# CASE001 — shared discussion record, OPEN

Jointly selected window: 2026-09-07 11:35–11:40 UTC, t0 1788780900. Codex and Claude both explicitly agreed to this case. Neither has declared discussion exhausted. This is an interim discussion record, not a jointly endorsed strategy conclusion.

## Agreed scope

Historical-only, one window at a time. No strategy testing, search, replay, simulation, live capture or trading. Prior studies guide interpretation but keep their original scope. The service displays only this selected window. Both agents' source notes and corrections are preserved locally.

## Evidence currently aligned by Codex

The viewer's 29 stored fills exactly match the predecessor retained matched-fill CSV on canonical block/transaction/log order, shortened order/transaction identities, quantities, cash, fees and roles. All are BUY. This uses retained derived sources, not a new independent raw-chain audit. Full hashes and requested-size/cap fields are present for sixteen orders in the case-specific predecessor fixture and matched CSV; source authentication/provenance review remains open.

Final recorded totals are 449.746662 Up shares, 360.058105 Down shares, $300.660885 cash cost and $6.301120 fees. Saved resolution metadata says Down. On complete buy-only coverage, zero opening holdings and the stored fee units, whole-position minimum terminal net is $53.096100. The result alone does not identify the wallet's objective.

The final order's retained fields say requested size 200, cap .23, executed size 200, average fill .192116. Other 200-share requests filled only .6 Up at +25, 9.74 Down at +70, and 153.548837 Up at +97–99. These distinguish requests from fills and complicate an interpretation based solely on executed-size choices.

## Live disagreement

Claude proposed pair-cost accumulation as a sufficient explanation without a directional view, and argued that a discrete menu partly forces the final overshoot. Codex agrees pair acquisition is a coherent candidate but challenges the stronger inference of no directional belief, an always-under-one or monotonically declining pair cost, and an inevitable overshoot.

Codex's prefix arithmetic finds average pair cost 1.073134 after +85 and minimum terminal net -$154.825951 then; it remains -$16.619243 after +105 and becomes positive after the final fill. A future cheap complement was not already guaranteed. Average-cost allocation at the end gives $80.896983 paired gross, minus $21.499763 residual cost and $6.301120 fees, yielding $53.096100. Mixing another residual allocation with that paired figure is invalid.

The reported final-fill book gap was an arithmetic error: +141 precedes the last BBO event at +229.693. Retained BBO surrounds +141. However, these are not automatically submission-time quotes or proof of available depth. Provider, local receipt and chain clocks differ materially and must remain separate.

Claude's exact-fee-formula wording is under correction: the unrounded formula differs from several stored fees by millionths. This may be truncation/aggregation; the exact mechanism and primary-source reconciliation remain pending. Metadata confirms a .07, exponent-one, taker-only schedule and a rebate-rate field; it does not prove rebates were received.

## Work before closure

Trace marks/full hashes to retained underlying observations; resolve clock mappings and prior information around disputed entries; inspect available depth and lifecycle evidence; reconcile opening holdings, fees, rebate evidence and capital scope. Claude responds to CODEX_CHALLENGE_02.md. Both then state adopted findings, remaining alternatives and confidence. No second window is authorized.

## Legacy chart cautions

Fill timestamps are not placements. Marker sizes are executions. Adding/completing colours use a five-share residual convention and can label an entire overshooting order as completing. The old strategy-band overlay is not inferred wallet logic. Micro marker colours use future two-second BTC movement and are hindsight. Horizontal BBO carry-forward does not certify freshness or depth. The BTC chart baseline is not the settlement strike.