formal: the deferred sg obligation is carried, and the driver checks the carry - #378
Merged
Merged
Conversation
There was a problem hiding this comment.
Your trial has ended. Reactivate Greptile to resume code reviews.
…the carry `Carry E cp pub cp' i`: the next same-curve proof's old accumulator `i` is `cp`'s opening `sg` with the wire's round challenges of `cp` — the vector `SgOk` commits. `AccOk σ a`: the accumulator equation on an accumulator alone. `sgOk_iff_accOk`: under `Carry`, `cp`'s `SgOk` is `AccOk` of what `cp'` carries — the deferred equation as a check on the next proof's input, without `cp`. Both have decidable mirrors, `accOk` and `carry`, reflected by `accOk_iff` and `carry_iff`. The halves driver gains a `carry` lane: per linked same-curve pair of the proof cache (wrap k−1 → wrap k through the step between them, step k−1 → step k through the wrap), `carry` on the two checked proofs, `accOk` on the accumulator, `sgOk` on the predecessor, and the last two agreeing; the unlinked accumulators (front pads, base-case slots) must satisfy `accOk` on their own. No accumulator in the file is taken from the prover's list on trust any more. Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
martyall
force-pushed
the
pickles/accumulator-carry
branch
from
September 18, 2026 07:26
594000a to
bfa50b0
Compare
martyall
changed the base branch from
pickles/halves-satisfiability-373
to
main
September 19, 2026 00:37
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Stacked on #377.
twoHalves_kimchiVerifycarriesSgOkas a hypothesis: pickles never checks a proof's ownsg, it hands(sg, round challenges)to the next proof on the same curve as an old accumulator, and only the terminator checks the equation outright. The design doc says the obligation is "discharged by the olds mechanism"; nothing in the tree said it, and the halves driver read every proof'soldsfrom the prover-stored list, so a prover who put an unrelated accumulator there would have passed every check we had.Statements (
TwoHalves.lean, besideSgOk)Carrynames what the step and wrap mains force through their message digests (not modelled yet; the handover is between consecutive proofs on the same curve, the circuit in between only computes the challenges).sgOk_iff_accOkis the transport: the deferred equation becomes an equation on the next proof's input, decidable by one MSM without the first proof.accOkandcarryare the decidable mirrors, reflected byaccOk_iffandcarry_iff. The fourtwoHalves_*theorems are unchanged.The driver
check-halvesgains acarrylane (on by default,HALVES=carryalone to run it). Per linked same-curve pair from the proof cache — wrap k−1 → wrap k through the step between them, step k−1 → step k through the wrap — it decidescarryon the two checked proofs,accOkon the accumulator,sgOkon the predecessor, and requires the last two to agree (the lemma, confirmed on data). Unlinked accumulators (front pads, base-case slots) must satisfyaccOkon their own: the dummy'ssgcommits the dummy challenges.On
SimpleChain.json: 14 carry runs (8 pairs, 6 pads), all green; about 50 s per pair (two proofs checked, the transcript and two MSMs), the corpus 37 runs in total. The CI lane picks it up unchanged.Gates run locally: build, dead-code (three new roots), shake on the CI list, runLinter on
Pickles, check-cs, check-halves.🤖 Generated with Claude Code