Recommendations
2 to acquire · 3 to dispose
| Ticker | Name | Close | Action | Action Price Where to act — a buy-limit on a pullback or a buy-stop above resistance, not the current price. Sanity-check vs Close: on a held name, an action price stranded above the close means the queued level was set when the stock was higher — stale, awaiting reprice. | Stop The risk-defined exit, set below the entry for a long (~1.2–1.8×ATR). A held long whose Stop sits at/above the Close is already below its stop — a stale box awaiting reprice. | 1wk Model estimate of the likely price at this horizon — a forward projection, not a guarantee, and separate from the trade's Action Price or Stop. Where the ML estimator is enabled these are per-ticker calibrated forecasts (~80% coverage band) that are refreshed periodically and can drift between refreshes; otherwise they're interpolated from the signal's swing target. Confirm against the settled close. | 4wk Model estimate of the likely price at this horizon — a forward projection, not a guarantee, and separate from the trade's Action Price or Stop. Where the ML estimator is enabled these are per-ticker calibrated forecasts (~80% coverage band) that are refreshed periodically and can drift between refreshes; otherwise they're interpolated from the signal's swing target. Confirm against the settled close. | 12wk Model estimate of the likely price at this horizon — a forward projection, not a guarantee, and separate from the trade's Action Price or Stop. Where the ML estimator is enabled these are per-ticker calibrated forecasts (~80% coverage band) that are refreshed periodically and can drift between refreshes; otherwise they're interpolated from the signal's swing target. Confirm against the settled close. | Rationale |
|---|---|---|---|---|---|---|---|---|---|
| KLAC | KLA Corporation | 209.37 | SELL 30% | 209.37 | 193.38 | 210.19 (+0.4%) | 217.76 (+4.0%) | 216.76 (+3.5%) | Banking a third of a position whose risk line has walked 30% away from it without a single share ever being sold. KLA's business is not what is being cut - the shares are up 118% over 52 weeks, a quarterly dividend is payable 1-Sep, and stay long the thesis - but a 1000-share holding whose stop has fallen from 272.00 to 191.24 with zero size reduction is a risk failure independent of price, and Seeking Alpha's current framing of the name as 'a great business at a dangerous price' is the case in one line. OPERATIONAL: ADR-0020 Trail Erosion FIRST FIRE, session 12 - Peak Stop 272.00 (2026-07-01 post-split) vs live 191.24 = -29.69% (-28.90% after tonight's raise to 193.38), never size-cut since the 07-04 opening balance; US-36 makes this first fire IMMUNE to the held-name sell veto stack (US-17/US-18/US-34/US-22/US-30) and fixes the tier at the 30% floor. Re-price of an unexecuted card 208.25 -> 209.37, NOT a US-19 escalation. The signal leg is deliberately absent and is NOT the basis (AVOID cv57, FailedBreakout_Down_20D, 2 prints). ADR-0012 TESTED WITH A CLEAN ARM FOR THE FIRST TIME AND STILL DECLINED: armed on the Gain_ATR limb (3.15 >= 3.0) with Unreal_PnL +19.13%, so US-26 passes outright and US-33 does not bind - but ZERO of four ticks fire: (a) no 2nd Exit_Warning code beyond PROFIT_ARMED, (b) conviction exactly unchanged, (c) Target_12Week 216.76 = 1.0353x Close, alive, (d) AVOID not NO TRADE. Valuation amplify (MoS_FV -5.68 <= -3, would bump SUPPRESSED by G2: Target_12Week 216.76 vs DCF_per_share 83.43 = 2.60x overshoot. Provenance: action price = anchor settled close 209.37; the row's printed Stop_Price 230.69 sits ABOVE the close - the S1 short-stop-on-a-long defect (US-29), not used. Trail on remainder 191.24 -> 193.38 (1.5 ATR). Keep; proceeds; realized ~+ (+19.13%). Est 12wk 216.76 exceeds the 83.43 DCF anchor (2.60x) - projection assumes growth beyond priced-in. Dividend record date 17-Aug: a Friday-open fill forfeits on, immaterial and not a reason to delay a 12-session-overdue trim. Adj conviction recorded; does NOT gate a risk-discipline card. · news |
| NBIS | Nebius Group, N.V. | 255.04 | SELL 30% | 255.04 | 215.37 | 254.42 (-0.2%) | 269.68 (+5.7%) | 286.29 (+12.3%) | Nebius delivered a genuine blowout - Q2 revenue up 454% year-on-year against a consensus, AI-cloud revenue up sixfold to, adjusted EBITDA swinging to + from a - loss, the 2026 outlook reaffirmed and the year-end contracted-power target raised to 5GW - and 210 of stay long every word of it. The coming off are not a verdict on the business; they are the overdue correction of a risk failure. This position has never been reduced while its stop sat 20% below the peak we once protected. Michael Burry disclosed on 12-Aug that he added to his short in the name, and the sell-side is openly split between 'could get cut in half' and '23% undervalued' - that argues for carrying less of it, not none. OPERATIONAL: ADR-0020 first fire, veto-immune (US-36), session 2 - pre-committed UNCONDITIONALLY on 08-12 to the 08-13 anchor and carded 08-13; it did not execute, so it carries, re-priced 259.20 -> 255.04. Erosion -20.23% vs Peak Stop 270.00 (2026-06-23), zero size reduction ever on. ADR-0012 is UNAVAILABLE not merely un-fired: Gain_ATR 1.49 < 3.0, T12_Progress 89.08 < 90, no Exit_Warning code at all - not armed, so the profit-take path cannot be and is not claimed as the basis. The signal leg is BULLISH (HOLD cv75 on AnticipationUp_VCP, up from NO TRADE cv0 two sessions ago) and that is not a defence - ADR-0020 is the non-signal origination class. Valuation amplify (MoS_FV -6.03 <= -3, would bump SUPPRESSED by G2: DCF_per_share is NA and Rev_Growth +855% is the hyper-growth/unreliable-anchor case G2 exists for. Provenance: action price = anchor settled close 255.04; the row's Entry_Price 262.35 is a BUY-stop pivot and Stop_Price 175.08 a loose long stop - neither used for a sell. Trail: 1.5 ATR = 211.40, BELOW the standing line -> stop 215.37 STANDS on the remainder; model 175.08 rejected (US-29). Keep; proceeds; realized ~+ (+20.53%). Est 12wk band [148.04-681.73] spans -42% to +167% - that +-4.7x width is the honest statement of model uncertainty here and is itself an argument for carrying less; DCF annotation impossible (DCF NA). PRE-COMMITMENT (1) CARRIES UNCHANGED: any settled close below 215.37, or a 2nd consecutive settled STRONG SELL - REDUCE, fires a mandatory 50% trim (105 of the remaining 210). · news |
| GOOGL | Alphabet, Inc.- Class A | 346.36 | SELL 30% | 346.36 | 332.13 | 346.97 (+0.2%) | 355.41 (+2.6%) | 386.87 (+11.7%) | Alphabet has now closed below the risk line we drew for it three sessions running. We committed on 08-08 that a settled close at or below would trigger a 30% trim; the closes have been 343.80, 343.54 and now 346.36, and the shares sit 19% below their May high after a leadership shake-up in the AI division that included Jeff Dean's departure - a competitive concern, not a price artifact - while the company plans the largest AI capex increase of any hyperscaler at roughly -205bn for 2026. This is a de-risk on the book's largest percentage winner, up 182%, not an exit from the Alphabet story: Google Cloud revenue is up 82% year-on-year to 8bn on a $514bn backlog, the 12-week view is still positive at +11.7%, and stay. OPERATIONAL: carried unexecuted from 08-12 and 08-13, re-priced 343.54 -> 346.36 (the delay has been worth +on). Decisive stop-loss break of a PRE-COMMITTED line is a valid signal-based exit under CLAUDE.md section 9 and is EXPLICITLY EXEMPT from G1/G2/G3; US-34's bar met three times over on the close-below-invalidation limb. RE-VERIFICATION PER US-27, STATED EXPLICITLY: a settled close >= 350.66 WITHDRAWS this card - tonight is (1.23%) short, and the name remains below EMA10, EMA20 350.73 and EMA50 352.98. Trail A11C HOLD cv52 -> A12C AVOID cv50 -> A13C NO TRADE cv0. TIER HOLDS AT THE PRE-COMMITTED 30% - not escalated on the label degradation (US-19), not de-escalated (US-32); the basis is intact, not partly dissolved. NOT the L2 combo: the 12wk is POSITIVE (386.87, +11.69%). ADR-0012 armed genuinely (Avg_Cost 123.00, Gain_ATR 23.54, Unreal_PnL +181.59%, T12_Progress 89.53, PROFIT_ARMED - US-26 passes cleanly) with ZERO of four ticks: (a) no 2nd code, (b) conviction is a FLIP TO NO TRADE, a label change not a comparable decay - declined here exactly as on KLAC 08-05/08-12 and GOOGL 08-05, and it must be declined consistently or tick (b) becomes a back door around US-17, (c) 386.87 vs 1.02xClose 353.29 alive, (d) one NO TRADE print not two. The profit-take path is NOT the basis. Valuation MoS_FV -0.74 too shallow to amplify and G2 would suppress anyway (DCF 145.29 = 2.66x overshoot). Provenance: action price = anchor settled close 346.36; Entry_Price blank on a NoTrade row. Stop on remainder 327.40 -> 332.13 (1.5 ATR), +2.5% over EMA200 323.89. Keep; proceeds; realized ~+ (+181.59%). Est 12wk 386.87 exceeds the 145.29 DCF anchor (2.66x) - projection assumes growth beyond priced-in. Per US-31 the 39-day-stale overlay override ('<345', which tonight's 346.36 un-invalidates) carries no force in either direction. · news |
| RKLB | Rocket Lab Corporation | 80.10 | BUY — QUEUED @72.89 | 72.89 | 65.86 | 80.12 (+9.9%) | 77.61 (+6.5%) | 76.87 (+5.5%) | Rocket Lab just posted record quarterly revenue of, up 62% year-on-year, and guided the current quarter to -265m against a consensus, with backlog at 36bn and demand for its new Neutron rocket described as very high. The market sold it anyway on a wider-than-expected loss, leaving the shares 9% above where our model wants to buy them - so we queue a limit at the pullback rather than pay up. The honest counterweights are that the company is not profitable, the end-of-year Neutron launch window is narrowing, and our own 12-week projection only works from the limit price, not from today's close. OPERATIONAL: adj conviction 71 = 69 raw + 0 val_adj (Val_Signal NA) + 2 scorecard (PullbackBuy_VWAP|buy, n=22, hit_1w.591 - the book's ONLY positive buy cell; L1 satisfied, this is a pullback not a chase). QUALITY GATE IS OFF BY THE NA CARVE-OUT, NOT BY MERIT - Quality_Pass is False on real computed data (F_Score 5, FCF_Yield -0.76%, ROE -8.1%, Quality_Rank); the gate is disabled only because EPV came out negative so Val_Signal reads NA. Note the inconsistency: ASX, with a BETTER F_Score of 6, is hard-blocked tonight purely because its EPV was computable. Logged as systemic finding #1 and the single reason this card exists - read it as an unprofitable hyper-growth name (revenue +55.8%) bought on a technical pullback. QUEUED not bought because Entry_Distance_ATR -1.19 (the 80.10 close is 1.19 ATR / 9.0% past the 72.89 model entry), a decisive fail of the ACT-NOW no-chase bar. Provenance clean: 72.89 is the anchor row's printed Entry_Price (Order_Type Buy Limit); entry range 71.37-74.41 (+-0.25 ATR of 6.0735). Signal is a FRESH single print (A11C WATCH cv49 -> A12C NO TRADE cv0 -> A13C BUY cv69), not a stable multi-snapshot signal - framed and sized accordingly. Sizing: (49.24/100) x x 20% = x 1.00 modulator (MoS_FV NA), / 72.89 = 270.2 -> =; well inside the 60% x cap and trivially covered by. Stop 65.86 = the model's own Stop_Price (1.16 ATR below the limit, adopted as-is); risk if filled (9.6% of). Target_Price 86.98 (+19.3% vs the limit) is the risk objective, distinct from the 4wk estimate. NO DCF ANNOTATION POSSIBLE - DCF_per_share is NA. WARNING: Est_12W 76.87 is 4.0% BELOW the 80.10 close; the forward view only works from the limit, which is exactly what the queue buys and exactly why chasing this at market would be wrong. L6 satisfied - buying the post-earnings pullback three sessions after the 08-10 print, not chasing the gap. HONEST CAVEAT: a 9%-below-market limit is unlikely to fill inside 3 sessions and the record agrees - no_chase_queue n=17, median wh_1w +1.51% [cost]; that is ADVISORY ONLY (US-28), not blended into conviction, and the no-chase bar is not relaxed to force a fill (US-20). · news |
| TSM | Taiwan Semiconductor Manufacturing Co | 430.49 | BUY-STOP — QUEUED @434.43 | 434.43 | 417.21 | 433.20 (-0.3%) | 438.40 (+0.9%) | 482.17 (+11.0%) | TSMC just reported its best month on record - July revenue of NT6bn, up 44.7% on the year - and has raised both its 2026 revenue guidance, to slightly above 40% growth in dollar terms, and its capital spending plan to -64bn. Bernstein lifted its target on a new CPU opportunity it calls the company's next serious growth engine. Despite all of that the shares still sit about 10% below their 52-week high. This is the highest-quality business the book has queued in more than a month - third-best quality rank of 52 names, a near-perfect financial-strength score, 34% return on - and we take a deliberately small starter position that only fills if the stock confirms by trading higher. OPERATIONAL: THE US-38 PRE-COMMITMENT ARMS EXACTLY AS WRITTEN on 08-13 - TSM printed BUY - ANTICIPATE at A12C and again at A13C = 2 of the last 3, conviction 81 >= the 55 anticipate_slot_conviction_floor, and Quality_Pass True so the gate caveat the pre-commitment anticipated does not arise. Adj conviction 71 = 81 raw - 10 val_adj (MoS_FV -14.25, clamped) + 0 scorecard (AnticipationUp_VCP|buy is n=1, REPORT-ONLY - no bias applied). QUALITY GATE PASSES ON MERIT: Quality_Pass True, Quality_Rank, F_Score 9, ROE 34.5%, FCF_Yield 7.7%. Provenance clean: 434.43 is the anchor row's printed Entry_Price (Order_Type Buy Stop), Entry_Distance_ATR +0.34, trigger ABOVE the 430.49 close so it fills only on confirmed strength and is self-limiting - the entire design of this slot; entry range 431.56-437.30 (+-0.25 ATR of 11.4778). Sizing: (60.76/100) x x 20% = x 0.5 size modulator (MoS_FV -14.25 <= -3), x 0.5 anticipate-slot factor = / 434.43 = 13.99 -> = NOTE: last night's US-38 sketch said '~'; the rule's own arithmetic lands at 14 and the written formula governs. STOP 417.21 = 1.5 ATR below the trigger; the model's Stop_Price 399.99 is REJECTED as too loose - 7.9% below entry is not a signal-defined risk line by this book's standard (US-29 applied to a new entry, same treatment as the DELL card of 08-13); risk if filled Target_Price 503.29 (+15.8%) is the risk objective. WARNING: Est_12W 482.17 exceeds the 45.37 DCF anchor by 10.6x, BY FAR the largest overshoot in the book (KLAC 2.60x, GOOGL 2.66x) - the EPV/DCF anchor is not meaningfully calibrated to a TSM ADR, so the -10 val_adj it drives is doing CONSERVATIVE work here rather than informative work. Recorded, not overridden. Signal stable across two prints (cv81 -> cv81); trail A11C AVOID cv62 -> A12C BUY - ANTICIPATE cv81 -> A13C BUY - ANTICIPATE cv81. · news |
| NET | Cloudflare, Inc. | 330.83 | HOLD | 330.83 | 301.75 | 330.09 (-0.2%) | 331.81 (+0.3%) | 365.66 (+10.5%) | Cloudflare rose another 6% to and we are still not adding. The stock is now 14% above the level our model considers a sensible entry, its momentum reading is stretched at 73, and the signal family it now sits on has the worst record of any buy pattern in this book. We hold the existing and raise the protective stop sharply to. OPERATIONAL: ADD DECLINED, 3rd session. adj 45 = 65 raw - 10 val_adj (MoS_FV -943.71, an EPV artifact on a name whose FV_per_share is) - 10 scorecard (MomentumContinuation_Up|buy, n=6, hit.0 - the worst actionable cell in the book); twenty below the floor, and US-16 requires raw conviction >= 85 on this family. RSI 73.30 independently fails the no-chase bar. Family rotated FailedBreakdown_Up_20D -> MomentumContinuation_Up; BUY cv84 -> cv65, Buy_Rank. The 08-13 re-arm (EMA20 +-0.25 ATR, conv >=65) is NOT triggered - EMA20 is now 289.63 and the close sits +14.2% above it. RE-ARM CARRIED, RE-BASED to the live EMA20 289.63 +-0.25 ATR (284.78 - 294.47) with conviction >= 65. Trail 282.89 -> 301.75 (1.5 ATR); model 271.28 looser, rejected (US-29). WARNING, 13TH SESSION: two NET rows in input/usportfolio.json (150 / 0); the loader takes the last, so this position reads Held=N to the scan and unledgered_drift to the monitor. DE-DUPLICATE. |
| NVDA | NVIDIA Corporation | 225.30 | HOLD | 225.30 | 215.45 | 226.98 (+0.7%) | 230.68 (+2.4%) | 240.81 (+6.9%) | Nvidia strengthened again to a 96 conviction reading and we hold all, raising the protective stop to. A partial profit-take would be justified on the technical picture, but we cannot verify the position's actual gain because the purchase cost is missing from the holdings file - and this book does not act on an unverifiable profit. Fill it in. OPERATIONAL: FIFTH CONSECUTIVE SESSION a fully-formed ADR-0012 tick is surrendered to a blank field. RSI_DIV + PROFIT_ARMED print together = tick (a) in textbook form, on an AnticipationUp_VCP row that G1 does NOT rescue (G1 covers only BreakoutUp_*/MomentumContinuation_*), with T12_Progress 93.56 arming on the T12 limb. US-26 DECLINES THE ARM OUTRIGHT: Avg_Cost is blank in input/usportfolio.json so Gain_ATR and Unreal_PnL_Pct are both blank and the gain is unverifiable; with no valid arm there is no armed position for tick (a) to fire against. This is the same missing datum blocking the same decision five times, not five judgement calls. HOLD cv93 -> cv96, +0.54% to 225.30, above EMA20 212.37 / EMA50 208.09 / EMA200 196.24. Trail 213.10 -> 215.45 (1.5 ATR); model 212.63 looser, rejected (US-29). Earnings 08-26. SHARED WITH ustech/ussemicon - EXECUTE ONCE (L7). See systemic finding #2 - fill Cost for NVDA, LLY and SPCX. |
| MU | Micron Technology, Inc. | 949.83 | HOLD | 949.83 | 851.72 | 955.03 (+0.5%) | 1033.59 (+8.8%) | 1097.77 (+15.6%) | Micron gained another 4.2% to and is now 10.5% above our cost, with a 12-week view still pointing 15.6% higher. The model has had no active opinion on it for eleven sessions running, so we simply hold and lift the protective stop by to - the largest single stop raise of this run. Michael Burry disclosed adding a short position in the name on 12-Aug; we note it and do nothing, because a third-party short is not a reason to sell. OPERATIONAL: 11th consecutive NO TRADE cv0. NoTrade_Reason verbatim: 'Consensus conflict + low conviction.' Above EMA20 897.30 / EMA50 890.95, far above EMA200 620.27. Not armed (Gain_ATR 1.38, T12_Progress 86.52, no warning code). ADR-0020 re-checked: Peak Stop 880.00 (07-22) vs live 815.47 = -7.33%, comfortably inside the 15% threshold; after tonight's raise -3.21%. NO FIRE. Trail 815.47 -> 851.72 (1.5 ATR). The Burry short disclosure is explicitly NOT a basis: it is not one of the three origination classes, and with no signal-based sell available US-17 bars acting on the bare NO TRADE. HOLD 100. |
| LLY | Eli Lilly & Co | 1209.00 | HOLD | 1209.00 | 1159.23 | 1218.80 (+0.8%) | 1244.25 (+2.9%) | 1285.32 (+6.3%) | Lilly slipped 0.9% to and the model's conviction went to zero, but nothing about the position changes. The stock remains above all its major moving averages with a 12-week view 6.3% higher, and the stop stays where it is at. OPERATIONAL: HOLD cv68 -> NO TRADE cv0. NoTrade_Reason verbatim: 'Consensus conflict + low conviction.' The move is a FLIP to NO TRADE, not a decay within a label, so it is NOT tick (b) - the same discipline applied to GOOGL and KLAC tonight. PROFIT_ARMED prints a 3rd session and US-26 DECLINES THE ARM: Avg_Cost blank, gain unverifiable. Even with a valid arm G3 would block - Target_12Week 1285.32 is +6.3% above the close on a name above EMA20 1186.11 / EMA50 1153.21 / EMA200 1034.18, RSI 55.7; cutting the winner. G2 independently disqualifies the valuation amplify (DCF 387.04 vs a 1209.00 close = 3.12x overshoot, so MoS_FV -3.36 cannot bump a tier). Trail: 1.5 ATR = 1147.60, BELOW the standing line -> stop 1159.23 STANDS. HOLD 70. FILL Cost (systemic #2). |
| OKTA | Okta Inc. | 154.95 | HOLD | 154.95 | 144.84 | 154.88 (-0.0%) | 166.21 (+7.3%) | 175.07 (+13.0%) | Okta rose 5.2% to, putting the position 10.8% ahead, but the model's conviction fell from - a real deterioration, though four points short of what would trigger a partial sale. The position also no longer qualifies as sitting on a large enough profit to arm a profit-take at all. We hold and lift the stop to. OPERATIONAL: the 16-point conviction drop is a genuine within-label decay but BELOW the >=20-point bar for ADR-0012 tick (b) - recorded, not fired. DE-ARMED this session: PROFIT_ARMED has dropped off the warning row entirely (Gain_ATR 2.25 < 3.0, T12_Progress 91.48 -> 88.51 < 90), so there is no armed position at all. Target_12Week 175.07 = +13.0%, alive. Trail 141.21 -> 144.84 (1.5 ATR); model 134.16 looser, rejected (US-29). Earnings 08-26. WATCH THE CONVICTION LINE: a further >=20-point drop from 45, or a second warning code, re-arms the trim conversation. HOLD 200. |
| SPCX | Space Exploration Tech Corp | — | HOLD | 146.15 | 129.00 | n/a | n/a | n/a | SpaceX has passed its lock-up test decisively. Since insider shares became sellable on 6-Aug the stock has run roughly 35% in five sessions, added about $500bn of market value and climbed back above its IPO price, with Morgan Stanley calling fundamentals largely unchanged and current levels an attractive entry. The remaining supply is also further off than we feared: the next major unlock is tied to third-quarter results in late October or early November, employee restrictions end in mid-December, and Musk's own stake is locked until mid-2027. We hold all with the stop at OPERATIONAL: permanently muted on signals (private-market row, no public quote -> no_close), so this is a qualitative read only and no Action/conviction/entry/target is synthesised. Classification: CONFIRMS-HOLD, strengthened - the 08-13 withdrawal of the 50% trim is reinforced, and there is no ~08-21 tranche-3 cliff in current reporting. Carried at the last confirmed 146.15 close for book marks. Stop 129.00 STANDS; RE-ARM on a settled close below 129.00. Cost still blank - fill it (systemic #2). |
Outcome & track record
Accuracy at the time of this record.
130 Graded decisions Past recommendations old enough to score against what actually happened (target hit / stopped out / closed). This is the evidence base — the bigger the number, the more history behind the calls. Click to inspect every one. See the decisions →
-0.148 1-week calibration How much of the eventual 4-week move typically shows up in the first week (0 = almost none, 1 = most of it). Use it to read the 1-week estimates: a low number means prices drift toward the target slowly, so don't over-read a soft week-1.
52 Published records How many dated end-of-day verdicts exist for this book — the length of the public track record you can browse in the Archive below.