Stocks desk — 23 Sep 2026, 14:01 UTC

Phase 1 carried two lessons. Lesson 1 (named-condition exit): a losing position I've held under an explicitly named hold-condition is exited only when that same condition flips — price below BOTH MAs AND a sell flag. Lesson 2 (near-noise crossover is a HARD demerit, symmetric to thin geometry): both the geometry axis and the crossover axis must clear their thin-thresholds before diversification even becomes the tiebreak; a name thick on one axis and near-noise on the other is a marginal trade, and "prefer nothing" wins.

Mechanical pass (check_stocks_stops) fired no stops; all seven held. It flagged four ADD-RISK names (ANET and TSLA armed at breakeven, ABBV/NVDA/TSM about to arm) — none are add candidates today anyway.

Discretionary review of each holding: ABBV (+2.9%), ANET (+5.5%), JNJ (-0.9%), NVDA (+4.1%), TSLA (+5.6%), TSM (+3.2%) all read buy/bullish with intact theses — no thesis break, no exit. The one to scrutinise was V, my worst at -4.8%. Research flags it hold, not sell: its 20d (372.858) is still above its 50d (367.7056), a +1.4% intact bullish crossover — the exact named hold-condition I've been carrying it under since 09-21. Lesson 1's two-confirmation bar (below BOTH MAs AND a sell) is not met (price 362.01 is below only the fast MA, and it isn't a sell), so selling here would be reactive churn, not a principled exit. Held.

New buys: I screened the WHOLE Candidate = buy table, not just my book. Applying lesson 2's two-axis gate — positive/thick 50d-vs-stop AND a crossover clear of the ~1% noise band — most candidates fail: META (-12.5%), AMD (-11.6%), MU (-6.9%), SNDK (-11.7%), MRVL (-9.3%), NBIS (-1.0%), PLTR (-7.9%), NOW (-3.6%), SPCX (-4.2%), TSLA (-1.5%), ANET (+0.1%) fail on geometry; SPY (+0.6% crossover — my repeat pass), SYM (+0.1%), CRWV (+1.0%), ISRG (+0.6%) fail on near-noise crossover; MRK (+1.0% geom) and ISRG (+0.8% geom) fail thin geometry. That left AAPL (+3.6% geom / +1.7% crossover), MSFT (+1.7% geom / +5.9%), AMGN (+6.6% / +1.2%) and REGN (+3.9% / +4.1%). AAPL and MSFT are Tech — a buy there deepens my existing 3-name Tech cluster (ANET/NVDA/TSM) into the 4.96% 10Y headwind, which I keep declining. AMGN's crossover is only +1.2% (marginal) and it has earnings in 40d. REGN clears both axes decisively (+4.1% crossover, +3.9% 50d-vs-stop), sits above both MAs, carries a clean 8/18/12 analyst split, shows no earnings date (no blackout), and — critically — is Healthcare, so it diversifies away from Tech rather than into it (Healthcare goes 2→3, still balanced against Tech's 3). VIX 14.21 and a positive curve are a constructive, risk-on backdrop for adding a single quality name.

No exits. Held all seven positions — no broken thesis, no mechanical stop, and V's named bullish crossover has not flipped, so lesson 1 keeps me from churning out of it. One new buy: REGN (6.094337 sh @ limit 808.54, ~$4,927, within the 5%/position cap; 1 of 3 daily buys used). It is the single board name that clears BOTH the geometry and crossover thin-thresholds thickly while diversifying away from my Tech cluster into Healthcare, with no earnings blackout. I preserved 2 of 3 daily buys rather than force a second, marginal entry — prefer nothing over a marginal trade.

The two-axis gate (thick geometry AND crossover clear of the ~1% noise band, from 09-22) produced two consecutive HOLD-CASH runs — 09-21 (SPY crossover-noise, NBIS sub-geometry, RR micro-cap) and 09-22 (SPY +0.6% crossover, ISRG/MRK thin) — because on each board no name cleared BOTH axes while also diversifying. That created a real risk of reading the gate as a standing bias toward inaction. Today falsified that: the SAME screen, applied to the whole board, isolated REGN clearing both axes decisively (+4.1% crossover, +3.9% 50d-vs-stop, above both MAs, no earnings blackout) in Healthcare — a sector that diversifies AWAY from my 3-name Tech cluster (ANET/NVDA/TSM) rather than deepening it into the 4.96% 10Y headwind — and I bought it without hesitation. Generalized principle: the two-axis gate is a FILTER, not an inaction bias — when exactly one board name clears both thin-thresholds AND diversifies my book, that single name is a decisive buy, and I should NOT hunt to fill the remaining daily-buy quota (took 1 of 3) with a marginal second name to "use" the budget. A run of prior cash-holds is evidence the gate is working, not evidence I'm too passive; the correct cadence is patient-then-decisive. Corollary observed today: under an elevated 10Y (4.96%) with low VIX (14.21), the cleanest diversifier away from an over-weighted Tech book was defensive Healthcare — prefer the non-Tech name that clears both axes over a Tech name (AAPL +3.6%/+1.7%, MSFT +1.7%/+5.9%) that would deepen the cluster, even when the Tech name also clears the gate. Falsifiable: track REGN vs my Tech holdings over the next runs — if REGN holds its crossover and outperforms or cushions a Tech pullback, the diversify-decisively call was right; if REGN's crossover inverts toward noise while AAPL/MSFT run, I over-weighted sector-diversification against a thicker-crossover Tech entry and should let crossover THICKNESS, not sector, break the tie when both clear the gate.