conic.ai

CONVICTION PICKS

AVOID / TRIM

• MU — Spike_only, 3m +124%, RSI 66 after -13% week -- classic FOMO trap.
• MRVL — +56% month, fwd PE 43, today's -7.6% is air leaving not entry.
• ARM — fwd PE 106, spike_only, +53% 12-1; uninvestable multiple even after -21% week.
• HPE — RSI 73, F=4, spike_only, +56% month -- blow-off in payback.
• SMCI — abs_mom False, spike_only, no quality floor -- streak 3 hit 100%.
• PL — F=2, fwd PE -5733 -- uninvestable.
• CRCL — Falling knife, no quality.
• CRWV — abs_mom False, fwd PE -121.
• AXTI — F=2, fwd PE 120.
• APLD — F=4, negative fwd PE.
• ACMR — F=2 quality floor missing despite +158% 12-1.
• WOLF — No F-score, negative fwd PE, pure spec.
• META — abs_mom False, -10.6% 3m -- cheap but no turn.
• MSFT — abs/ts both False -- SaaS lagging chips.
• ESTC — spike_only, abs_mom False.

POSITION PLAN & RISK exposure x1.0 · open P&L +0.1% · realized -5.5% · 1 open / 5 closed

COOLDOWN · no re-entry

• AVGO locked until 2026-06-11 (stop hit)
• NVDA locked until 2026-06-12 (stop hit)
• VRT locked until 2026-06-12 (chandelier trail hit)

BACKTEST · $100k since inception paper book (with stops) vs the plan basket

VALIDATION / EDGE

BELIEF UPDATES · vs yesterday

• UPGRADE AVGO accumulate→buy Post-print purge now -18% in a week, RSI 45, but F=8, fc=0.55, abs/ts pass, fwd PE 20 -- this is the kind of quality drawdown the playbook was built for. Upgrading to buy / high conviction; the weight stays modest (HRP cap) but the stance reflects that the setup improved, it didn't break.
• UPGRADE NTAP avoid→watch RSI cooled from 77 to 73 with a -9% week, hqm_quality True, abs/ts pass, F=6, fc=0.71. Still too extended to buy but quality is real -- moving to watch for a deeper pullback toward RSI 50.
• UPGRADE CSCO avoid→watch F=8, hqm_quality, fc=0.77, abs/ts pass, RSI cooled to 56 from 79 with a -4.9% week. Quality is undeniable; not a buy yet (fwd PE 25, +55% 3m still rich) but the VRT-style pullback we waited for is starting -- move to watch.
• KEEP VRT buy→buy Still the oversold quality AI-power name: -12.6% week / -21% month, RSI 33, abs_mom+tsmom pass, F=5. Streak -4 / hit 20% is painful but every miss has been beta-driven; thesis (datacenter cooling/power capex booked years out -- echoed today by Taiwan grid-eq makers full to 2029) is reinforced, not invalidated. Keep as 1 high-conviction buy; size with HRP not Kelly.
• KEEP TSM buy→buy Top of universe on quality: F=9, fc=0.85, hqm_quality, both momentum gates pass, only -2% on the week vs SMH -6.5%. Holding up best when the group sells off = exactly the relative-strength signal we want. Keep high conviction.
• KEEP NVDA accumulate→accumulate hqm_quality, abs/ts pass, RSI 40, fwd PE 16.4, -5% on the month -- still no chase. SpaceX in-orbit reference design = another structural demand line. Keep accumulate medium; not upgrading until validation gates open.
• KEEP MU avoid→avoid Spike_only still True, 3m +124%, 12-1 +618%, RSI 66 after a -13% week. This is the textbook FOMO trade Han banned -- holding avoid.
• KEEP MRVL avoid→avoid +56% month, -11.5% week, fwd PE 43, spike-y action even though hqm_quality flips True today. Wait for the digestion to actually finish; not chasing a -7.6% reversal day.
• KEEP ARM avoid→avoid Streak 4, fwd PE 106, spike_only -- multiple still uninvestable, -21% week is just air coming out.
• KEEP SMCI avoid→avoid abs_mom False, spike_only, no quality floor. Streak 3 / hit 100% on avoid -- thesis compounding.
• KEEP HPE avoid→avoid RSI 73, F=4, spike_only, +56% month. Late-innings blow-off; the -12.5% week is the start of payback, not an entry.
• KEEP PL avoid→avoid F=2, fwd PE -5733, streak 5 / hit 100%. Uninvestable, compounding.
• KEEP ACMR avoid→avoid F=2, +158% 12-1, spike-y bounce -- quality floor missing.
• KEEP WOLF avoid→avoid No F-score, negative fwd PE, pure spec -- nothing changed.
• KEEP SMTC avoid→avoid Spike_only, fwd PE 42 -- still no fundamentals visibility.
• KEEP CRCL avoid→avoid Falling knife, no quality -- nothing to do.
• KEEP CRWV avoid→avoid abs_mom False, fwd PE -121 -- unchanged.
• KEEP AXTI avoid→avoid F=2, fwd PE 120 -- unchanged.
• KEEP APLD avoid→avoid F=4, negative fwd PE -- unchanged.
• KEEP META avoid→avoid abs_mom False, -10.6% 3m. Cheap but no turn yet.
• KEEP MSFT avoid→avoid abs_mom/tsmom both False; SaaS still lagging chips. Keep on the bench.
• KEEP ESTC avoid→avoid spike_only, abs_mom False -- nothing changed.

VS PRIOR DAYS

Yesterday's calls (VRT/TSM/NVDA/AVGO buy/accumulate; 14 avoids) are all still PENDING the 3-session maturity window, so no actual graded outcome yet. The carryover prior shows buy/accumulate hit-rate at 41% (n=17), high-conviction bucket only 33% (n=3) -- I have NOT earned the right to widen the high-conviction sleeve, so I am NOT adding new high-conviction names today, only upgrading AVGO from accumulate to buy where the dataroom evidence is unambiguously better. Avoids continue to compound (MU/ARM/HPE/SMCI streaks intact). The lesson reinforced today: the right move on big group pullbacks is to upgrade existing quality names that improved, not to start new positions.

MACRO & REGIME

X / FINTWIT TRENDS

• AI in orbit: SpaceX AI1 launches next year using Nvidia Rubin/GB300, also TPU-ready -- another non-hyperscaler demand vector for Nvidia silicon (dnystedt).
• Taiwan grid-equipment (Fortune Electric, Shihlin, CHEM, Allis) full to 2029 on AI-datacenter + chip-fab power orders -- structural confirmation of the VRT/CEG/power-grid thesis (dnystedt).
• Apple expanding Private Cloud Compute to GCP on Nvidia -- broadens NVDA's enterprise/regulated-AI footprint (nvidia / dnystedt).
• DeepSeek V4 1.6T -- 100x inference perf improvement in 26 days on GB300/MI355X/B200; reinforces the cycle of model launches re-creating compute demand (SemiAnalysis/dylan522p).

FULL WATCHLIST · reference only, descriptive not a buy list