i built the parts of my brain that don't work well, outside my head.
not a notes app. one fixed loop that takes a thought i've frozen on and hands back a move small enough that i'll actually make it — then watches which moves i finish, and leans toward those.
if i can't visualize it, i can't do it.
the moment it exists for is the freeze.
not “capture a thought.” the instant an open-ended thing feels expensive to get wrong, and i stall. break the sentence, get one reversible move. this is a real entry — press it.
the wall · closed loopsthe whole design is in the shape of one figure.
two of my systems are top-0.1%. the one that holds state while i think sits between them, at the 42nd percentile.
reasoning
speed
memory
that trough is the entire reason this exists. it isn't an organizer — it's a prosthesis for the one weak system, driving the two strong ones. everything is form, never prose.
nobody's brain can hold a movie.
a film set is a working-memory prosthesis built out of people and paper. call sheets, slates, shot lists, dailies — a hundred years of externalized executive function, invented because nobody's brain can hold a movie.
gollwitzer's implementation intention — “if [trigger], i will [action]” — is the most replicated trick in behavior science. film crews have been printing it every night since 1910. they call it a call sheet.
my brain doesn't need a notes app. it needs what a movie set has.
every entry takes the same five steps, in order.
the order is the point — the one place a number means something. nothing skips ahead.
i checked what it thinks it learned. most of it was invented.
the findings below are computed from 2,821 real prompts across 1,387 sessions. a “move” is one thing i asked for. it counts as finished if an edit or a commit landed before i spoke again. 757 of them did — 26.8%.
your best hours are the ones you don't use.
asking as a question helps. a little.
two findings didn't survive.
“small unfreezes you.” untestable. the only proxy i have is how long before i speak again — and that's long precisely because work was happening. it measures the answer, not the ask.
“questions move. figure out stalls.” across 2,821 prompts i wrote “figure out” once. the finding was invented, and the tool was built to nudge me away from a phrase i never use.
these were never neutral pictures. each one became a feature — propose smaller, nudge toward a question, schedule the hard thing at 10am. two of the three were tuned to fiction.
a project i can't see stops existing.
in inattentive adhd a dormant intention exerts no force. it has to be physically, visually present or it vanishes from the decision. a project that scrolls off-screen is a project that dies quietly.
so every project stays pinned to one wall. the ones i've moved on this week keep their color. the ones i haven't go gray.
which of my projects are actually alive? — not, how do i hold all of them comfortably.
motivation isn't a willpower tank. it's a reward signal, and it's measurably weaker.
the guardrails are the design.
reward-deficit plus anxiety plus perfectionism means any shaming signal backfires and burns more than it recovers. it only ever adds reward and matches difficulty down. it never punishes a low day. a movement break counts as progress. nothing here can be failed.
it's built. i open it most mornings.
local-first, single user, no backend. one rule settles every tradeoff: pick whatever gets the learning loop running on real data sooner. a live, learning, imperfect system beats a complete one that isn't running — time-in-use is the ingredient you can't buy back.
and if you want to test the idea before building anything: run the five steps by hand, on paper, on one real decision today. if it hands you a concrete reversible move, the design is sound. if it doesn't, you just saved yourself the build.