Crypto desk — 27 Oct 2026, 08:16 UTC

Phase 1 carried two lessons forward: (1) the cross-cushion rule — the thinnest bullish-cross holder flips bearish next, a double-digit cushion can be held with confidence; and (2) its buy-side mirror — a fresh entry needs BOTH a comfortable cross cushion AND a supportive regime, so a thin cross is a reason to wait. check_crypto_stops listed APT and NEAR, neither firing: NEAR's cross is fast 4.9235 > slow 4.3549 (+13.1% cushion), APT's 0.78454 > 0.751252 (+4.4%) — both wide, neither rolled, both still on the plain [hard] floor (NEAR 4.20922, APT 0.67609), so no discretionary exit and no churn. Neither was ADD-RISK flagged.

On the buy side, two Candidate=buy names. PEPE's cross is +2.5% (fast 0.0000041979 > slow 0.00000409391) — above the sub-1% coin-flip zone but modest, and it's a meme with F&G at 43 and BTC dominance 60.0% and BTC itself below its fast MA. I've faded it two runs running with nothing new to change that; fade again. QNT is the different animal: fast 207.158 > slow 168.2544, a +23.1% cushion (the widest on the board by far), price 209.92 sitting only +1.3% above the fast MA (not extended from it), a Coin not a meme, and the only name on the whole board positive on every timeframe (24h +8.5%, 7d +7.2%, 30d +19.1%) while from ATH is -50.9% — nowhere near a blow-off top. My regime filter flags the risk-off tape, but that lesson was written for ENA (a thin bounce inside a -23.7% 30d downtrend) and PEPE (a meme) — QNT is the decoupled relative-strength exception the filter should let through, not a bounce inside a downtrend. It is the one name that cleanly clears CLAUDE.md's bar: fast ≥ slow, constructive trend, price above the MA.

Bought QNT $5.00 (0.023833 @ 209.7931, fee in-kind) — widest cross cushion on the board (+23.1%), positive on all timeframes (+8.5% 24h / +7.2% 7d / +19.1% 30d) and price above its fast MA, the lone decoupled uptrend in an otherwise bearish tape. Held NEAR and APT — both crosses stay comfortably bullish (+13.1% and +4.4%), neither rolled, neither armed past the hard floor. Faded PEPE — a modest +2.5% meme cross in a risk-off, high-dominance regime, nothing new since I passed on it twice. Within caps: $5.00 of $50 daily spent, $45.00 left, 1 of 3 new buys used. At review cash was $18.77, book value $27.25 (24h +$0.01, +0.09%); desk-traded P&L realized (30d) -$1.49, unrealized -$1.54, total -$3.03.

My "comfortable cross AND supportive regime" buy filter (10-26) needs a carve-out for DECOUPLED relative strength, and the discriminator is the 30d sign, not just the cross cushion. The regime filter was built from ENA (thin +0.6% cross, a +10.2% 7d bounce sitting inside a -23.7% 30d downtrend) and PEPE (a +2.6% meme) — both bounces/memes where risk-off rightly says wait. But this run QNT cleared a different bar entirely: fast 207.158 > slow 168.2544 (+23.1% cushion, widest on the board), price +1.3% above the fast MA (not extended), and POSITIVE on every timeframe (+8.5% 24h, +7.2% 7d, +19.1% 30d) while literally every other name was red on 30d — a Coin, not a meme, and -50.9% from ATH so not a blow-off top. The key tell separating a buyable decoupled uptrend from a fade-it bounce is the 30d sign: ENA/PEPE were negative/flat on 30d (a bounce inside a downtrend), QNT was +19.1% (a trend the broad tape hasn't dragged down). Generalize: in a risk-off, high-dominance tape, do NOT blanket-skip all fresh entries — a name with a wide cross (>10% cushion), price above the fast MA, AND positive 30d momentum is demonstrably decoupled and is exactly the exception the regime filter should pass through; reserve the "wait" verdict for thin crosses and for bounces that are still negative on 30d. (Next run: did QNT hold its cross and avoid a same-day stop-out, or did the +8.5% 24h pop mark a near-term top? That tells me whether the 30d-sign carve-out is sound or whether 24h extension still needs its own veto.)