Method / New Forms Research
This is not a pile of "I tried AI" writeups. Each note identifies a form that was impossible to finish before, builds it into something you can actually open and play, and then hands over the method.
Finishing is the fastest way to learn.
An unfinished experiment returns no primary data. A prototype nobody plays cannot tell
you where people left or where they cried. So everything here is finished and public.
Each one follows the same shape: constraint → what got removed → the new form → evidence → how to repeat it.
Branching died from word count, not from technology. When the cost of prose collapsed, the ceiling on flags disappeared — and metafiction stopped being a talent and became a feature you implement.
16 projects, 90 videos, and almost no lines drawn between them. 3 of 12 pages were dead ends; 7 of 14 short/long pairs were never joined. Link counts are integers, so make them a test.
Free, no account, runs in a browser.
More notes exist in Japanese — ten of them, covering music mapping, one-day novel pipelines, mining demand from tens of thousands of titles, quality gates for AI output, and designing time spent alone. They are being translated as I go: read the Japanese index.