v0 wrap-up: the stub inventory + a full usage guide#8
Open
WWStoryMode wants to merge 1 commit into
Open
Conversation
docs/v0-stub.md — every stand-in in the system, in one place, so nothing in a screenshot can be mistaken for a finding. The five fixtures; the prompt hash; the triptych's model signature; the ghost-cloud perturbation; the identity rotations. Plus what is NOT fake and survives v1 untouched. Two things fell out of writing it down: - NOT ONE of the five authored phrases returns the fixture it was written for. Type "a body remembers a place it cannot return to" — the project's own signature phrase — and you get a body sliding sideways. That is what "the system does not understand what you type" looks like written down honestly. - sum(ord(c)) for "snapmogen" and "kimodo" are congruent mod 5, so those two triptych panels ALWAYS select the same base fixture. Two thirds of the triptych's visible difference is seeded noise on one identical motion. docs/usage.md — the guide: running it, every view, every control, the full keyboard table (they were undiscoverable), the boot flags, how to read each of the four registers, the fixture-authoring trap, and how to swap the backend. fix(app): T outside performance mode jumped straight to 0.25x. The tempo index was seeded for performance mode's opening state (0.5x) rather than the bench's (1x), so the first press skipped a step. Found while documenting it. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Closing out v0 with the two documents it was missing: an honest account of what is fake, and a guide to using what is real.
docs/v0-stub.md— what v0 fakesEvery stand-in in the system, in one place, so nothing in a screenshot can be mistaken for a finding. The five fixtures, then the four fakes: the prompt hash, the triptych's per-model signature, the ghost-cloud perturbation, the identity rotations. Then what is not fake — the schema, the
Generatorcontract, the renderer, the lineage tree, all four registers — and a table of what retires each fake at v1.The one-sentence version:
Two things fell out of actually writing it down
Not one of the five authored phrases returns the fixture it was written for. I ran the real hash over them. Type "slip away" and the body reaches for something. Type "a body remembers a place it cannot return to" — the project's own signature phrase, the one in the abstract — and you get a body sliding sideways, which is not remotely what that fixture was authored to express. This isn't a bug to fix; it's what "the system does not understand what you type" looks like when you write it down honestly, and it's more pointed than the README's careful phrasing conveyed.
Two of the three triptych panels always show the same fixture.
sum(ord(c))for"snapmogen"and for"kimodo"are congruent mod 5, so for every prompt those two panels select the identical base motion. What separates them on screen is nothing but the seeded jitter I added to avoid a "triptych of twins". So two thirds of that view's visible difference is noise on one identical motion. The in-UI honesty banner turns out to be doing much heavier lifting than I realised when I wrote it.docs/usage.md— the usage guideRunning it; the Lab Bench; the lineage tree; the ghost-cloud; all four registers and how to read each glyph; the triptych; performance mode; re-authoring fixtures (including the translation-only-FK trap that once sank
gather's feet 7 cm through the floor); swapping the backend.The full keyboard table — the shortcuts were entirely undiscoverable, which is what prompted this. Also documents the mouse controls and the four boot flags (
?registers=1,?compare=1,?perform=1).One real bug, found by documenting it
Toutside performance mode jumped straight to 0.25×, skipping 0.5×. The tempo index was seeded for performance mode's opening state (0.5×) rather than the bench's (1×). Fixed; it now steps 1× → 0.5× → 0.25× as the table says.Verified
tsc --noEmit+pnpm buildclean. Every factual claim in both documents was read out of the code or computed from the fixtures, not written from memory — the hash mapping table is the actual hash, run over the actual prompts.🤖 Generated with Claude Code