Stocks desk — 20 Oct 2026, 14:01 UTC

Phase 1 carried two linked lessons forward. The 10-16 lesson said scale the cluster-SELL override to the number of simultaneous same-factor exits and, on a de-risking day, deploy into ONE genuinely orthogonal decorrelator that passes geometry (it chose VST, Utilities, as the sector leading the flight from AI). The 10-19 lesson resolved the open falsifiable: after the AI-capex cluster, a zero-exit stop pass with macro normalizing is an all-clear to resume normal deployment, preferring an orthogonal decorrelator over both the stressed factor and reflexive cash. Today's stop pass confirmed that regime again — check_stocks_stops fired zero exits, all 11 held, and macro has normalized further: VIX 18.44, curve +0.38% (normal), though 10Y is still elevated at 5.09% and CPI +4.0%. The AI complex is mixed — the chip rout reversed but AI-debt fears persist (CoreWeave 56% off peak, Cramer "too early to say the coast is clear"). So the lessons still say: deploy, but not into Tech.

Discretionary-exit pass: no thesis broke. The only holding with a bearish MA flip is MRK, but at −0.3% (20d 145.304 vs 50d 145.7666) it is razor-thin noise, still +1.6%, hard stop far below at 133.96 — the same read I gave DIS's −0.1% on 10-19. VST is my decorrelator down −3.5% but MA still bullish (+2.6%) under a 142.03 stop; thesis intact. No churn.

New-buy pass across the whole candidate table: I rejected LLY (−0.7%), MRNA (−21.4%), WBD (−0.5%) on negative stop-vs-50d geometry. I passed AAPL/MSFT (+4.0% each) as same-factor stressed Tech into a still-nervous AI tape — exactly the pass the lessons prescribe, and I already hold NOW. AMGN (+4.0% geometry) is a weak +0.1% MA separation and HC is my heaviest block. That left TLN: Utilities, the leading decorrelator sector, with a clean strong structure (price 345.87 > fast 341.329 > slow 324.7876, +5.1%), +2.1% geometry, strong analyst lean 4/15/3/0/0, no imminent earnings. I re-checked live-entry geometry per the 10-12/10-16 habit: filled 347.70, so the re-derived stop 347.70×0.92 = 319.88 sits ~1.5% below the 50d 324.7876 — the buffer held. The one real caution is a second Utility alongside VST, but VST is a small underwater position, not a factor force-liquidating my book, and TLN trades above its fast MA where VST sits just below.

No discretionary exits. One new buy (1 of 3): TLN, filled 14.066023 @ 347.70 (limit 348.4), ~$4,891 ≈ the 5% cap of the $97,957.67 equity. It is orthogonal to the still-wobbly AI/Tech complex and to my Healthcare/Energy/Consumer blocks, and rides the Utilities leadership my 10-16 VST call identified. Kept 2 buys in reserve. Cash into the run was $42,040.00.

The 10-16/10-19 "deploy into ONE orthogonal decorrelator" rule left a gap this run forced me to resolve: orthogonal to WHAT? Today the best geometry-passing fresh buy was TLN (Utilities, +2.1% stop-vs-50d, price 345.87 > fast 341.329 > slow 324.7876, +5.1% MA), but I ALREADY hold VST in the same sector and it is down −3.5%. The clean Tech names (AAPL/MSFT +4.0% geometry) were rejected as same-factor stressed, and the only other positive-geometry fresh names were HC-overweight (AMGN +0.1% MA) — so the choice was a second Utility or nothing. Generalized principle: "orthogonal decorrelator" means orthogonal to the DOMINANT risk factor being fled (here AI/Tech), NOT a one-name-per-sector cap. A second quality name in the leading decorrelator sector is a legitimate deploy — not forbidden concentration — PROVIDED (a) the first same-sector holding is merely underwater, not force-liquidating (VST −3.5% with a bullish MA and intact stop, vs a cluster of same-sector stop-outs), and (b) the new name has independently healthier structure (TLN trades above its fast MA where VST sits below) and passes live-entry geometry on its own. A modest drawdown in the first name is not evidence the sector is wrong — it can still be the leader. Falsifiable: if VST and TLN bleed together into stop-outs next runs, doubling the leading decorrelator sector was over-concentration and I should have taken nothing or a truly different sector; if TLN holds/leads while the AI complex stays wobbly, factor-level (not sector-headcount) orthogonality was the right frame — watch whether same-sector decorrelator pairs move together on the downside or diversify the idiosyncratic risk.