Stock-market dispatch — 2026-07-20

Headless run (Mon, covers the 07-17→07-20 weekend). Research output, not advice — brain never trades. Marks are the 2026-07-17 (Fri) close (markets closed over the weekend). Supabase sync + push are performed by the wrapper, not this run.

Top of mind

The book's breadth problem got its cleanest fix in weeks: today's three net-new emissions are 100% thin-vertical (healthcare + financials), 0% AI-infra — and two of them are corroborated first-party by the same earnings call. glp1-substitution-to-bariatric-surgery-decline is the standout — a 4-step, all-confirmed chain: GLP-1 efficacy now approaches bariatric-surgery-like magnitudes (retatrutide 28.3%), metabolic surgery volume fell −34% 2022→2024, and the Medicare GLP-1 Bridge Program (live 2026-07-01, $50 copay, ~3.8M eligible) structurally expands the drug population. ISRG's Q2 call confirmed its own US bariatric cases fell high-single-digits "impacted by rising GLP1 usage." The long-only expression is the winner leg — LLY (~60% US obesity share, oral Foundayo) — emitted at 0.60.

The same ISRG call anchors the second healthcare chain, aca-subsidy-cliff-to-deferred-procedure-volume: the ACA enhanced-subsidy expiry (end-2025) more-than-doubled subsidized premiums (+114%), ~2.2M lose coverage, and deferrable elective procedures get deferred — HCA cut 2026 guidance with hard surgical-volume declines, and ISRG's US da Vinci growth fell to 12% (from 14%) on the same "changes in patient coverage and premium dynamics." Not emitted — every leg is a short/avoid (HCA cleanest); the long-only trader can't express it (same precedent as the 07-16 OEM value-transfer).

Third: client-float-interest-to-custodian-exchange-nii — under higher-for-longer, exchanges (CME/ICE, near-zero-beta mandatory collateral) and custodians (STT/BK/NTRS, deposit betas >100% → muted) earn near-costless float NII. Emitted the clean exchange leg (CME 0.46, ICE 0.44).

Strongest-conviction buys

Ranked shortlist (4c). Prices 2026-07-17 close, twelvedata (105/109; ASE/BESI free-plan-gated, GNKG/HXSCL OTC).

⚠ LLY caveat — the chain is right but LLY sits ~1% off its high; the thesis is confirmed, the entry isn't cheap. Emitted undervalued_vs_base: marginal, adoption at-knee / steep / high runway.

Watchlist

New chains to investigate (hypothesis-stage)

• aca-subsidy-extension-outcome-to-insurer-direction — OSCR / CNC / MOH. The swing variable across every healthcare loser chain: a legislative subsidy extension blunts the whole ACA cliff. Two-sided (insurers hurt by enrollment loss, whipsawed up on extension headlines) — a policy-contingent pair, not a clean long. To graduate: a dated legislative vote + a directional read on the pool.
• stripe-block-paypal-bid-to-card-network-disruption — PYPL (long, takeout premium); watch V/MA. $53B Stripe+Block+Advent bid (Reuters 07-15); Chamath flags a plausible higher/competing bid. Not emitted — low-medium, a merger-arb catalyst with steps 3–4 open (card-network disintermediation is years-out). Catalyst watch.

New theses (now active)

• glp1-substitution-to-bariatric-surgery-decline — LLY (winner) / ISRG (loser). All 4 steps confirmed; dated Medicare Bridge catalyst. Emitted, 0.60.
• client-float-interest-to-custodian-exchange-nii — CME/ICE (clean, near-zero-beta) + STT/BK/NTRS (muted, >100% beta). Emitted CME 0.46 / ICE 0.44.
• aca-subsidy-cliff-to-deferred-procedure-volume — HCA/THC/UHS/CYH losers, ISRG corroborating. Medium-high chain, not emitted (short-only; long-only book).
• open-weight-sputnik-to-frontier-lab-derate — Kimi K3 "Sputnik" → frontier-lab premium derate, silicon neutral-bullish (Jevons). Not emitted — low-medium, the short leg is private (OpenAI/Anthropic), and NVDA is already ~18×; one-sided vs the capex-down read.

Updated theses (existing active)

• intuitive-surgical — Q1's hedged ACA "watch" resolved into a quantified Q2 impact (US growth 12% vs 14%, bariatric −HSD). New CEO Dave Rosa; GI-endoscope 510(k) filed.
• open-source-share-shift-bullish-for-compute / ai-roi-reckoning / phantom-data-center-load strengthened (Kimi K3; Ramp 21× token spend → CFO earnings-miss channel; NY statewide datacenter moratorium).
• biogen-tau-failure-to-brain-shuttle-rerate resolved — diranersen Ph2 landed (inverted dose response); reads as ASO-tolerability, favors IV brain-shuttle (ARWR/DNLI) + ALNY intrathecal.

Contradicted / weakened

• JPM figures — the NII autoresearch's snippet-sourced Q2 numbers ($21.2B NII/$6.14 EPS) conflict with first-party jpmorgan ($16.9B/$7.70). Earnings call trusted; snippet not merged. No thesis moved — a source-quality flag, not a falsification.
• No active thesis was falsified today; 0 calibrations filed.

Open questions worth a human's eye

• ⚠ The calibration gate still never fires. Both live flags < −10% sector — PWR −14.6%, TJX −12.2% — land in medium (0.50 / 0.40), and step 4e only reads high/med-high. CEG sits at −9.3%, just inside. Third consecutive mark where every flagged name is medium. The monitor remains aimed where the failures aren't (pre-existing; calibration owed on the gate spec itself, filed 07-17).
• INTC's live exit-pressure eased. The live book now shows INTC at −3.1% / −5.8% sector (was −21% hypothetical on 07-17) — the position is no longer the standout drag. Not re-dated (thesis unchanged); worth a human's eye on whether the 07-17 exit flag still stands.
• 5 rows still below the 0.4 gate (cuda-moat-erosion NVDA 0.32, ASTS/RKLB 0.34, DG/ROST 0.39) — retiring them is a reconcile --prune call this run was told not to make.
• vera-cpu-to-arm-datacenter-royalty page/feed still disagree (page low-medium+hypothesis, feed emits 0.42 active). Pre-existing.

What I looked at

• Run status: headless; steps 0.5–7 in-turn; price fetch paced 8/min — 105/109, one 429 self-corrected (api.twelvedata.com reachable; the 4 failures are the known ASE/BESI plan-gating + GNKG/HXSCL OTC). PAPER-LEDGER not due (weekly; last 07-17, next ~07-24).
• Breadth (2a): ai-infrastructure 53% ⚠ over → net-new steered entirely to thin verticals (healthcare bucket 8, financials bucket 9). Post-emit 50% — 3 new rows, all other-cluster; the denominator moved the right way for once.
• Ingestion veins (0.5–0.7): podcast 5 diarized ($1.26; budget $11.91/$50 July); earnings 1 (ISRG Q2 — deferred from 07-17; flagged partial, condensed host transcript). SCHW/IBKR (07-21) and UNP (07-23) hadn't reported. Feeds a no-op — SemiAnalysis/Fabricated-Knowledge/Apricitas all deduped; only new item was a Construction Physics reading-list roundup (skipped, no chain).
• Steps 1–2: 2 autoresearch (ACA/GLP-1 deferred-procedures; brokerage/bank NII), both thin-vertical. 2b prospect-chains SKIPPED — the 3–5/day hypothesis cap is already spent by manual prospecting (5 mechanisms + 1 question + 8 concepts from today's ingest).
• Sources: 7 promoted (4 → stock-market, 3 multi-context → vault/sources/ + threads/artificial-intelligence), 7 ingested → 5 new mechanisms, 8 new concepts, ~26 company + 8 person entities, 1 new question. Capital Allocators (LP-allocator interview) + 3 stale May/June clippings correctly left off-topic. Zero-novelty: 0.
• Valuation (4b): signal feed re-marked to 07-17 (48/48 instruments priced); 14 concept snapshots re-marked.
• Live feedback (4e): 24 rows, 2 flags < −10% sector (PWR, TJX), both medium → 0 auto-calibrations (see Open questions). Best: DNN +9.9%, WST +6.6%, AMZN +6.8%.
• Signal feed (7): full active/armed set, 51 rows, validated 51/51 against the v3 contract. 48 re-marked to 07-17 closes; as_of carried forward on all 48 (no existing emitted mechanism moved a step today — the financials anchor is the new client-float mechanism + the warsh question update, not a change to an existing row). +3 rows: LLY (0.60), CME (0.46), ICE (0.44). Deliberately not emitted (auditable): aca-subsidy-cliff (short-only, long-only book), open-weight-sputnik (private short leg + one-sided), stripe-paypal (merger-arb, steps open). Supabase sync performed by the wrapper, not this run.