AVOID / WATCHLIST. P(beat QQQ) = 0.34; score 42/100. OKLO is the better 6-12 month advanced-nuclear expression; wait through XE's September 1 lockup test.

AVOID NEW MONEY / UNDERWEIGHT. P(beat QQQ) = 0.36; score 42/100. Public proxies hedge nuclear beta, not STDN-specific unlock supply.

NEUTRAL / HOLD. Prefer long $1 INTC versus short $0.92 AMD; the Sep $100/$130 call spread is acceptable only at no more than $8.25.

AVOID / WAIT FOR PROOF at $5.84. No trade until filed financing, 45X monetization, or named non-Trina offtake resolves the funding and customer-quality gaps.

NO BLANKET NAKED CALL SALES. Test only remote ADBE 350/400 and CRM 250/300 December call spreads; defer TTD until Q2. Use the premium for QQQ and GOOGL upside protection plus IGV downside convexity.

STARTER BUY WITH CONFIRMATION ADDS. Curiosity-loop P(beat QQQ) = 0.44, independent forecast = 0.52, expected absolute return = 10.1%, score 64/100.

His 2007 playbook, not a top call: +32% H1 being long the AI epicenter while holding ~40% cash. His bearish public calls miss 8/10 times; his cash-while-long positioning preceded his greatest trade. The stark divergence: Tepper 40% cash vs BofA FMS 3.9%.

AVOID as core; rent the catalysts. P(beats QQQ)=0.22 (historical rolling-year win rate vs QQQ: 9.9%); 2x-to-$42 0.40 vs halve 0.75, the most negative skew in coverage. The one new fact: 2026's rally was bought intraday (+287%) against a lifetime overnight-only pattern.

AVOID as core; squeeze-and-catalyst trading vehicle only. P(beats QQQ)=0.18; P($18 consensus touched)=0.20; 2x 0.35 vs halve 0.68. All gains arrived overnight (+9,015% lifetime) while sessions lost 98.9%; every covering firm was a paid placement agent; targets were marked at the top days after the $1B raise.

Thesis real, instruments scarce: best structures are long NB into the EXIM close, the BE-puts-vs-NB/SRL pair, and SCY.V as the un-rerated cousin. No futures/ETP exists; REMX is not a proxy. Nov 2026 truce checkpoint is the dated catalyst.

HOLD / add on confirmed catalysts. P(beats QQQ)=0.37; 2x-by-Jul-2028 0.18; 3x-by-Jul-2029 0.07; halve 0.25. Institutional tape (intraday accumulation, negative momo coupling), Q2 print 23 Jul is the near binary.

The slogan is wrong; the instrument is right. 33.1% (not most) of M2 is post-2020; equities/M2 hit a record 4.93x in Q4 2025 vs 4.31x in 1999; credit/GDP has been flat since 2019 (belief-leverage leg fails); money-as-energy breaks on conservation (banks create, repayment destroys) but survives repair as 'money flows to credible printers because it is not conserved.' Calibrated: P(real GDP >=8%

AVOID as core vs QQQ; policy-catalyst position at most. Axion P(beats QQQ)=0.37, P(2x by Jul 2028)=0.30 vs P(halve by Jul 2027)=0.23. The 'trades with semis' effect is beta (partial corr 0.10); the momentum-crowd coupling is real (partial 0.47–0.52, shift p=0.02) and sector-wide. Decorrelates only on MP-sized policy shocks; the dated one is the 10 Nov 2026 MOFCOM suspension expiry.

AVOID-leaning, negatively skewed. Axion P(beat QQQ) = 0.22, curiosity-loop 0.41 (15 Jun run). 25 Jun refresh: faded ~35% from a $19.87 ATH close (2 Jun) to ~$13 on sell-the-news; the convex read is P(double to ~$26 in 24mo) = 0.34 vs P(halve to ~$6.50 in 12mo) = 0.62. Can 2x on scarcity + government catalysts, but the skew is down.

HOLD / starter vs QQQ. Axion P(beats QQQ)=0.38, curiosity-loop 0.48 (small buy on convexity). The Apple-modem cliff and handset/licensing erosion land before the back-loaded data-center pivot pays; the catalyst and the cliff are on different clocks.

A call option, not an easy 2x. P(double,12mo) 0.23 vs P(halve,12mo) 0.48; 5x 0.07, 10x 0.025. ~205M shares unlock ~10 Nov into an ~80M float (Arnold/Centaurus among them). Cape first power 0.52, 500MW Google PPA 0.27, beats-QQQ 0.34. Starter size only; size for the lockup.

Convex bet, not an index substitute. ~0.32 to double in 24mo vs ~0.22 to halve in 12mo; Specs-100k 0.20, re-rate-to-2.5x-sales 0.25. P(beat QQQ)=0.30 is one lens. Own small for the asymmetry.

HOLD / fair value. P(beat QQQ) = 0.47. The pivot is real but low-trigger: P(a compute business by 2027) = 0.16.