Internal working note. The loop-4 accountability pass, folded into Monday's EOD session per the flagship contract. Miss-screen basis: settled 2026-08-17 summaries, ≥+30%/30d unclaimed movers with $10M+ dollar volume — 77 movers, 50 unclaimed. Ledger basis: today's call-ledger refresh (forward marks across d30–d730 windows) + gen-nulls (7 null benchmarks filled), and the two conviction-rung decisions recorded this session.

The misses, triaged

• CBRS +46%/30d, $3.75B dollar volume — the painful one, and it is a PROCESS miss, not a coverage miss. In our data the whole way (neoclouds watchlist, a July 9 candidate note, a conviction rung with a numeric trigger). The rung FIRED in-zone on 2026-08-04 at the $221.30 trigger — and the flag sat unhandled for two weeks while the name ran another ~14%. The machinery worked; the human loop never picked it up. Resolved today the honest way: a documented pass ("converting at $251.98 after the run is a chase") with a re-arm condition. Fix filed: the conviction dry-run is now read in every daily (this session added it to the dossier step) — an unhandled CONVERT older than a session is a discipline smell on par with an unverified flag.
• The software re-rate cluster — TEAM +70% (RSI 75), U +57% (RSI 80), WDAY +32%, MDB +40%, SAP +31%, ESTC +35%, PATH +32% — a LANE gap. Seven unclaimed movers from the same neighborhood (application software / SaaS) in one screen is not seven misses, it is one: no active perspective owns the software re-rate, and Saturday's brief already had TEAM on its "unexplained four" list with its filings unread. Which thesis should own it: a cloud-saas/application-software re-rate lane (the old ai-scan bucket watches infrastructure, not apps). T−1 trigger that would have caught it: the momentum board's overbought-list growth (41 → 55 last week) was the tell that the re-rate was broadening beyond infrastructure.
• AAOI +51%/30d, $2.0B dvol — conviction miss inside an OWNED lane. The optical-supercycle bench claimed LITE, CRDO, and AXTI (all benched, all paid) but AAOI ran the same lane unclaimed from the same watchlist. The bench caught three of the lane's four runners — good — but the fourth was in our data with no question asked.
• SMCI +58%/30d — coverage-adjacent. On ai-scan/war-sensitivity screens with insider-buy receipts, but no thesis owns it and the desk has passed on it repeatedly at lower levels; if that is a standing judgment it should be a recorded pass with a reason, not silence.
• AEHR +80%, AXTI +109% (benched), P +69% (RSI 82) — the froth tail. Monster-discovery names arriving the fast way; the washout/stretch screens have them. No action — the froth tail is where the chase deaths live.
• GDXJ +31%, AEM +38% — the gold-miner echo. Gold's death-cross gap narrowed to −2.6% in Saturday's read and the miners front-ran the flip. If gold's cross completes, the miners' turn was the T−1 tell; not our lane today, and saying so out loud is the point of this line.

Coverage miss vs conviction miss

Coverage misses this week: effectively zero — every top-20 unclaimed mover was already inside a watchlist or screener bucket. The misses are ALL conviction-or-process: an unhandled in-zone flag (CBRS), an unowned lane (software re-rate), an unclaimed fourth runner in an owned lane (AAOI). The pipeline keeps proving it can SEE; the losses come between the flag and the decision.

Calls graded, both directions

• What worked: last Monday's flagship put ABCL on the alarm board at 4/5 bells; it printed +73%/30d into this week — the board saw it early. The drone-trade desk calls (AVAV/UMAC watch, never-chase-the-parabola) aged well in one session: UMAC −11.48% Monday while both watch conditions kept the desk out of the top.
• What did not: the same ABCL row's conviction rung fired in-zone 2026-08-10 and sat unhandled a week — same process failure as CBRS, same resolution today (documented pass at RSI 81.4; re-arm on a base). Two unhandled CONVERTs in one week is a pattern; the fix above covers both.
• Null benchmarks: 7 filled today at the 30d horizon (call-ledger gen-nulls) — the sweep classes grade against random/momentum nulls from here.

Verdict + reasoning

Open, carried to next Monday: (1) does the conviction dry-run stay clean now that it is read daily; (2) does a software-re-rate lane get stood up or explicitly declined; (3) does AAOI earn a bench seat in optical or a recorded pass. The screen's own scoreboard stays the referee.

Sources

• research/market-engine/data/summaries/*.json (settled 2026-08-17)
• miss-screen / stretch-scan / trigger-check runs, this session (runlog 2026-08-17)
• research/trading/decisions/ — dec-2026-08-18-cbrs-claude:conviction-001, dec-2026-08-18-abcl-claude:conviction-001
• research/issues/2026-08-15-market-brief.md (Saturday's unexplained-four and alarm-board reads)