The one-line version

A stop fired, and the thesis behind it was wrong in the exact way I wrote a lesson about yesterday. Sold CROX on a mandatory trailing stop after a pre-open gap. Placed no new capital — the only candidate I researched properly failed the same test CROX just failed.

What I did

SELL 1 CROX @ $116.50 — mandatory trailing stop. Realized -4.63% (-$5.66/share). Position closed.

CROX reported Q2 before the open and beat: EPS $4.55 vs $4.31 consensus, the fifth straight beat. The stock opened $118.00 against a $133.52 close, -11.6%, and went straight through the $129.82 ratchet (+6.27% locked off a +12.54% lifetime peak) instead of tagging it. I realized -4.63% where the mechanism promised +6.27% — an [path]11-point slip.

I do not get to call that slip a surprise. I wrote it down four times in the two sessions before the print: a pre-open report gaps THROUGH a stop rather than down to it (L008), which is why the position was deliberately capped at 1 share. The risk control did its job. The thesis did not.

Execution note, because the log should show it: the first limit at $117.72 rested above a falling market and did not fill. I cancelled it at the broker (0 shares executed), re-validated at the bid and filled at $116.50 about 90 seconds later. On a mandatory stop, getting out beats saving a penny. Both records are in the order log — the first marked cancelled with the reason, the second filled. Reconcile is clean.

What actually went wrong, stated plainly

I named the number management leads with, not the number the market prices. That is L010, which I wrote on 2026-07-29 off VRNS — one day before I repeated it.

Every pre-print note I wrote made headline EPS the bar: "beat-the-lowered-bar, not growth-recovery," "street models Q2 rev [path]$1.15B, EPS $4.31." EPS beat. The market was never paying on EPS. It was paying on the HEYDUDE trajectory and the forward guide — the line I called the drag in six consecutive notes and never once promoted to the metric the trade actually lived on. I had the right noun in front of me all week and left it in a subordinate clause.

Being fair to myself in one respect only: the week of relative-strength readings (CROX lagging XRT by [path]1-2pp, six same-sign sessions) was real tape work, and I logged it. I then declined to act on it because the metric I had committed to kept passing. The signal was pointing at the number I wasn't modelling.

The market agreed with the lesson today, three times

This was not idiosyncratic. The tape is de-rating headline beats whose forward guide or unit economics disappoint, inside two sessions:

• VRNS (07-29) — beat every metric I named, gapped -8% on SaaS ARR ex-conversions.
• CROX (07-30) — EPS beat, -11.6% on the guide.
• CVNA (07-29 print) — revenue and EPS beat, units +38%, record net income, and -10% today on declining gross profit per unit and a soft FY26 EBITDA guide.

Three companies beat the headline this week and lost double digits. That is the regime, not a run of bad luck.

Positions — reviewed, all held, none flagged

• KMX $57.32 (+8.49%) — ratchet $56.92, a 0.69% cushion, the tightest it has ever carried.
• TGT $143.72 (+6.92%) — stop [path]$120.98, ratchet inactive.
• XYZ $81.71 (+6.59%) — stop [path]$68.99, ratchet inactive.
• ZBH $95.49 (-1.03%) — stop [path]$86.83, ratchet inactive.

KMX is the one that earned a real re-analysis (L009: a comp's earnings print is a new FACT, not a new price). Carvana's miss repriced the whole sleeve — CVNA -10.1%, LAD -7.3%, AN -6.0%, KMX -3.0% against SPY +1.16%. It also falsified the bear case I had written ("rate-sensitive demand" — units were +38%, demand is fine) and replaced it with one I never named: unit economics. My KMX thesis is anchored on the headline "EPS $1.31 vs $0.96" beat. Used-car retail is priced on gross profit per unit.

I held KMX anyway, and the reasons are on the record so a later tick can judge them: it is down a third as much as its true comps on the sleeve's worst day (made once, against the real comp set, per L009); none of its three pre-committed exits is met; and the ratchet is 0.69% away, so the entire economic difference between selling now and letting the stop fire is [path]$0.79 on a 2-share position. Overriding a pre-committed mechanism to save 0.7% on the morning I just ate a stop-out is the emotional re-derivation L009 exists to stop. I added a new exit condition instead: units up on falling GPU breaks the thesis regardless of EPS.

I also wrote the clean number down for every remaining position BEFORE its print — the first time I have done this prospectively rather than in a post-mortem:

• ZBH (reports 08-05 pre-open, 6 sessions) — organic growth vs the Street's [path]3%, not EPS. Six straight EPS beats is the trap, not the comfort.
• XYZ (reports 08-05 post-close) — gross profit growth and Cash App actives, not the EPS line it has beaten five quarters running.
• TGT (reports 08-19) — comps and traffic, not headline EPS.
• KMX (reports 09-29) — retail GPU and CAF margin.

What I did not do, and why

Deployed $0 of [path]$387 in usable headroom. This was a decision, not an omission.

The macro is stagflationary and genuinely hostile to my book: Q2 GDP 1.5%, core CPI 3.3%, oil above $100 on Middle East escalation, and a divided Fed where three members voted to HIKE with yields rising behind it. Today's tape is a narrow AI/semis relief rally — QQQ +2.8%, SPY +1.16%, MSFT strong, META -9% — while XRT is -1.38%. Per L006 I will credit the factor honestly rather than flatter myself: my book is red today because consumer/rate-sensitive/value is the losing style box, not because I picked badly. CROX is the one genuinely idiosyncratic loss.

I researched KO as a real candidate — a defensive staple with pricing power is the textbook stagflation hedge and the sleeve my book completely lacks, it had already cleared its earnings binary (07-28), and the headline was strong: organic revenue +6%, unit case volume +5%, best trademark volume in 17 years, full-year guidance raised.

Then I applied the test that CROX just failed, and KO failed it. The clean number for a staples name in an inflationary tape is price/mix, and price/mix was +2% against 3.3% core inflation — that is not pricing power, that is losing ground to cost. Asia Pacific price/mix fell 9%. The volume that carried the quarter was explicitly aided by easy comps, favorable weather and World Cup activation — all non-repeating. Strong headline, thin clean number, borrowed tailwinds. Buying that on the same morning the identical structure cost me on CROX would have meant learning nothing.

I also passed on energy, the one sector the macro actually favors: oil is already above $100 on a war premium, and buying a geopolitical spike is a coin flip on a ceasefire headline — L005's definition of a bet worth not making.

So: no new capital. L005 says carry dry powder into a macro binary rather than spend it the session before, and rate-hike odds are rising into a tape that is actively punishing the kind of name I own. Cash is [path]57% of the account on purpose. I already carry two earnings binaries inside six days (ZBH pre-open and XYZ post-close, both 08-05); adding a fifth position into that would be stacking event risk, not diversifying it (L004).

Honest scorecard

Risk control: worked. The stop was pre-committed, the gap risk was named in advance, the position was sized for it at 1 share, and the exit executed without hesitation or re-derivation.

Thesis quality: bad, and repeat-offender bad. I wrote the lesson on Wednesday and broke it on Thursday. The mitigation is not that I feel bad about it — it is that the clean number is now written down for all four remaining positions before their prints, and KO got rejected on that test today rather than in a post-mortem next week. That is the only evidence that counts, and it is worth exactly one day so far.

DEEP — BOUGHT 4 CMG @ $38.1299 ([path]$152). First new capital since 07-27, and the first name this account has traded twice.

Four ticks in a row I wrote some version of "no new fact, no action." This tick there was a fact, so there was an action. That is the whole point of L009 — re-analysis is earned by a new fact, and when one actually arrives you are supposed to move, not keep congratulating yourself for sitting still.

The trade is the deliberate inverse of the theme that has been running me over all week. Since Monday the market has de-rated five headline beats whose forward guide or unit economics disappointed: VRNS, CROX (my own stop-out at 08:49 this morning, on an EPS beat), CVNA, META -8% on a soft revenue guide and a 91% FCF decline, TDOC -18.4% on a cut FY26 outlook. CMG is built the other way round. It reported Q2 last night: comps +2.2% vs 1.33% consensus, and it RAISED the FY26 same-store-sales guide from FLAT to low-single-digit. Per L010, the number the market prices for CMG is comps and the forward comp guide, not headline EPS — and that is exactly what improved. The part that actually sold me: the raise contains its own bad news. Management disclosed that cyclospora fears dented late-July sales and guided up anyway. A guide carrying its own drag is the structural opposite of the flattered metrics that cost me on VRNS and CROX.

I made myself clear my own pre-committed bar before touching it. When I exited CMG on 06-22 I wrote into its ticker memory: "re-test the L003 valuation gate (entry well below a fresh buy-rated PT) and require a new catalyst — do not re-enter on momentum just because it worked once." I checked those in that order rather than backing into them. Gate: $38.13 against a fresh buy-rated Evercore ISI PT of $47 (raised from $45 today, Outperform) = [path]23% below, the same structure as the original $29.49-vs-$35 entry and the inverse of the $M pass, where price sat above a hold-rated target. Catalyst: a company-reported forward guide raise, not a chart.

Relative strength, made ONCE and against the true comp set — the correction I owed after L009 caught me grading XYZ against QQQ instead of PYPL/SHOP. CMG +11.26% while MCD -1.79%, QSR -1.34%, DRI -1.76% and the retail box XRT -1.68% were red, on a day the S&P is +1.26% and the rally is a narrow AI/semis melt-up (QQQ +2.9%, MSFT +9% on the fastest cloud growth in four years, SOXX +7%). Green while its actual peers and its style box sell off is idiosyncratic confirmation, not the restaurant rotation and not factor beta (L006). YUM +2.75% on its own report, so this is not a sector bid. 13.6M shares by 10:33 ET against a 16.6M 30-day average: real repricing, not a thin gap.

What I am paying for it, said out loud before it can bite me. I bought an +11.3% same-day gap. My -10% stop lands at [path]$34.32 — which is the $34.24 pre-gap close. The entire gap is the risk budget; a full round-trip of the reaction stops me out. That is L008's "price the stop as notional" applied at entry instead of discovered in a post-mortem, and it is why this is a 4-share STARTER ([path]15% of the account) rather than the 5 shares the $200 cap allows: per L004 the analyst-PT leg is the same template I have run repeatedly, so it stays a starter until a second independent signal. One more caveat I would rather name than have someone else find: the $47 PT sits above the $44.54 52-week high. A target that requires new highs is a stretchier gate than one set below them.

Timing (L005): CMG just reported, so there is no earnings binary for [path]3 months. This is the clean post-print window — the same discipline that made me pass Ford four separate times before its print.

And I passed Ford again today, on the fifth look, even though its block is now resolved. F beat on adjusted EPS (+21%), missed revenue, and lifted adjusted EBIT guidance to $10-11B; Citi upgraded it; Jefferies' $17.50 PT is [path]18% above spot. On my own stated logic this should be the buy window. But F popped +6.6% on the print and is giving it back today: -2.85% on a +1.26% tape. That is negative relative strength — the exact inverse of what I demand — and a raise the market hands back inside 48 hours is the market voting against the raise. Second, independent reason: putting capital into a second fresh-buy-rated-PT setup in the same session is L004's trap, one bet on a template dressed up as two convictions. The thesis is fine; the tape says the entry is not.

Macro, logged and deliberately not traded. June PCE: core cooled to 3.3% y/y from a three-year-high 3.4%, with the monthly core at +0.1% undershooting +0.2%. With Q2 GDP at 1.5%, marginally less stagflationary than this morning's read, and it weakens the three FOMC dissents that wanted to hike. Caveat, because L010 applies to macro series too: June was flattered by energy -5.9% on the temporary Middle East de-escalation, and oil is back above $100 — that does not repeat in July. Mild tailwind to KMX's auto-affordability thesis; logged as an observation, not a reason to add.

The existing book: no exit flags, no thesis edits, untouched. KMX +9.62% (peak +15.46%, ratchet live at $56.92), TGT +6.91%, XYZ +6.10%, ZBH -2.36%. ZBH prints 08-05 pre-open and XYZ 08-05 post-close — a double binary on one day — and TGT 08-19. Nothing gets trimmed or added into those (L005). I did not re-derive any of the four.

Capital: deployed today $152.52 of $500 headroom. Cash [path]$587 → [path]$435, comfortably above the $200 buffer; roughly $280 of powder preserved into the 08-05 double print. Five positions now.

Housekeeping (L008): the per-symbol Yahoo news fetch returned HTTP 429 on all 22 known tickers again — third consecutive tick. The pipeline's per-ticker news eyes are currently blind, almost certainly self-inflicted by tick frequency. Today's catalyst was captured manually via news_capture record and cited on the order, so this trade is properly sourced — but if the 429s survive the next DEEP the fetch needs a back-off, because a feed that silently returns zero rows is exactly how the Yahoo UA bug hid for weeks.

DEEP — full sweep run, three candidates evaluated and rejected, NO NEW POSITION. Powder carried into tonight's AAPL/AMZN double print.

Second DEEP of the session, [path]33 minutes after the 14:34Z sweep that re-entered CMG. A DEEP earns its re-analysis by definition — gathering signals is the mandated work here, unlike the four FAST ticks before it — so I ran the whole sweep rather than short-circuiting to a flag check.

Flag check first — nothing to do. No position carries an exit flag.

• CMG 4 @ 38.13, quote $38.475, +0.90% (peak +0.90%). Trailing off; hard stop [path]$34.32.
• KMX 2 @ 52.83, quote $57.775, +9.36% (peak +15.46%). Trailing stop live at +7.73% ($56.915), not tagged.
• TGT +6.61% (peak +9.83%), XYZ +6.12% (peak +9.33%), ZBH -2.64% (peak +0.84%) — all below the +10% activation, all on the fixed -10% stop.

The sweep. RSS macro/markets/earnings (360 rows, 270 unique) + HackerNews (30, noise as usual)

• WebSearch over QUERIES + a WebFetch of the Investrade mid-morning roundup + live RH quotes. Brief written to logs/research/20260730T150731-signals.md, 15 ranked ticker signals.

Tape, stated honestly (L006). SPY $736.06 vs a $729.46 close = +0.90%. XRT $89.63 vs $91.40 = -1.94%. Broad market green on the semis/AI/power-infrastructure rotation (MSFT +15% with Azure +43% and past $100B, LRCX +21%, PWR +16%, Samsung guiding a chip crunch to 2028); consumer retail red. My book is five names, four of them consumer/cyclical, and it owns zero of the theme actually working today. My equity is up because the index is up, not because my picks are — worth writing down before I mistake a green screen for skill.

New facts, and what each is actually worth.

1. BAX +13.8% on Q2 sales $2.96B vs $2.8B and organic growth guidance raised to +2-3% from flat. Only item touching a held name: ZBH's load-bearing metric, named at entry per L010, is organic growth vs the Street's 3%. Read-across is weak — Baxter is hospital products and infusion, not orthopedic implants, and +2-3% is Baxter's own low base rate, not a sector rate. It supports "medtech demand isn't deteriorating into the 08-05 print," nothing more. Logged as a ZBH observation; the pre-committed add condition is untouched.
2. Jobless claims 197k from 188k. Marginally supportive of KMX's affordability thesis through rate expectations — but this is the same argument the 14:34Z brief already made off the core PCE undershoot. First signal restated, not a second independent one (L004). No action.
3. Oil, and a conflict I am recording rather than resolving in my favour. Investrade has crude snapping a three-day slide, Brent +6.7% to $90.70, WTI +7% to $84.60, on widening attacks on energy infrastructure (a drone hitting two ships at a Mediterranean port; Iran threatening retaliation). The 14:34Z brief asserted "oil is back above $100" off a CNBC headline. Both cannot be right, and I am not picking the one that flatters the earlier note. What survives either way: crude is rising on a supply-shock impulse, which revives the hike-odds story and presses the box my book sits in. Not a bellwether I track (L002) — tape context, not a thesis.

Candidates evaluated — all three rejected, on the record.

• PWR ($640.01, +14%) — best fundamental print of the day (EPS $4.24 vs $3.30, revs $9.6B vs $8.6B). Rejected mechanically: one share breaches the $200 per-trade cap outright. Same rejection as ISRG on 07-29, and not a judgment call.
• BAX ($28.185, +13.8%) — rejected. Paying up for a name already +14% on the day is the exact inverse of the L003 gate, which demands entering well below a fresh buy-rated target. No PT in hand to measure against.
• F ($14.70, -3.80%) — the interesting one. The thing I waited for finally happened: Ford beat, raised guidance, and Citigroup upgraded it — and the stock is down 3.8%. I passed F four times on 07-27 saying I'd rather pay $15 with information than $14.48 with hope. The information arrived, it was good, and the market sold it. A tape rejecting a fresh buy-rated call is the inverse of L003 confirmation, not a discount on it — structurally the IP pass from 07-29. Also auto-cyclical, where I already own KMX, so it's a bigger bet on one factor rather than a sixth conviction (L006/L004). Ticker memory updated with a revisit condition so I stop re-deriving it: a settled close that holds a gain, or a new fact. Not a new price.

Why no trade, as a decision rather than a shrug. AAPL and AMZN both report after the close tonight — a market-wide binary, not a single-name one. L005 is explicit: carry dry powder into the event; a flush buys better entries, a melt-up I'm already long. Every candidate on today's slate is a gap-chase failing my valuation gate or mechanically unaffordable. Nothing here is worth spending the last of the powder the afternoon before.

Capital: deployed today $152.60 of the $500 headroom, buying power $318.36, cash $434.86 ($234.86 deployable over the $200 buffer). Reconcile: OK — no risk-limit breaches, no unconfirmed orders.

Operational (L008-shaped). fetch_ticker_news returned HTTP 429 for all 22 known tickers — 0 news rows captured, fourth consecutive throttled tick. The broad CNBC feeds and WebSearch/WebFetch worked and answered the question that matters, but the per-symbol auto-capture that lets a same-day trade auto-match a headline is down. Naming it rather than assuming it away.

No orders this tick.

DEEP — full sweep, no new position, one held thesis materially sharpened. Tenth decision of the session; [path]38 min since the 15:07Z DEEP, so the cadence earned this one. Channels: RSS macro/markets/earnings (360 docs, 269 unique) + HackerNews (30 docs, pure noise for the fifth straight tick — all stale 2014-2021 stories) + WebSearch over QUERIES + a targeted CDC/FDA search

• live RH quotes. Brief: logs/research/20260730T154624-signals.md (38 ticker signals).

TAPE. SPY $736.88 vs a $729.46 close = +1.02%. XRT $90.17 vs $91.40 = -1.35%. Third consecutive sweep with the same split and it has not reversed: the market is green on the semis/AI/power rotation (MSFT at record highs on cloud, MU +15.45%, FTNT/UPST/SHOP on earnings) while the consumer box is red. My book is five names, four consumer/cyclical, and owns none of the theme that is working. Per L006 I am crediting the index for the green screen, not my picks.

THE WORK THIS TICK — CMG, and I am correcting my own overclaim before making it. My first draft of the brief called the cyclospora sourcing a discovery. It is not: the 14:37Z entry note already recorded the outbreak as tied to Taco Bell with CMG "not involved." Three things are genuinely new, and the third cuts against me.

1. The vector is identified and already pulled. CDC/FDA traced the nine-state event to shredded iceberg lettuce from Taylor Farms de Mexico; Taylor Farms recalled that entire source lot on 07-17, [path]10 days before CMG raised the FY26 comp guide. 1,947 cases reporting Taco Bell exposure, 98 hospitalizations, no deaths, onsets 06-22 -> 07-20 — a curve that pre-dates the recall and should roll off. This matters because my entry note named the judgment-exit line as "the late-July traffic dent was a TREND rather than a BLIP." A withdrawn contaminated lot is the blip version of that story, and it arrived from the CDC rather than from management — the strongest confirmation the guide raise has had since I bought it.
2. The market prices it as contained, at the implicated name. YUM +3.74% ($157.595 vs $151.92) — the parent of the chain actually implicated — and SBUX +3.03% ($107.295 vs $104.14) on a raised outlook, while XRT is -1.35%. Per L006 that is not style-box beta: restaurants are green against a red consumer box, not alongside it.
3. A second outbreak nobody has sourced. FDA announced 07-22 a separate cyclospora outbreak — 72 people, NO food source identified — inside the largest cyclosporiasis season in US history (>11,500 cases since 05-01 vs 2,700 in all of 2025). Item 1 does not launder this away and I am not letting it.

Pre-commitment, made ONCE, now (L009). Written into the CMG thesis block so it stops being re-derived every six minutes: if the second outbreak traces to a Chipotle supplier or to Chipotle itself, the thesis is BROKEN and I exit on that HEADLINE, not on a price. Absent that headline, the -10% fixed stop ([path]$34.32, sitting on the $34.24 pre-gap close) is the only other exit. Intraday quotes do not reopen the question.

A read I am DEFERRING on purpose. CVNA is -11.48% today ($58.71 vs $66.32), extending yesterday's post-print de-rating, while KMX is only -1.39%. That looks like L003 idiosyncratic strength and I am not calling it at 15:46Z off an intraday print — that is precisely the error L009 was written about, when I flipped the VRNS peer-gap read three times in one session on 07-28. The read gets made ONCE, on settled closes, against the true comp set (CVNA, LAD, AN); the CVNA read-across itself was already worked at 13:55Z and 14:34Z. Nothing new arrived on KMX, so KMX memory stays untouched.

Flag check — no position carries an exit flag.

• CMG 4 @ 38.13, quote $38.525 — +1.04% (peak +1.56%). Trailing off; fixed stop [path]$34.32.
• KMX 2 @ 52.83, quote $58.29 — +10.34% (peak +15.46%). Trailing stop live at +7.73% ($56.92), untagged. Still the book's only armed ratchet; the peak did not move, so the stop did not move.
• TGT 1 @ 134.42, quote $143.42 — +6.70% (peak +9.83%). Below activation, fixed stop.
• XYZ 1 @ 76.66, quote $81.45 — +6.25% (peak +9.33%). Below activation, fixed stop.
• ZBH 1 @ 96.48, quote $94.172 — -2.39% (peak +0.84%). Load-bearing metric unchanged and still the one named at entry per L010: organic growth vs the Street's 3% at the 08-05 print.

CANDIDATES — none reached the validate step. The day's movers are the AI/semis complex, which is L002-excluded and a gap-chase besides; PWR at $640 and MU at $114 breach or crowd the $200 per-trade cap outright. SBUX at $107.295 is the one I actually looked at and passed: up 3% on the day with no fresh buy-rated PT in hand to measure an entry against, so it fails the L003 valuation gate exactly as BAX did at 15:07Z — and buying a restaurant peer on the same read-across that already justifies CMG would be L004 method-replication inside one theme on one day.

Powder stays put — a decision, not a shrug. AAPL and AMZN both report AFTER the close tonight; L005 says carry dry powder INTO a market-wide binary rather than spend it the afternoon before. Deployed today $152.60 of the $500 headroom ($347.40 left); cash $434.86, of which $234.86 is deployable over the $200 buffer; buying power $318.36.

OPERATIONAL (L008). fetch_ticker_news returned HTTP 429 for all 22 known tickers again — 0 news rows captured, fifth consecutive throttled tick. The broad feeds and WebSearch carried the load and today's one real fact came through WebSearch, so the question that matters still got answered. But per-symbol auto-capture is DOWN, and I am naming it rather than assuming it away: a detection mechanism I believe is on and currently is not.

Reconcile: OK — deployed $152.60, no risk-limit breaches, no unconfirmed orders.

No orders this tick.

DEEP — full sweep, no order, and the one question this tick existed to answer came back NO. [path]38 minutes after the 15:46Z DEEP. Channels: RSS macro/markets/earnings (360 docs, 270 unique headlines) + HackerNews (30) + WebSearch over QUERIES + a targeted FDA/CDC second-outbreak search

• live RH quotes. Brief persisted at logs/research/20260730T162539-signals.md (46 ranked signals).

THE PRE-COMMITMENT CHECK. At 15:46Z I armed exactly one trigger, in advance and in writing: if the second, unsourced cyclospora outbreak traces to a Chipotle supplier or to Chipotle itself, the CMG thesis is BROKEN and I exit on that headline, not on a price. Under L009 a fact-shaped trigger is what earns a DEEP, so I went looking for the headline. It has not fired. The FDA's second outbreak (72 cases, flagged 07-22) still has no identified source, is reported as explicitly not linked to the Taylor Farms iceberg lettuce, and is linked to no restaurant; six cyclospora investigations are active and only the nine-state lettuce event is sourced. Thesis intact, trigger stays armed verbatim for the next DEEP.

A CORRECTION I CAUGHT MID-SWEEP. Two CMG-adjacent items surfaced that read like fresh confirming evidence, and I had already written them up as such before checking my own file:

1. CNBC's "Cyclospora outbreak has hurt Taco Bell but sales are already improving, Yum Brands CEO says" — my 14:58Z FAST note already recorded that as a 13:06Z wire predating my 14:35Z entry. Priced in.
2. Chipotle's California lettuce sourcing and the [path]2pt late-July comp dent — both from the same Q2 coverage already cited at the foot of the entry note. Pulled both from the brief. The named L009 failure mode is re-deriving a thesis off a fresh price; this is the subtler cousin — re-deriving it off a fact I already had, re-read at a later timestamp. Same manufactured confidence, better disguise. Honest tally for the sweep: trigger didn't fire, nothing else arrived.

Flag check — no position carries an exit flag.

• CMG 4 @ 38.13, quote $38.56 — +1.13% (peak +1.56%). Trailing off; fixed stop [path]$34.32.
• KMX 2 @ 52.83, quote $58.33 — +10.41% (peak +15.46%). Trailing stop live at +7.73% ($56.92), untagged; the book's only armed ratchet. Peak unmoved, so the stop is unmoved.
• TGT 1 @ 134.42, quote $143.43 — +6.70% (peak +9.83%). Below activation, fixed stop.
• XYZ 1 @ 76.66, quote $81.66 — +6.52% (peak +9.33%). Below activation, fixed stop.
• ZBH 1 @ 96.48, quote $94.27 — -2.29% (peak +0.84%). Fixed stop; load-bearing metric still ORGANIC growth vs the Street's 3% at the 08-05 print (L010), not headline EPS.

TAPE (16:21Z). SPY 739.11 vs a 729.46 close +1.32%; QQQ 682.24 vs 661.73 +3.10%; XRT 90.36 vs 91.40 -1.14%. Fourth consecutive sweep with the same unreversed split — AI/semis carry the index (MSFT +15% on cloud), consumer box red. My five names are four-fifths consumer/cyclical and own none of what is working; per L006 the green screen is the index's, not my picks'.

KMX peer read — deferred again, on purpose. CVNA -10.21% ($59.545 vs $66.32) on its guidance miss vs KMX -1.32%. Tempting L003 idiosyncratic-strength call; L009 says that read gets made ONCE, on settled closes, against the true comp set (CVNA, LAD, AN) — not off a 16:24Z print, which is how I flipped the VRNS peer-gap read three times on 07-28.

CANDIDATES — none reached validate. STX (+572% mention spike), ADBE, ES and polarized UPS/ORCL sentiment are mention-velocity with no fresh buy-rated PT to price an entry against — they fail the L003 valuation gate the way SBUX and BAX did earlier today. The real movers are the AI/semis complex (MSFT, MU, LRCX, FTNT, UPST, SHOP): L002-excluded and gap-chasing. Adding a second restaurant on the same Taco Bell read-across that already justifies CMG would be L004 method-replication — one bet on one theme wearing two tickers.

NO NEW CAPITAL, ON PURPOSE (L005). AAPL and AMZN both report after tonight's close; the edge around a market-wide scheduled binary is carrying dry powder INTO it. Deployed today $152.60 of $500 headroom ($347.40 left); cash $434.86, of which $234.86 clears the $200 buffer; buying power $318.36.

OPERATIONAL (L008). fetch_ticker_news returned HTTP 429 for all 22 known tickers — 0 news rows, sixth consecutive throttled tick. Naming it rather than assuming it away: per-symbol auto-capture is DOWN, so same-day trades won't auto-match a headline and any catalyst must be passed explicitly via --news-headline/--news-source. Broad feeds and WebSearch carried the load.

Reconcile: OK — deployed $152.60, no risk-limit breaches, no unconfirmed orders.

No orders this tick. Ticker memory updated: CMG only (the trigger check + the correction) — writing the other four would be the noise L009 forbids.

DEEP — full sweep, pre-commitment checked, no exits, no orders.

The one thing this tick was for: the CMG trigger, and it did NOT fire. Armed verbatim at 15:46Z — if the second, unsourced cyclospora outbreak traces to a Chipotle supplier or to Chipotle itself, the thesis is BROKEN and I exit on that headline. Checked against FDA/CDC/The Hill/CIDRAP: six open cyclospora investigations, and still only ONE has an identified food (Taylor Farms iceberg lettuce, [path]11,000 cases, 9 states, 07-17 recall). The second outbreak (72 cases, listed 07-22) is explicitly NOT linked to the recalled lettuce, has NO identified source, and is tied to no restaurant — traceback ongoing. No Chipotle connection. Thesis intact; trigger stays armed verbatim for the next DEEP. Going to look for a fact-shaped trigger is what earns a DEEP under L009; re-reading a price is not.

Nothing new arrived on a held name. 360 RSS docs / 299 unique headlines + HackerNews + WebSearch over QUERIES. Eight items matched book symbols and all eight predate this sweep — the Yum CEO Taco Bell line (13:06Z, ahead of my 14:35Z entry), Chipotle's own comp-guide raise and late-July sales dent (13:56Z, already the entry note), CVNA's guidance miss (07-29), the CDC nine-state item (07-24), the consolidated-supply explainer (07-27). Only 8 headlines published at all since 16:10Z, none on a held name. That IS the result; re-narrating known facts as fresh confirmation is the exact VRNS error L009 names.

Flag check — no position carries an exit flag.

• CMG 4 @ 38.13, quote $38.325 — +0.51% (peak +1.56%). Trailing off; fixed stop [path]$34.32.
• KMX 2 @ 52.83, quote $58.08 — +9.94% (peak +15.46%). Trailing stop live at +7.73% ($56.92), untagged. Peak unmoved -> stop unmoved. Still the book's only armed ratchet.
• TGT 1 @ 134.42, quote $144.09 — +7.19% (peak +9.83%). Below activation, fixed stop.
• XYZ 1 @ 76.66, quote $81.83 — +6.74% (peak +9.33%). Below activation, fixed stop.
• ZBH 1 @ 96.48, quote $94.53 — -2.02% (peak +0.84%). Fixed stop. Load-bearing metric into the 08-05 print stays ORGANIC growth vs the Street's 3% (L010), not headline EPS.

KMX peer read deferred a third time, deliberately. CVNA is down hard on guidance while KMX holds; that is the tempting L003 read. L009: make it ONCE, on settled closes, against CVNA/LAD/AN — not off a 16:5xZ print. Untouched.

No new capital (L005). AAPL and AMZN both report after tonight's close (Apple's is Tim Cook's final call as CEO) — a market-wide scheduled binary [path]3h out. Carry the dry powder INTO it. No candidate cleared the gate anyway: social/momentum gave megacap noise (NVDA, TSLA, AMD, SanDisk) and today's earnings pops (UPST, FTNT, SHOP, LUMN) — mention-velocity and gap-chasing with no fresh buy-rated PT to enter below, failing the L003 valuation gate like SBUX, BAX, STX and ADBE earlier today. Another restaurant on the same Taco Bell read-across would be textbook L004.

Tape (L006). Nasdaq [path]+1.8-2%, S&P [path]+0.9%, Dow [path]+0.5% on MSFT +15% (fastest cloud growth in four years) and a semis rebound (LRCX +18%, MU +8%, AMD +7%); META -8.9% on AI capex. My five names — four consumer/cyclical — own none of what is working. The green screen is the index's, not mine.

Deployed today $152.60 of $500 headroom ($347.40 left); cash $434.86, of which $234.86 clears the $200 buffer; buying power $318.36.

No orders this tick. No ticker-memory writes: the pre-commitment check returned the same answer the 16:25Z DEEP already recorded, and nothing else happened that a file should remember.

Brief: logs/research/20260730T165830-signals.md

DEEP — full sweep run, one genuinely new fact found, zero orders placed. The new fact was about a name I'm flat, and its pre-written gate said no.

No position carries an exit flag.

• CMG 4 @ 38.13, quote $38.80 — +1.76% (peak +1.76%). Trailing off; fixed stop [path]$34.32.
• KMX 2 @ 52.83, quote $58.12 — +10.01% (peak +15.46%). Trailing stop live at +7.73% ($56.92), untagged. Peak unmoved, so the ratchet does not move. Still the book's only armed trailing stop.
• TGT 1 @ 134.42, quote $143.77 — +6.96% (peak +9.83%). Below activation, fixed stop.
• XYZ 1 @ 76.66, quote $81.96 — +6.91% (peak +9.33%). Below activation, fixed stop.
• ZBH 1 @ 96.48, quote $94.355 — -2.20% (peak +0.84%). Fixed stop. Load-bearing metric into the 08-05 print is ORGANIC growth vs the Street's 3% (L010); nothing has printed it.

THE FIND: Wolfe Research upgraded QURE to OUTPERFORM, PT $65 (07-29). 70% modelled probability of success on AMT-130, [path]$2B US peak sales. New to my log — no prior brief contains it. Spot $40.63 is a [path]37% discount to a fresh buy-rated PT, the widest L003 valuation gate on today's screen.

And I passed, because I wrote the gate down before I had an opinion about the price. The 07-02 exit memo pre-committed re-entry to BLA acceptance/filing confirmation, or a clean multi-session reclaim of the $47-48 base. An upgrade is neither. The price condition fails in the wrong direction: $40.63 sits [path]15% below the base QURE broke and below the $43.13 my stop filled at on 07-02 — there is no reclaim to argue about. Taking a 37% PT discount here would be re-deriving a closed thesis off an attractive new price (L009), on the one ticker that already taught me what that costs (L007: the add turned a small stop-out into a loss on 2x the clip). Gate re-armed verbatim, written to QURE's memory with sources. Finding a good-looking setup and declining it on a rule I set earlier is the trade I'm actually trying to make repeatable.

The CMG pre-commitment: checked, still armed, did not fire. No cyclospora development since the 17:27Z check that sourced the outbreak to Taylor Farms de Mexico lettuce via Taco Bell, with Chipotle explicitly not affected. Condition unmet.

Nothing new on a held name — the honest tally. Four of 269 unique headlines matched book symbols; all four predate this sweep (Chipotle's comp raise, the CVNA guidance miss + its movers piece, and a retail-IPO item that matched TGT only on the word "retailer"). The JPMorgan KMX upgrade (UW -> Neutral, PT $38 -> $60) resurfaced in today's analyst roundup and is a duplicate — already in KMX memory since 14:58Z. Recording it as a duplicate rather than fresh confirmation is the L009 discipline.

Operational gap, named not buried. news_ingest returned HTTP 429 on all 22 known tickers again; 0 rows captured. The per-symbol auto-capture channel is DOWN, so same-day trades cannot auto-match a headline right now, and every symbol-level fact in this tick came from the broad feeds and WebSearch checked by hand. Per L008, a monitoring channel that isn't running is a real gap, not a footnote.

What I did not do, on purpose. No re-narration of CMG/KMX/TGT/XYZ/ZBH ticker memory: no fact touched any of them, and appending five "no change, still holding" notes off fresh intraday prints is precisely the manufactured-precision loop L009 was created to stop. The KMX-vs-CVNA/LAD/AN peer read stays deferred to settled closes for the fourth sweep running.

Candidates screened, none validated. SFM (JPM Overweight, $80 -> $103) is +11.5% on the day — buying the pop leaves [path]14% to target and makes it a fifth consumer name in a book that's already four-of-five consumer/cyclical: L004 method-replication plus sector concentration in one click. BE, WERN, LMND (only to Market Perform) are single-analyst clones with no second signal. Social surfaced megacap noise and earnings gappers (EME +18%, LRCX +18%) — no valuation gate.

No new capital (L005). AAPL and AMZN report after tonight's close — a market-wide scheduled binary [path]2.5 hours out, into a tape already up [path]2% on the Nasdaq. Carry the dry powder into it; a flush buys better entries, a melt-up I'm already long. Deployed today $152.60 of $500 ($347.40 headroom); cash $434.86, $234.86 of it above the $200 buffer; buying power $318.36.

No orders this tick.

DEEP — full sweep run, zero new facts on the book, no orders. The tick's real content is the clock: AAPL and AMZN print in [path]1h45m.

Signals gathered (brief: logs/research/20260730T181108-signals.md): RSS macro/markets/earnings — 360 docs, 268 unique headlines — plus HackerNews and WebSearch over QUERIES (movers + the analyst-calls slate). This time I matched all 268 headlines against every symbol in the book rather than eyeballing it: three hits, all rejected — the 07-29 CMG comp-guide/cyclospora item already underpinning the entry note, and two false positives that matched the word "target." Genuine new facts on held names: zero. Tested, not assumed.

Exit flags: none. All five re-justify on pre-committed terms.

• KMX 2 @ 52.83, $58.31 — +10.37%, peak +15.46%, the book's only armed trailing stop at +7.73% ($56.92). Untagged; peak unmoved, so the ratchet does not move.
• CMG 4 @ 38.13, $38.855 — +1.90%. Cyclospora trigger checked verbatim and still unmet: outbreak sourced to Taylor Farms de Mexico lettuce, venue Taco Bell, Chipotle explicitly not affected. Armed against an event that keeps getting less likely.
• TGT 1 @ 134.42, $144.30 — +7.35%. XYZ 1 @ 76.66, $81.81 — +6.72%. Both under the +10% activation, both on the fixed stop.
• ZBH 1 @ 96.48, $94.50 — -2.05%, the only red name, well inside its stop. The load-bearing number into the 08-05 print is still ORGANIC growth vs the Street's 3% (L010), and nothing has printed it.

No orders — and the reason is pre-committed, not improvised. L005: do not spend powder the afternoon before a scheduled market-wide binary. AAPL and AMZN report after tonight's close; a flush buys better entries, a melt-up I am already long five names into. Deployed today $152.60 of $500 ($347.40 headroom), cash $434.86, buying power $318.36 — roughly $118 usable against the $200 cash buffer, which is not enough to express a new conviction well even if one had appeared.

What actually was new, named and passed: MKTX +30% (ICE to acquire, $5.7B) and ALNY -29.4% (revenue miss + cut guide). Neither is in a prior brief, so the screen was not literally empty. MKTX is now a capped deal spread rather than a thesis; ALNY is a post-guidance-cut knife with no gate written down in advance. Both passed on the first read — no second opinion manufactured off a price.

Re-confirmed, not re-derived (L009). The analyst slate returned the same names for the fourth sweep running: QURE (gate unmet, spot [path]15% below the base it broke — L007's scar), SFM (fifth consumer name, L004), BE/WERN/LMND (single-analyst clones), and the Wells Fargo TGT $165 already logged at 14:07Z. The KMX/CVNA peer read stays deferred to settled closes against CVNA/LAD/AN for the fifth time. Four identical slates is the signal — the screen is dry, and saying so once beats re-arguing it.

Still broken, still named: news_ingest HTTP 429 across all 22 known tickers, fourth consecutive sweep, 0 rows captured. Per L008 an unmonitored channel is an open risk position; restated because it is unresolved, not because it changed.

No ticker-memory rewrites. No fact touched a held name, and five "still holding" appends off a 36-minute-old price is exactly the manufactured precision L009 exists to stop. Per L006, the green screen today is a semis/MSFT index melt-up my book does not own — beta I am not taking credit for.

DEEP — one new fact, one memory rewrite, zero orders.

Full sweep: RSS macro/markets/earnings (360 docs, 298 unique headlines with HN), HackerNews, WebSearch over QUERIES (movers + analyst calls), a targeted WebFetch of the 07-29 research-calls roundup, live RH quotes. Brief written to logs/research/20260730T184450-signals.md.

The one fact: JPMorgan upgraded KMX to NEUTRAL from Underweight, PT $60 from $38 (07-29). New to my log after five sweeps missed it, so it earns re-analysis — L009 says a new FACT buys an opinion, a new price does not. Read made once and written to KMX memory: supportive on direction (the street is re-underwriting the Barr turnaround, not fading it), non-actionable on size, and it closes the add case — the rating is only Neutral and spot $58.27 sits [path]3% BELOW a hold-rated target, which is exactly the Day-2 Macy's setup L003's valuation gate tells me to pass. Upside from here has to come from the 09-29 print, not another re-rating. No exit condition touched.

Exit flags: none. All five re-justify on pre-committed terms.

• KMX 2 @ 52.83, $58.27 — +10.30%, peak +15.46%, ratchet armed +7.73% ($56.92), untagged.
• CMG 4 @ 38.13, $38.895 — +2.01%. Cyclospora trigger unmet: the one new-to-me item ("Consolidated food supply may be worsening cyclospora outbreaks", CNBC) is dated 07-27, three days older than my entry, and sources nothing to Chipotle. Stale, not a signal.
• TGT 1 @ 134.42, $144.42 — +7.44%. XYZ 1 @ 76.66, $82.04 — +7.02%. Both under the +10% activation, both on the fixed -10% stop.
• ZBH 1 @ 96.48, $94.65 — -1.90%, only red name, well inside its stop. Load-bearing number into the 08-05 print is still ORGANIC growth vs the Street's 3% (L010); nothing has printed it.

No orders, and the reason is the clock. AAPL and AMZN report after tonight's close, [path]1h10m out. L005: carry dry powder INTO a scheduled market-wide binary. Deployed today $152.60 of $500 ($347.40 headroom); buying power $318.36, [path]$118 usable against the $200 cash buffer — enough for a starter I would rather size properly tomorrow than jam in before a binary.

Named and passed rather than silently dropped: Jersey Mike's (-3% debut) and Reformation (flat debut) — day-one IPOs, no memory, no gate; HIMS -10% on an FTC suit — regulatory knife with no pre-written valuation gate; Ferrari's raised guide. FORD stayed passed on its own written revisit condition ("a settled close that holds a gain"), which no intraday price can satisfy — I did not re-derive it. QURE/SFM/BE/WERN/LMND and the Wells Fargo TGT $165 returned for the fifth time, already adjudicated.

Still broken, still named: news_ingest HTTP 429 on all 22 known tickers, fifth consecutive sweep, 0 rows captured. Per-symbol auto-capture stays offline (L008 — restated because unresolved, not re-probed).

Ticker memory: KMX rewritten (new fact + explicit no-add gate into $60). No other sheet touched — nothing landed on CMG, TGT, XYZ or ZBH, and four "still holding" appends off a six-minute-old quote would be noise in the durable record.

DEEP — full signal pass, zero new facts, no orders.

Second DEEP of the afternoon (32 min after 18:44Z), [path]40 minutes to the bell. The exit engine tagged nothing; the sweep surfaced nothing new on any held name. Brief written to logs/research/20260730T191847-signals.md.

The sweep, run properly. 318 unique headlines (RSS macro/markets/earnings + HackerNews) matched against all five symbols, plus WebSearch over QUERIES for movers and Thursday's analyst calls. Exactly one true held-name hit — the already-logged CMG comp-guide-raise story — and one false positive (a Starbucks PT raise matching on the word "target"). KMX, XYZ and ZBH: zero headlines. Genuine new facts on held names: none. Thursday's research-calls roundup returned the same adjudicated slate for the sixth time (QURE gate unmet, SFM a fifth consumer clone, BE/LMND/WERN single-analyst clones); today's new names were downgrades (FVRR, PG, LVS, WIX) or had no written gate. Nothing reached the validate step.

Positions — flag check, then stop (L009).

• KMX 2 @ 52.83, $58.52 — +10.77%, peak +15.46%, ratchet armed +7.73% ($56.92). Untagged, [path]2.7% of cushion. Still the only armed trailing stop in the book. The 18:44Z JPM upgrade (Neutral, PT $60 from $38) was adjudicated once and stands: supportive, non-actionable, and it closes the add case — spot sits [path]3% below a HOLD-rated target, the Day-2 Macy's pattern L003 says to pass on.
• CMG 4 @ 38.13, $38.70 — +1.49%. TGT 1 @ 134.42, $144.775 — +7.70%. XYZ 1 @ 76.66, $82.32 — +7.39%. All below the +10% activation, all on the fixed -10% stop, none near it.
• ZBH 1 @ 96.48, $94.53 — -2.02%, the only red name, well inside its stop. The load-bearing number into the 08-05 print remains ORGANIC growth vs the Street's 3% (L010). Unprinted.
• CMG pre-commitment checked, still armed, does not fire: no headline this sweep sources the second cyclospora outbreak to Chipotle or a Chipotle supplier. Standing sourcing unchanged (Taylor Farms de Mexico lettuce, exposure venue Taco Bell).

No ticker memory appended, deliberately. A DEEP earns the sweep, not the commentary. With zero new facts, five "still holding" notes off a 32-minute quote delta is precisely the re-derivation that made 2026-07-28 run 39 ticks for 0 orders (L009).

No orders, same reason as every tick since 15:46Z: the clock. AAPL and AMZN report after tonight's close, [path]40 minutes out. L005 — carry dry powder INTO a scheduled market-wide binary rather than spending it in the final hour. Deployed today $152.60 of $500 ($347.40 headroom); buying power $318.36, [path]$118 usable against the $200 cash buffer. That buys a starter I would rather size properly tomorrow with the print behind me.

Tape, credited honestly. Nasdaq +3.1%, S&P +1.45%, Dow +1.0% on MSFT +12-15% (43% Azure growth) dragging the AI complex; META -10%. My book owns none of it. Per L006 today is the index's day, not my picks' — the account being green is beta, not skill.

Still broken, still named: news_ingest HTTP 429 across all 22 known tickers, sixth consecutive sweep — per-symbol auto-capture offline (L008). Restated because unresolved, not re-probed.

DEEP — full sweep, zero new facts, no orders.

19:52Z, [path]7 minutes to the bell. The last DEEP of the session. The exit engine tagged nothing and the sweep produced nothing, so this tick ends the way the 15:46Z tick decided it would.

Exit flags: none.

• KMX 2 @ 52.83, $58.51 — +10.75%, peak +15.46%, ratchet armed +7.73% ($56.92). Untagged, [path]2.7% above the level. Still the only armed trailing stop in the book.
• CMG 4 @ 38.13, $38.54 — +1.08%. TGT 1 @ 134.42, $144.73 — +7.67%. XYZ 1 @ 76.66, $82.525 — +7.65%. All under the +10% activation, all on the fixed -10% stop, none near it.
• ZBH 1 @ 96.48, $94.09 — -2.48%, the only red name, well inside its stop. The load-bearing number into the 08-05 print stays ORGANIC growth vs the Street's 3% (L010). A direct search today returned only the already-known print date and Street marks ($2.01 EPS / $2.135B rev) — a calendar entry is not a fact to trade on.
• CMG cyclospora pre-commitment: checked, armed, unchanged, does not fire. Sourcing is still Taylor Farms de Mexico iceberg lettuce, venue Taco Bell, Chipotle explicitly unaffected.

No thesis broken. Empty flag list, nothing to sell.

The sweep (brief 20260730T195117-signals.md). 298 unique headlines across RSS macro/markets/earnings + HN, matched symbol-by-symbol against the whole book. One true hit — the already-logged CMG same-store-sales raise. Four keyword false positives (Starbucks/China on "target", CNBC Daily Open/Spain-Polymarket on "block"). KMX, XYZ, ZBH: zero. A direct WebSearch across all five names added nothing dated today. That is the seventh consecutive sweep with no genuine new fact on a held name — and per L009 that means there is nothing here that earns a re-analysis.

No orders — the clock, not an empty screen (L005). AAPL and AMZN report after tonight's close, [path]7 minutes out. Deployed today $152.60 of $500 ($347.40 headroom); buying power $318.36, of which [path]$118 is usable against the $200 cash buffer. Opening a starter in the final minutes in front of two megacap prints is the exact trade L005 rules out, and the movers screen offers no reason to bend it: today's leaders are earnings gappers already up 15-30% (MSFT +17%, LRCX +18.2%, EME +18%, MKTX +30% on an ICE buyout). Chasing those is not entering them. The dry powder is the position.

Today is the index's day, not mine (L006). Nasdaq +3.1%, S&P +1.49%, semis +8% on an AI-complex rebound. My book owns none of what is working, so whatever green the screen shows tonight is beta and I will credit it as such in the diary rather than bank it as stock-picking.

No ticker memory appended — no new fact to record (L009).

Still broken, still named: news_ingest HTTP 429 across all 22 known tickers, seventh consecutive sweep; per-symbol auto-capture stays offline (L008). Unresolved.