File the handoff's open items as todos, before this session closes

[syncytium2]

A sibling session read #523 and found the defect: the handoff files no todos and adds no INDEX row, so nine open items existed only in a file the tree instructs the next session to delete.

It also made the sequencing argument, which is the better half: the session that wrote the handoff has the context to file them in one pass, and once it closes the next reader has to reconstruct which lines were open items and which were narrative. That's harder than writing them — and it's how an item on the sibling thread ended up with no todo at all.

Right on both counts. This is that pass.

Nine todos, each self-contained

todo	what it is
which-surrogates-enter-the-screen	Tony's call; gates everything downstream
build-the-surrogate-screen	the three tiers, blocked on the above
the-dead-time-floor-is-the-producers-number	τ, and why fitting it is circular
the-design-trains-quiet-and-deploys-busy	the −0.24 vs +0.12 collision
four-recordings-carry-an-unflagged-contaminant	motion-correction frame-floor pinning
nobody-has-read-hansen-1973	a library job; routes already exhausted, listed so they aren't redone
the-canonical-surname-is-gregers-hansen	actionable now, needs no new source
the-export-contract-does-not-mention-field-steps	and the loader drops both columns silently
the-encoder-clips-onsets-onto-the-boundary-frames	inert on real data; fires on every surrogate

Two INDEX rows

One for the self-supervised thread — keyworded on what someone would actually grep (dither, jitter, dead time, leak, no labels) rather than what's in the filenames. One in Known traps for the encoder clip.

The handoff

Now links every item and says so at the top, so deleting it when the thread lands loses nothing. The note names the defect rather than quietly fixing it, because the next handoff author needs the rule, not the tidy result.

🤖 Generated with Claude Code