big spec you can't picture yet.
the usual way to get a composition out of an image model is to describe it and re-roll until the dice land. give it the geometry instead — and stop asking the slot machine to be your camera operator.
we gave three companies' nightly reports three different personalities — and the personality, not the automation, turned out to be the point.
we've been quiet for a while, building. this is the lab opening the notebook — starting with the bet behind all of it: the useful part of ai is almost never the part everyone's excited about.
most productivity tools quietly work against you — more pressure, more guilt. we built the opposite, and learned that friction is emotional before it's operational.
something we figure out for one company becomes something we can apply at the next — not by copying it whole, but by reusing the parts that fit.
the building could happen anywhere. what el salvador gives us is the setup — incentives, proximity, and a reason to grow this here, in place.