Weekly money-loop pass — 2026-07-19

First pass under the standing questions (ratified 2026-07-19). Mechanical block green end-to-end; LADDER.json stamped. Data through the 2026-07-17 close.

Scoreboard + loop state

• Decision loop: 584 rows scored (848 complete windows). Conversion (7d): 22 rungs armed, 1 fired, 0 converted, 0 declined. Counterfactual: no blocked gate intents. RL readout appended (first meaningful d30 ≈ 2026-08-08).
• Omission loop: 31 cautious tags priced — passing SAVED +6.8% avg; 0 passes ran >+25%; 9 correct avoids. Deep-dive stream remains the best passer (+13.2% edge on 16 passes). Best avoids: CIEN −34.3%, INFQ −33.3%, VPG −30.7%. Worst miss: RGEN +16.4%.
• Ladder: all books hold their rungs. Druckenmiller-mirror (+7.96 med30) and Buffett-mirror (+15.84 med30) lead; the four core paper books sit slightly negative on median with the early-exit drag documented below.

Attributions (16 required rows cleared)

The queue's story is one sentence: the books' exit rules sold the AI melt-up. execution ×7 — ARM −68.9%/−35.3%, CRWD −55.3%, SNOW −28.2% across all four books (same rule, four books — the rule is the variable), plus the autotrade too-young-regime entry (−25.2%). All filed as evidence for the live trailing-stop forks (readout 2026-10-08). within-plan ×8 (13F-mirror trims BE/ON/GOOGL/LSCC — mirror-strategy risk, not error; yolo mandate variance; BRBR's stop that sold a bear-rally correctly per the 07-19 EDGAR re-affirm). research ×1: ABT — the death-cross "bounce zone" thesis was wrong; low RSI is not cheap (the June 25 doctrine, learned again).

Flagged-call verdicts (all four resolved)

• META — in-zone (07-10 at 669.21) VERIFIED then failed; close 646.01 back under the 665 trigger. No entry taken; call open, second attempt needs a fresh close above 665. Jul 29 print next.
• NTLA — invalidation CONFIRMED (12.95 ≤ 14.20 on 07-13; now 11.38). The call is dead and renders broken; re-entry requires a new base, not this call.
• CRDO — the 07-09 reclaim failed hard (202.68, −21% under trigger) in the AI-semi derate. No entry; invalidation (~160) not threatened; call open.
• MXL — the 07-09 in-zone failed (71.84 vs 91). Earnings Jul 22 decide it; under the 1-yr VWAP (~50.66) the call dies.

Omission-board verdicts (5 prompts, all examined)

• WDC intact-broke — stands by design: gated on the Jul 29 print (tag carries the breach note).
• ELAN / ILMN / NTRA / RVMD caution-overridden — all four examined: the near-high strong-up tape is real, but every push is on sub-0.6× relative volume, failing the volume bar the desk's own expression-ladder doctrine (written this afternoon) requires. Tags stand as intact-but-priced with as_of bumped and 52wk-high reclaim legs armed (~$27.72 / $196.66 / $288.04 / $193.82, auto-derived). ELAN gets the desk call (frontmatter): the flip trigger is a volume-confirmed high reclaim or the ~early-Aug Q2 print — the wait-for-pullback-becomes-missed-breakout risk is now a watched trigger instead of a silent drift.

Origination queue (18 rows, all given dated reads)

Watch (7): BE (margin inflection +9.7% vs −1.9% prior-median is real; −34.6% 30D washout — interest on the first higher low + 20-day reclaim), HWM (quality in-trend; add only on a flag toward the rising 20-day, never the vertical), ATI + CRS (the metallurgy pair, both in textbook flags −9/−11% off highs; 20-day reclaim on volume = candidate entry), SCCO (margin elite but copper leg still rolling — no entry while COPX deteriorates; FCX print Jul 22/23 is lane info), LSCC (margin inflecting +15.3%, weak-down; base + reclaim condition), AEHR (print-gated — FY26-Q4 ~late July decides). Pass, dated (11): AWK (defensive utility, no thesis edge), OUST (existing intact-but-priced tag covers it; spec end of the lane in derate), UMAC (spec, margin −89.7%), IBRX (margin −157.9%), AVTR (tools-tier shape but +15.6% 30D mid-move; re-look at the ladder's conditions on the next refresh), PTGX (RSI 69 at the high — step-0 chase shape), BLK/COF (quality, not our lane), AS/MDLZ (no thesis), PYPL (RSI 80 parabolic bounce — chase).

Standing question 1 — bear-case attack: MEMORY (rotation 1 of 4)

Where a short-seller attacks today: (1) the pricing-power story is contractually bounded — Micron's own floor-and-ceiling structure caps ASP torque on ~40% of revenue exactly where the market extrapolates it (the July 13 ASP-repricing read); (2) breadth inside the thesis already broke — STX/WDC/EWY June setups all invalidated; the complex narrowed to MU/SNDK survivors, and Jul 28–29 prints are a dated catalyst with asymmetric downside if HDD/NAND pricing wobbles; (3) distribution tells — MU insiders 1 buy / 54 sells; the cohort traded 120–152% above lifetime VWAP at the June froth scan; (4) demand quality — the ramp is contract-pricing-driven with shipments flat/down in some BUs; if the next MU guide disappoints on HBM4 pricing, "demand-anchored" loses its anchor. Did this week feed the bear? Yes, two real bricks: the KLIC filing read showed the advanced-packaging mix SHRINKING at the equipment layer (the AP pull-through story is unproven where the bull cites it), and the COHR/LITE read downgraded the "committed dollars" floor to unverified narrative. Neither falsifies demand (MU diverged from SNDK and fell least; the EDGAR ramp + $100B contracted stand) — but this is an earned "the bear got stronger this week," not a nothing-new.

Standing question 2 — concentration read (baseline)

Shared factor: AI-derivative demand. Three of four PRIMARY slots (memory, electricity, physical-AI BoM) monetize the same AI-capex impulse, and the open-call book leans the same way — of 50 open desk calls, the heavy majority are AI-lane (CEG/GEV/LEU/CRDO/MXL/META/the memory complex). Rough read: ~3/4 of PRIMARY conviction and well over half of open-call exposure ride one macro bet. The single print that hits all of it at once: an NVDA guide miss — or hyperscaler capex language softening in unison, for which GOOGL's Jul 22 print is this week's first read. This is the baseline pass; next week reports the delta. The non-AI ballast on the board today: ELAN, the biotech ladder names, BG/food-security — thin by comparison.

Sources

• research/trading/decisions/LADDER.json (this pass's stamp) + LEDGER.json via decision CLI receipts
• research/market-engine/config/thesis-tags.json (board verdicts, 2026-07-19)
• research/investigations/2026-07-19-thesis-tag-omission-loop.md · 2026-07-19-stopped-out-reexams-and-hubs-verdict.md · 2026-07-19-optical-lane-filing-reads-klic-axti-cohr-lite.md
• research/investigations/2026-07-13-memory-selloff-is-an-asp-repricing-not-sympathy.md · 2026-07-18-sell-side-regime-read.md
• research/trading/decisions/ORIGINATION-QUEUE.json (18 rows, tape re-read at 2026-07-17 close)