Every founder has the same graveyard: a notes app full of 2:47 AM thoughts that were, at the moment of writing, obviously important. "Onboarding should branch by team size." "Maya said the renewal blocks on actor filtering." "Stop building settings nobody opens."
Six weeks later you ship an onboarding revamp that doesn't branch by team size. The note existed. It just wasn't anywhere that builds things.
Why the best insights arrive at the worst times
Product truth doesn't schedule itself. It shows up in the last two minutes of a sales call, in a support thread you skim at lunch, in the shower, at 2:47 AM. Which means the deciding factor isn't how smart your insights are, it's the capture path: what happens in the ten seconds after the thought occurs.
If the path is "open Notion, find the right doc, file it under the right heading", the thought dies. If the path is a notes app, the thought survives but never votes: it isn't there when you write the spec, brief the agent, or review the PR.
Notes that vote
The fix is making capture end where building starts:
Ten seconds, back to bed. The note is now embedded in your product brain next to every ticket and call transcript. It does nothing, costs nothing, until the day you run scriptonia plan "redesign onboarding", and there it is in the evidence section, cited as [cd537a72], shaping the acceptance criteria of a plan your agent is about to execute.
The compounding part
One rescued note is a nice moment. The habit is a moat: every call transcript, every rant, every 2:47 AM thought makes the next plan sharper, because retrieval works across all of it. Your judgment, which used to evaporate, starts to accumulate. That's the whole idea of agent product memory: nothing your customers teach you should ever be lost between the moment you learn it and the moment something gets built.