lab notebook over demo theater

demos are fun. they are also a terrible archive. six weeks later nobody remembers why a tool felt promising or where it fell apart.

what we tried

short notes after each meaningful run: setup, result, decision. not essays — paragraphs. if it takes longer to write than the experiment took to run, we cut it.

what worked

killing projects with a sentence. “paused: latency under load is worse than the alternative we already ship.” that sentence saves the next person a week.

publishing some of those notes. outsiders ask better questions when the work is specific.

what’s next

keep the home page essay-led. the latest note should always be the loudest thing on the site — not a hero animation.