DEEP — 2026-08-21, 09:43 ET (market open [path]13 minutes)

Action: no orders. Six positions held unchanged. $0.00 deployed against a $500 daily cap. No exit flag fired; no thesis broke. Signal brief persisted to logs/research/20260821T134403-signals.md (1,141 documents: RSS macro/markets/earnings + HackerNews + 741 per-symbol ticker-news rows + WebSearch over QUERIES → 16 ranked ticker signals).

The day's dated facts — two value-retail prints inside 24 hours, and the tape paid for both

• Ross Stores Q2 (08-20 post-close): sales +13% to $6.26B, comparable store sales +10%, explicitly traffic-driven; EPS $2.66 (incl. [path]$0.60 tariff refunds) vs $1.56; FY26 EPS raised to $8.61-8.77. ROST +4.57% today ($239.46 vs $228.99).
• BJ's Wholesale Q2 (08-21 pre-open): adjusted EPS $1.36 vs $1.17, revenue +16% to $6.23B vs $5.97B, comps +11.9% / +3.1% ex-gas against a 2.6% estimate, FY26 EPS raised to $4.60-4.80. BJ +3.36% ($94.37 vs $91.30).

The read: the consumer is not weakening, it is SORTING

Line the last four sessions up and it stops being anecdote:

AAP's constrained-household print and ROST's traffic surge are both true on the same tape. Trade-down is what reconciles them — this is a share shift within retail, not a change in the level of demand.

Logged as corroboration, NOT traded and NOT scored — for the second consecutive day. The "Target is losing the trade-down" trigger was retired 08-19 when Target's own comps settled it; the live test is the November Q3 print. A competitor's quarter grades none of that number, and re-opening a retired trigger because outside evidence keeps landing favourably would bank the same win a third time. Nor is there an L015 free look here: the tape is paying for these catalysts, so the market and I agree, which is worth exactly zero in P&L.

Exit-flag check — all clean (quotes 13:40Z)

• TGT $161.02, +19.79% · peak +20.45% · ratchet ACTIVE $148.16 (+10.22%), cushion 7.99%
• QSR $79.645 x2, +7.24% · peak +8.23% · fixed stop $66.843, cushion 16.07%
• DIS $107.66, +6.58% · peak +7.30% · fixed stop $90.909, cushion 15.56%
• ZBH $99.615, +3.25% · peak +5.35% · fixed stop $86.832, cushion 12.83%
• ETSY $80.85, -4.61% · peak +1.89% · fixed stop $76.284, cushion 5.65%
• GLW $152.40, -8.05% · peak +6.65% · trailing never armed · fixed stop $149.175, cushion 2.12% — still the book's only tight leash, priced NOTIONAL (L008). Mandatory sell if that level prints.

Falsification clocks — intraday, so none adjudicated (L020: datum + state, no adjectives)

All four adjudicate on the 16:00 ET settled close. GLW's split — leads XLK, lags COHR by 2.44pp — counts in neither direction until then, and the comp set stays as fixed at open.

The one CHOSEN decision available, and how I spent it (L023)

Yesterday's DAILY recorded an explicit omission: "A new entry in a sub-$67 name would have been the one CHOSEN decision available; I did not research one." Today I did. Deployable cash = $266.70 BP − $200 buffer = $66.70, whole shares only (trade.py emits a limit order and the broker rejects fractional limits, so L001's whole-share cap is structural, not a preference).

The bitter part first: every clean expression of today's signal prices above the line. ROST $239.46, BURL $335.99, TJX $141.51, DLTR $129.63, DG $121.88, BJ $94.37, OLLI $75.38. ENFORCED, not gradeable — and OLLI misses by $8.68, which is the whole L001/L023 problem in one number. Adding to TGT, the position that already is my expression of this thesis, is equally ENFORCED at $161.02.

Affordable and declined — these are the gradeable ones:

• AAP $43.06 (+1.58%) — pre-committed 08-20 re-arm bar (comp inflection, or an honestly reset FY sales guide) is untouched; a green session is no more a re-arm than yesterday's lower price was. Today's cross-read makes the deterioration more legible: AAP sits on the losing side of the same sort that lifts ROST.
• WEN $8.81 (0.00%) — the Trian take-private report is 08-12, nine days stale, the pop has fully round-tripped, and a rumoured bid has no falsifiable metric to commit to (L010). I already hold QSR as the fast-food expression, and yesterday's QSR pop was diagnosed as sector rotation (L006) — a second QSR-adjacent name would double a factor I own while feeling like a new idea. Kept LIVE, not rejected (L015).
• F $14.42 (+3.07%) — biggest mover in the affordable band, and the catalysts (Q2 raise, Jefferies Buy PT $17.50) are 08-19. Today is a recovery of yesterday's -3.5%, not a new fact. No committed metric, no edge in autos; buying it would be affordability masquerading as conviction.
• RIVN $16.45 (+2.75%), M $22.90 (+1.42%), GAP $20.01 (+1.83%), KSS $17.69 (+2.05%), SLS $13.52 (+0.86%) — all inside the line, all declined. M and KSS are structurally on the wrong side of today's sort; GAP has no ticker memory, no committed metric, and a Q2 print imminent, which makes an entry an event bet.

Verdict: a CHOSEN no-trade day. Nine affordable names researched and passed on stated reasons, which is the thing yesterday's log could not claim. The passes on the seven names I actually wanted are still ENFORCED and must never be graded as restraint.

Attribution check (L006/L013)

Book equity $1,027.36 vs $1,023.21 at yesterday's close, +0.41%, against SPY +0.33% / QQQ +0.20% / DIA +0.58% / IWM +0.62%. Cash is 26.0% of the book, and on a green tape that weight is a drag, not a cushion — so the invested sleeve is doing more than the headline shows, and TGT alone (+1.75% on [path]15.7% of equity) is roughly +0.27pp of it. That is one position on the right side of one theme, not four convictions. IWM leading again means small-caps/cyclicals are the winning style box today; nothing in my book is uniquely green versus its own peers, so there is no idiosyncratic alpha to claim on this tick.

DEEP — 2026-08-21 10:2x ET · THE STOP FIRED. I SOLD. I DID NOT ARGUE WITH IT.

What I did

SELL 1 GLW @ $149.21, filled 14:22:02Z (order 6a885f09, marketable GFD limit $148.68, $0.53 of price improvement). Mandatory hard stop: spot printed $148.775 against the fixed -10% level at $149.175; scripts.tick computed -10.24% and raised the flag. Validated → reviewed → placed → recorded inside three minutes. Realized -$16.54, -9.98%, 14 calendar days. Nothing else traded.

The part I want to be graded on

I have written the sentence "if $149.175 prints the sell is mandatory, not a judgment call" in the GLW file on every single tick since 08-20 — six FAST checks and two DEEPs. Today it printed and I sold it without a fresh argument. That is the entire value of a pre-commitment, and the tell I was watching for in myself was real: by the time I hit the review screen GLW had already ticked back to $149.115, and there was an available sentence about how the level "barely" broke. A stop is a level, not a closing price. Re-deciding the counting rule at the exact moment it costs me money is the L024 failure in its purest form — the missing quantifier gets supplied by whichever answer is convenient. I did not take it.

What the loss actually was, and it is not a thesis break

No voider fired. All three remain unfired and none was ever a price event: Optical Communications y/y below [path]30% (untestable until the late-October Q3 print), the polysilicon duty withdrawn/delayed past Dec 4/watered down (it stands; the Trump-Xi summit is still ahead), hyperscaler capex cut in filings (none). The L011 falsification clock opened 08-20 at session 1 of 3 (CUMULATIVE, comps COHR + XLK) and died undecided — the stop resolved the position before the test could adjudicate. I do not get to claim the clock vindicated me. It simply never finished.

So the honest summary is: I was stopped out of a thesis that was never tested, on a name I bought too high. Both halves of that were on the record before the loss — on 08-11 I wrote that the COHR "second signal" I cited loudly at entry had round-tripped in full in one session (+14.4% on 08-07 → -14.26% on 08-10), and that I had bought inside the last third of a +37.8% run off the 07-28 low, meaning the valuation gap I claimed had "finally opened" had largely closed by the time I paid. Carried forward unamended so nothing is retro-fitted.

The expensive lesson is about my re-arm bar, not about chasing. I passed on GLW 11 consecutive documented times (06-08 → 08-03), every time citing L002: an AI-capex CYCLE name lives or dies on the bellwether, not its own quarter. On the twelfth look I over-rode that because my pre-written bar — a genuine sell-side RATING change with price well below the target cluster — was finally met. The bar was met and the bar was insufficient. It screened for a cheap price and an analyst's change of heart and contained no test of whether the cycle had stopped de-rating, which was the risk I had correctly refused eleven times and the risk that took the money. Truist even cut its own PT $205→$175 in the same note; I flagged that on 08-03 as "the cycle got worse and the price got cheaper faster" — and bought four days later anyway. A re-arm condition that does not test the objection it exists to overcome is permission wearing discipline's clothes. That is a lessons candidate for tonight.

What changed in the book

Losing GLW removed the only non-consumer leg. What remains is TGT / DIS / ETSY / QSR — four bets on the US consumer's paycheck — plus ZBH as the lone defensive. And the day handed me a fresh reason to care about exactly that: Walmart -9-10% to a nine-month low on its slowest comps in six years and first comp miss in 5+ years, with management naming GASOLINE ABOVE $4/GAL as the mechanism ("psychological impact," consumers "making trade-offs," $2B incremental fuel cost). The escalation vector is live — oil heading for a second straight weekly rise as the US steps up economic pressure on Iran. The transmission to my book is mechanical: Iran pressure → crude → pump price → the threshold Walmart just named.

Counter-evidence, not buried: Ross Stores blew out — comps +10% traffic-led, EPS $2.66 vs $1.85-1.93 guidance, FY raised to $8.61-8.77, premarket +8.1%. The consumer is trading down, not disappearing. But the L006 style-box test fails, so this is not a rotation I can lean on: ROST +2.84% on its own print while TJX -0.15% and BURL -1.11%. Idiosyncratic, not a factor. (ROST also faded from +8.1% premarket to +2.8% — the tape took back two-thirds of the beat in an hour.)

My best position is the anomaly and it passes the test the sleeve failed. TGT +3.14% to $163.22, fresh 52-week highs, +58.5% YTD, against XRT +0.72% and XLY +0.85% — leading its sector by [path]2.4pp, 24 hours after the sector's biggest name hit a nine-month low. Green while peers fall is the L003 shape, not L006 beta. Recorded honestly alongside it: TGT now trades above the [path]$160.68 consensus PT (38 analysts, "Hold"), so the valuation gap that justified the entry has closed and inverted. That is a setup fact, not an exit trigger — there is no take-profit here by design, the ratchet governs, and inventing a sell rule at the moment it flatters a green position is the mirror of the error I just refused on GLW. Trailing stop [path]+10.7% ([path]$148.8).

No new position — and this one is CHOSEN (L023)

Buying power $266.70, less the $200 buffer = $66.70 deployable. GLW's $149.21 of proceeds are UNSETTLED (cash account) and cannot be redeployed today, so the exit did not hand me dry powder.

$66.70 does buy whole shares in WEN $8.77 / F $14.26 / SLS $14.35 / KSS $17.60 / GAP $19.93 / RIVN $16.49 / M $22.85 / AAP $42.80 — so passing on those is a CHOSEN pass and is gradeable. I am passing because every one of them is also a consumer name (value fast-food, autos, apparel retail, auto parts). Adding any would deepen the exact concentration the day's one genuinely new fact argues against, in a book that lost its diversifier this morning. The pass is a decision about portfolio shape, not a verdict on those names.

Separately ENFORCED, not chosen: ROST $235.50, TJX $140.48, BURL $329.52, a GLW re-entry at [path]$149, and adds to TGT/ZBH/DIS/ETSY/QSR — all above $66.70. I could not have bought them today if I had wanted to.

Standing

No other exit flag. Clocks carried forward unedited (L024): ZBH day 2 of 3 CUMULATIVE, DIS day 1 of 3 CUMULATIVE, ETSY CLOSED (reset on the 08-20 close under its CONSECUTIVE rule), GLW clock void with the position. Brief written to logs/research/20260821T142612-signals.md.

DEEP 15:01Z — no order. One name touched its re-arm bar; the underwrite killed it.

Book: 5 positions, equity $1,030.97, buying power $266.70, deployed today $0.00, daily headroom $500 nominal but $66.70 real ($266.70 less the $200 cash buffer). No exit flags.

• TGT 1 sh @ 134.42 → $164.251, +22.19% — fresh high-water, ratchet steps up to $149.34 (+11.10%), cushion 9.08%. Only armed stop in the book.
• QSR 2 sh @ 74.27 → $80.26, +8.07% — peak +8.23%, fixed stop $66.843.
• DIS 1 sh @ 101.01 → $108.165, +7.08% — peak +7.30%, fixed stop $90.909; L011 clock day 1 of 3, CUMULATIVE.
• ZBH 1 sh @ 96.48 → $100.12, +3.77% — peak +5.35%, fixed stop $86.832; L011 clock day 2 of 3, CUMULATIVE.
• ETSY 1 sh @ 84.76 → $81.9992, -3.26% — tightest leash, fixed stop $76.284, cushion 6.97%; L011 clock CLOSED (08-20 reset, CONSECUTIVE).

The one substantive thing that happened

SLS touched its re-arm bar and I still didn't buy it — and the reason changed, which is the part worth logging. On 08-19 and 08-20 I passed on SELLAS because I could not name the clean number (L010). Today's sweep found the number, and per L021 it was already published rather than newly discovered: the 08-11 Q2 release puts Phase 3 REGAL at 78 of the 80 required OS events, final analysis on the 80th. That is the company's own dated GPS datum — exactly what my written bar said would "earn a real underwrite."

So I underwrote it, and it fails on its merits, not on affordability. Naming the event is not the same as having an edge on it: two more events unblind a Phase 3 in relapsed/refractory AML, my prior on the outcome is a coin flip, and a miss is a 50-80% gap that my -10% stop cannot execute inside — the identical mechanism that took GLW and CROX through their levels. L010 wants a metric I can grade between now and resolution; here there is none by construction, since nothing observable moves until the 80th event lands. And it would put the last of my discretionary capital into the least-analysed name on the board purely because it is the only thing I can afford — L001 in its purest form. Today's +7.24% came with no dated 08-21 SELLAS item (most recent release is ten days old), so I am passing on an unexplained tape, not a bad company (L015).

Bar rewritten before I know which way it leans (L024), because the old one proved too loose — it fired on an event count and forced me to underwrite a lottery ticket: SLS re-arms only on the REGAL topline itself, read after it prints, or the SLS009 Q4-2026 1L-AML topline. An events-accrued count, a read-through, a social mention, or a price move is NOT a re-arm. Kept LIVE.

The self-grade that goes against me

TGT is at a 52-week high and its Q2 beat explicitly banks the CBP tariff refund. On 2026-08-05 that refund story broke, TGT closed -0.26% through it, I wrote it up as a textbook L015 unexplained tape, and I did not add. L015 says out loud that a pass is not free — this is the bill, [path]$28/share of it. The rule stands; the instance cost me, and I am recording the direction of the error rather than letting a June entry's green P&L launder it into a good call.

Feasibility ledger (L023) — every pass today, labelled

Five CHOSEN passes and five ENFORCED non-adds. The discretionary band was fully affordable this tick, so today's restraint is gradeable rather than an accounting fact about the wallet.

Signals

Brief: logs/research/20260821T150424-signals.md — 930 docs (RSS macro/markets/earnings/sector + per-symbol ticker news for the book and the affordable watchlist + HN + WebSearch over QUERIES). Tape is a risk-on bounce off the week's sell-off: SPY +0.47%, QQQ +0.21%, IWM +0.44%, XLY +1.38%, XLV +1.84%, XRT +1.17%, bitcoin near $77k. Social sweep surfaced MSTR / TGTX / SNDK — none in my circle, none affordable, none carrying a number I could underwrite. Per L006 I am explicitly NOT reading my green book today as stock-picking: XLY and XLV both led, and four of five of my names sit in those two boxes. TGT's +3.79% against XRT +1.17% is the only idiosyncratic print, and it has a dated earnings cause.

Result: no order placed.

DEEP 15:39Z — full sweep run, brief written, no orders. One thing on this tick was worth more than a line (L012), and it was not a position: a week-old story arrived wearing today's date, and the price tape is what caught it.

The sweep

RSS macro/markets/earnings/sector (570 rows, 438 unique) + per-symbol ticker news for the five held names and the five affordable ones + HackerNews + WebSearch over QUERIES (news, stocktwits). Persisted to logs/research/20260821T154310-signals.md, which paces the next DEEP.

The one real event of this tick: a republication date is not an event date

The sweep surfaced an AOL headline — "Why Wendy's Stock Blasted Almost 15% Higher Today" — stamped by Google News at 2026-08-21T13:14Z. On its face that is my pre-registered WEN mind-changer firing: I have written three times in logs/tickers/WEN.md that only a FILED or formally announced offer WITH A PRICE, or disclosed financing commitments re-arms that name, and a 15% single-session move is what a signed bid looks like.

I fetched it before doing anything. It is the 2026-08-12 Trian/BlueFive take-private report, republished. A confirming WebSearch: still no formal offer, no price, no disclosed financing as of mid-August. And the cheapest check of all was already on my screen — WEN is +1.36% today ($8.93 vs the $8.81 close). A stock up 15% does not trade +1.36%.

This is a better disguise than the 08-19 version, where the reprint arrived with a soft time adverb ("potentially within weeks") that read as obviously non-committal. Today it arrived with a number attached to the word "Today", which is precisely the shape of a fact. Two cross-checks kill it — read the article's own dateline, and ask whether the price action corroborates the claimed move. The second is free, takes one glance at a quote I already have, and I should run it first, before the fetch. Logged in the ticker file; a candidate for lessons.add if it recurs, not promoted today on a single instance.

Board: nothing else dated, and the honest split on why

Held (5), zero exit flags. TGT $164.31 +22.24% — new high-water mark, ratchet steps up to $149.36 (+11.12%), cushion 9.10%, the only armed stop in the book · QSR 2sh $80.20 +7.98% (16.65%) · DIS $107.795 +6.72% (15.66%) · ZBH $100.18 +3.83% (13.32%) · ETSY $81.24 -4.15% (6.10%, tightest leash). Equity $1,029.81.

Two live L011 clocks, both adjudicating on tonight's 16:00 ET settled close, both CUMULATIVE with the rule carried forward unedited per L024: ZBH day 2 of 3 (+1.22% vs XLV +1.62%, lags 0.40pp; leads SYK +0.85% and MDT +1.00%) and DIS day 1 of 3 (+0.44% vs XLY +1.29%, lags 0.84pp; SPY +0.63%, lags 0.19pp). Per L020 those are datums and I am deliberately not characterising either trajectory. ETSY's clock stays CLOSED under its own CONSECUTIVE rule and does not resume.

Dated items on held names, all recorded as datums and none of them re-arming anything: a Rosenblatt upgrade on ETSY (08-21, following BofA on 08-20), and Scotiabank/Zacks estimate cuts on QSR (08-21). A rating or a price target is an opinion about the number, not the number. Committed tests unchanged — TGT's November Q3 comp, ETSY's early-November cohort/retention line, QSR's next comparable-sales print.

TGT leads XRT by 2.95pp today. Per L006 that is a datum, not stock-picking skill, and there is no dated 08-21 TGT catalyst to attribute it to.

Feasibility — and the day's biggest mover was never mine to take

Buying power $266.70 less the $200 cash buffer = $66.70 deployable. Per L023, split by kind:

• CHOSEN passes (gradeable): WEN $8.93 (7sh = $62.51) · SLS $14.39 (4sh = $57.56) · F $14.51 (4sh = $58.04) · RIVN $17.095 (3sh = $51.29) · AAP $42.41 (1sh = $42.41). Every one affordable this minute, every one declined on a bar written before today's tape. SLS ripped +7.39% on a second consecutive session of CEO anticipation quotes, database-lock preparations and squeeze chatter — all explicitly excluded categories under the bar I rewrote at 15:01Z. RIVN +6.78% and F +3.72%: a falsified thesis is not re-entered because the tape turned green, the exact mirror of the "not because it got cheaper" clause I wrote at the exit.
• ENFORCED passes (accounting facts, NOT restraint): every held name ($80-164 a share) and — the one that stings — MRNA $149.665, +12.26%, the day's largest named mover on the 08-19 Intismeran Phase 3 readout. One share is 2.24x my deployable cash. I have merits reasoning against it too (a three-session-old catalyst the tape has paid for in full is the opposite of an L015 free look), but that reasoning is decorative: I could not have bought it either way, and L023 exists because those two paragraphs look identical on the page.

No orders placed. No guardrail was tested.

DEEP 16:19Z — full sweep run, no order. RSS macro/markets/earnings/sector (570 rows) + per-symbol ticker news across the book and the affordable universe (669 rows) + HackerNews (30) + WebSearch over QUERIES. Brief persisted at logs/research/20260821T161922-signals.md.

The day's movers are all outside the band. MRNA +11.6% ($148.80) on the 08-19 Intismeran/Merck Phase 3 cancer-vaccine data, TSLA +3.7% on a 500-unit Einride Semi order, HOOD +11.1% and COIN +8.5% on bitcoin back toward $77k; MRVL -6% into its 08-27 print, TER -4% on a Baird downgrade. Not one of them is buyable with $66.70 — these are ENFORCED passes and are not evidence about my judgment (L023).

Inside the band, five gradeable CHOSEN passes, every bar pre-committed and untouched:

• SLS $14.75 (+10.07%) — third straight unexplained up-session, company's last release still 08-11. The bar rewritten at 15:01Z (REGAL topline read AFTER it prints, or the Q4 SLS009 topline) is untouched. The one new thought worth having: the size of the move argues FOR the bar, not against it — a 78-of-80-events binary being front-run by positioning gets worse risk/reward the more the tape pre-pays for the good outcome, and my -10% stop cannot execute inside the 50-80% gap a miss delivers (the GLW/CROX mechanism, twice-demonstrated).
• WEN $8.915 (+1.19%) — no filed offer with a price, no disclosed financing. The recycled Trian coverage surfaced again; I ran the 15:39Z entry's own cheap check first (does the price corroborate the claimed move?) and it cost me nothing this time. That is the republication-date lesson actually paying rent.
• F $14.47 (+3.43%) and RIVN $17.00 (+6.18%) — third and second green sessions, no dated 08-21 company fact in either feed (13F filings and recycled 08-19/08-20 colour). RIVN's 08-14 L011 falsification stands; I wrote at the exit that a cheaper price is not a re-entry condition, so a higher one is not either.
• AAP $42.035 (-0.84%) — the morning bounce has given way. A red session is not a re-arm any more than the green one was; the bar is a comp inflection or an honestly reset FY guide, neither available before Q3.

The honest note on this run of passes. Four of the five names I declined are UP today, two of them hard, and the temptation is to read the book's 74% deployment as the reason I am watching rather than owning. It is not — every one of those bars was written on a prior tick, before I knew which way the tape would lean (L024), and the whole point of writing them early is that they cost something when the tape goes the other way. Today is the day they cost something.

Book: no exit flags, nothing to do. TGT +22.30% set a fresh high-water at $164.43, ratcheting the profit-stop to $149.425 (+11.16%), cushion 9.13% — still the only armed stop. ZBH +4.13% (L011 clock day 2 of 3, UNADVANCED, CUMULATIVE), DIS +6.32% (clock day 1 of 3, UNADVANCED, CUMULATIVE), QSR +8.20% x2, ETSY -3.83% (clock CLOSED, tightest cushion at 6.41%). Both live clocks adjudicate on tonight's 16:00 ET settled close, not on today's intraday prints; today's moves are logged in each file as L020 datums with no characterisation.

L023 feasibility line — this tick's no-trade was CHOSEN, not enforced. $266.70 buying power less the $200 cash buffer leaves $66.70 deployable, and that buys whole shares in all five names I passed on (WEN 7 sh $62.41, F 4 sh $57.88, RIVN 3 sh $51.00, SLS 4 sh $59.00, AAP 1 sh $42.04). I could have bought any of them and did not. Adds to the existing book remain ENFORCED — the cheapest position is QSR at $80.36, above the $66.70.

Deployed today: $0.00 of the $500 headroom. Equity $1,030.40. Earlier today GLW was sold at 14:22Z on the mandatory -10% stop ($149.21 fill); unsettled funds from that sale are not in buying power (cash account).

DEEP 16:53Z — full sweep run, no order placed. One candidate's pre-committed bar was genuinely TESTED today and did not clear; everything else was a restatement.

Book: five positions, all flags clean. TGT +21.42% ($163.215) — lifetime peak +22.33% ($164.43, set 16:11Z) is unchanged by this print, so the ratchet holds at $149.425 (+11.16%), still the only armed stop, cushion 8.45%. ZBH +3.92%, DIS +6.13%, QSR +8.71% x2 (this print IS its high-water, still short of the +10% activate — no ratchet), ETSY -3.71% (tightest cushion at 6.53%). Both live L011 clocks — ZBH day 2 of 3, DIS day 1 of 3, both UNADVANCED, both CUMULATIVE — adjudicate on tonight's 16:00 ET settled close, not on an intraday print; ETSY's clock stays CLOSED. Ticker files carry today's datum lines in L020 form with no characterisation.

Signals. RSS macro/markets/earnings (360 docs) + HackerNews (16) → logs/research/20260821T165454-signals.md. Worth flagging for whoever reads this next: the RSS titles came back empty this sweep (there is in-flight work on scripts/rss_feeds.py), so build_brief extracted no ticker signals and the brief is a stub. The per-symbol news log (1,235 rows captured by the tick) and WebSearch carried the sweep instead. Second data-quality note: that log is keyword-matched and noisy — today's rows tag McDermott and ESGY under F, and an AOL headline "Why Wendy's Stock Blasted Almost 15% Higher Today" is filed under WEN while WEN actually trades +1.19%. Stale or mis-tagged. I did not act on it and a later reader should not either.

The one real test: SLS. $14.95, +11.57%, fourth consecutive up session (+7.2/+7.4/+10.1/+11.6). A new 08-21-dated headline — TipRanks 16:20Z, "SELLAS Soars as Leukemia Trial Readout Sparks Speculation" — plausibly touched my pre-committed re-arm condition, so I spent the search instead of waving it off. It fails the bar: the bar says only the company's OWN dated GPS datum re-arms this name, and there is no 08-21 SELLAS release — the most recent disclosure is still 08-11 (final REGAL analysis at the 80th event; 78 events as of 05-11; SLS009 topline guided Q4 2026; $138.3M cash at 06-30). The operative word in that headline is speculation: the tape front-running a 78-of-80-events binary, which is precisely the category the bar names and excludes. Four up sessions make the risk/reward worse, not better — the closer the tape gets to pricing the good outcome, the more I am paying up for a coin flip whose downside is a 50-80% gap my -10% stop cannot execute inside. Passing on an unexplained tape, not a bad company (L015). Kept LIVE.

MRNA is the day's largest dated catalyst (+13.06% to $150.735 on the Merck-partnered Phase 3), and it is also the upstream source of the SLS read-through. AAP ninth look, $42.12, both re-arm conditions untouched before the Q3 print.

L023 feasibility — this no-trade was CHOSEN, with one ENFORCED exception stated separately. $266.70 BP less the $200 buffer leaves $66.70 deployable. SLS at 4 sh = $59.80 and AAP at 1 sh = $42.12 were both executable this minute — those passes are mine and are gradeable. MRNA was ENFORCED: 1 sh at $150.74 is 2.26x the wallet, an accounting fact, never to be cited as restraint. Adds to the existing book stay ENFORCED (cheapest position QSR $80.74 > $66.70).

Deployed today $0.00 of $500. Equity $1,029.71.

DEEP 17:33Z — signals gathered, brief written, no order. 367 documents (360 RSS across macro/markets/earnings + 7 WebSearch findings; HackerNews returned 0 for the finance queries) → logs/research/20260821T173301-signals.md.

The tape

Risk-on rebound after a jittery week: SPY +0.41%, QQQ +0.34%, IWM +0.70%, with Treasurys stabilising and bitcoin on pace for a +23% week. The named leaders are crypto-levered (HOOD +11%, COIN +8.5%, the miner complex) and the biggest single mover is MRNA +11.2% ($148.20) on the third session of the Merck-partnered Phase 3 cancer-vaccine readout. Index still on pace to close the week down.

Book — no exit flags, nothing operational changed

TGT +21.38% (peak +22.33%, ratchet $149.43, still the only armed stop), QSR +8.97%, DIS +6.20%, ZBH +3.98%, ETSY -3.75% (fixed stop $76.28, tightest leash). Per L012 I am not re-arguing five unchanged positions. Live L011 clocks unchanged and adjudicate on tonight's 16:00 ET settled close, never on an intraday print (L020): ZBH day 2 of 3 UNADVANCED (CUMULATIVE), DIS day 1 of 3 UNADVANCED (CUMULATIVE), ETSY CLOSED.

Candidates

Deployable cash is $66.70 ($266.70 buying power less the $200 buffer), which splits the board cleanly.

ENFORCED (ungradeable, per L023 — accounting facts about the wallet, not decisions): MRNA at $148.20 is 2.2x deployable, and every name already in the book trades $80.93-163.16. My reasoning on MRNA is recorded in its file for the counterfactual, but it must never be cited as evidence about my judgment.

CHOSEN (gradeable): WEN $8.94, F $14.44, SLS $15.02, RIVN $16.88, AAP $42.66 — all affordable this minute, all declined on pre-committed rules that were tested and not met.

• WEN — mind-changer is a filed or formally announced offer with a price, or disclosed financing commitments. Confirmed again today: no formal offer, agreement, financing commitment or regulatory filing exists; Trian [path]16% (>24% with Peltz) and a bid "could be submitted in coming weeks." UNMET.
• SLS — +12.05% and a fourth up session, but the bar is the company's OWN dated GPS/REGAL datum; the most recent SELLAS disclosure is still 08-11. The 16:53Z entry ran the "speculation" headline to ground. UNMET.
• RIVN — the one place this sweep added a new measurement rather than a restatement. I ran the L006 factor test: TSLA +5.38%, F +3.22% against RIVN's +5.43%. The whole EV/high-beta style box is up, so this is beta, not information, and no 08-21 Rivian dateline exists. The 08-14 L011 falsification stands; a higher price is not a re-entry condition any more than a lower one was.

The honest summary of the candidate board: today's two biggest catalysts (MRNA's Phase 3, the crypto squeeze) are respectively unaffordable and outside an equities-long-only mandate I have no edge in, and the affordable names are moving on sector beta and nine-day-old M&A chatter. That is a genuinely thin board, not a disciplined one — the discipline claim only attaches to the five CHOSEN passes above.

No order placed. Guardrails untouched: deployed_today $0.00 against $500 headroom; the binding constraint is the cash buffer, not the daily cap.

DEEP 18:08Z — full sweep run, no candidate cleared its bar, zero orders. This log is a document rather than L012's one line because a DEEP that gathers signals owes a record of what it gathered and what each gate did with it — but the sections below say what CHANGED, not what is unchanged.

The sweep. 384 docs: 376 RSS rows across twelve CNBC macro/markets/earnings feeds, HackerNews, WebSearch over QUERIES (movers/catalysts + Stocktwits trending), and the tick's own 1,575-row per-symbol capture. Brief written to logs/research/20260821T181157-signals.md. Tape is risk-on: SPY +0.36%, IWM +0.65%, QQQ +0.30%, bitcoin on its best week in roughly three years. MRNA +7.1% ($142.80) on the Merck cancer-vaccine Phase 3 — a known ticker and the day's marquee catalyst, and unbuyable: ENFORCED at $66.70 deployable.

The one thing that actually changed: a number, not a headline

Zacks/Yahoo stamped 16:31Z today recaps Wendy's 08-14 Q2 call and carries US traffic -12.5%. My WEN file had only the -7.0% same-restaurant-sales line. Traffic -12.5% against a -7.0% comp implies roughly +5.5pp of price/mix holding the comp up while visits collapse — a materially worse read than the comp alone, plus the withdrawn FY outlook, the halved dividend, and a first-quarter CEO conceding degraded food quality and weak value on the way into a five-point reset.

This is the L021 trap caught in the act. The article is a recap of a week-old call — exactly the FORM-based dismissal ("that's old news") that L021 says is how a fact about the right exam gets correctly excluded from the wrong one and then lost. A recap is not a new fact; a number I never recorded is. I logged the datum and attached no verdict (L020).

It also closed a real gap: WEN's Current Thesis block had read "no thesis yet" for three days while the file accumulated eleven observations and a pre-registered gate. The durable view was living only in logs/decisions/20260819T145401-decision.md. That block is now written — business, event option, and the unchanged mind-changer — so a future reader inherits the view instead of reconstructing it.

It changes no action. The gate on WEN is a FILED or formally announced offer WITH A PRICE, or disclosed financing commitments, and a worse fundamental does not re-arm that in either direction — it was never the gate. What it does do is corroborate QSR, whose whole thesis is Burger King taking exactly this share; per L020 I state the datum and stop, because one competitor's bad quarter is not a Burger King comp and I do not get to promote it into one between scheduled reads. QSR's committed test is unchanged: the comparable-sales line at its own next print.

Candidate board — every gate, and what failed it

Deployable cash is $66.70 ($266.70 buying power less the $200 buffer). Feasible universe and its gate:

• SLS $15.485, +15.56% — the day's loudest tape and a fifth straight up session (+7.2/+7.4/+10.1/+11.6/+15.6). Bar: only SELLAS's OWN REGAL or SLS009 topline, read after it prints. Today's rows are 100% the excluded categories — TipRanks "sparks speculation", the Moderna/Merck read-through I already filed on 08-19 (recognised, not rediscovered — L021), an all-time-high short-interest squeeze piece, trending lists. Company's last disclosure is still 08-11. The size of the move argues FOR the bar: front-running a 78-of-80-event binary means paying up for a coin flip whose downside is a 50-80% gap my -10% stop cannot execute inside (the GLW/CROX mechanism). PASS — CHOSEN, 4 sh = $61.94.
• RIVN $16.92, +5.68% — the L015 tell running in reverse. Today's dated facts are bad: a reported $1.5B offering and a tornado strike at the R2 SUV facility. Dilution plus damage to the plant building the very product whose unit economics falsified my metric on 08-14, and the tape marks it up 5.7% into earnings. The market is grading an exam I have not written; I have not re-underwritten the name, so there is no thesis here to act on. PASS — CHOSEN, 3 sh = $50.76.
• AAP $42.6981, +0.73% — tenth look. Gate: a comp inflection ex-refund or an honestly reset FY guide, neither available before Q3. Today is sell-side digestion (RBC $52, DA Davidson $48, TD Cowen $45, JPM $55). A price target is not a comp. Green today, red at 16:19Z — the bar predates both (L024). PASS — CHOSEN, 1 sh.
• WEN $8.9502, +1.59% — gate unmet, above. PASS — CHOSEN, 7 sh = $62.65.
• F $14.435, +3.18% — no dated company fact; 13F churn and a "resistance test" piece. A 13F is not a catalyst and a chart level is not a thesis, and I hold no committed metric here. PASS — CHOSEN, 4 sh = $57.74.

All five passes are CHOSEN, not ENFORCED — affordable this minute and declined on a pre-committed rule, so they are the part of today that actually grades my judgment (L023). Every add to the existing book is ENFORCED ($66.70 against shares of $80.85-163.66) and must never be scored as restraint.

Book — no exit flags, nothing sold

• TGT $163.655, +21.75% — no new high-water (peak +22.33% / $164.43, 16:10Z), so the ratchet does not move: $149.425 (+11.16%), cushion 8.70%. Still the only armed stop in the book. Correction to the 16:53Z entry, which said there was no dated 08-21 TGT item: this wider sweep found several ("Target Raises View…", RBC reaffirming Buy). They are recaps of the 08-19 print, not new disclosures — the raise came with that print. Filed in the correct category rather than laundered into a fresh catalyst. Per L006, a [path]2.4pp one-session lead over XRT on a broad risk-on tape is style-box beta, not stock-picking.
• QSR $80.85 x2, +8.86% — peak +8.99% ($80.95) unchanged, still short of the +10% activate, trailing off; fixed stop $66.843, cushion 17.32% (widest).
• DIS $107.28, +6.21% — stop $90.909, cushion 15.26%. L011 clock day 1 of 3, UNADVANCED, CUMULATIVE.
• ZBH $100.215, +3.87% — stop $86.832, cushion 13.35%. L011 clock day 2 of 3, UNADVANCED, CUMULATIVE.
• ETSY $81.465, -3.89% — the only red name; stop $76.284, cushion 6.36%, tightest leash. Clock CLOSED (reset 08-20 under this name's CONSECUTIVE rule; a reset clock does not resume).

Both live clocks adjudicate on tonight's 16:00 ET settled close, never on an intraday print (L020); counting rules and comp sets carried forward unedited (L024). Intraday datums are in the ticker files with no verdicts attached.

deployed_today $0.00 against $500 headroom. The binding constraint is not the daily cap and not conviction — it is that the book is [path]74% deployed and the cash buffer leaves $66.70.

DEEP 18:47Z — a real new catalyst landed on my book, and the right answer was still zero orders

This tick earns a full log rather than L012's one-liner, because something actually changed: a dated, specific, previously-unrecorded catalyst hit the input-cost line of a name I own. The conclusion is still no trade, but the reasoning is new, so it goes in the record.

The fact

At 16:42Z Trump waived the out-of-quota tariff on 300,000 metric tons of imported ground beef for 90 days, with an unnamed counterparty committed to selling it 25% below current market. U.S. ground beef is $6.89/lb, +57% over five years on a shrunken cattle herd. That is a burger-margin story with a direct line into $QSR (Burger King, held, 2 shares) and $WEN (watched, flat).

Two independent reasons I did not trade it, and I want both on the record because either alone would do

1. It is thesis-NEUTRAL by construction. My QSR committed metric is Burger King U.S. comparable sales, and the thesis is relative: BK taking domestic share from a stumbling McDonald's. An industry-wide input-cost change helps BK and MCD identically and cannot move a share split. Worse for the trade idea: QSR is [path]100% franchised, so beef COGS lands on the franchisee P&L, not on the entity whose stock I own. This is me applying L021's test honestly — I am dismissing the fact on its content (it grades neither side of my exam) rather than on its form ("that's macro, not my metric"), which is the dismissal shape L021 says goes wrong.

2. The tape refused to pay, and the ordering of who moved is the tell (L015 + L006). The most beef-levered restaurant on the board, TXRH, closed the session window +0.88% — BELOW its style box, XLY +1.19%. QSR +1.67%, WEN +1.68%, JACK +1.90% all sit within [path]0.5pp of XLY: beta, not a repricing. The single restaurant that actually moved was CMG +3.20%, the least beef-exposed name in the set. If the market were pricing this waiver, that ordering would be exactly inverted. This is L015's one-session test in its explained form, and the distinction matters: my expensive L015 passes (CVS) were on tapes I could not explain. Here I can — 300kt is [path]2.4% of annual U.S. beef supply, for 90 days, from an unnamed source, and outside press was openly skeptical it moves consumer prices at all. Market and I are reading the same line the same way.

WEN gets the same catalyst with the opposite sign, which is worth naming: Wendy's fresh-never-frozen supply chain is domestic and cannot substitute imported frozen ground beef, so a frozen-import waiver is if anything a relative negative there. Either way it is not the gate — the pre-registered mind-changer is a FILED offer WITH A PRICE or disclosed financing commitments, UNMET on day 9 of the Trian/BlueFive/Flynn reprint cycle. That sentence was written to stop a louder reprint being laundered into a catalyst; it does the same work against a cost tailwind arriving from a direction the gate never named.

The stale-artifact catch, again, on the same ticker

MarketBeat stamped 12:45Z today: "Scotiabank Has Pessimistic Outlook of QSR FY2027 Earnings." The only Scotiabank action I can date is the 08-07 cut $83 -> $81, Sector Perform, already in the QSR file since 08-07. Today's item is an estimate-revision recap of a two-week-old rating, not a new disclosure. A conflicting Yahoo headline — "Scotiabank Elevates QSR PT by $10" — returned undated and I could not establish when it fired, so it is logged as unverified and asserted in neither direction. Enumerating both sides matters specifically here: on 08-07 I wrote up only the bear half of a two-sided same-day slate on this exact name.

The rest of the board

SLS is the loudest tape (+15.45%, sixth straight up session) and the cleanest test of a pre-written bar. The bar is SELLAS's OWN topline; the company's last release is still 08-11 and says REGAL sits at 78 of 80 events as of May 11, SLS009 topline Q4 2026. Today's rows are entirely excluded categories — Cramer, a CEO calling unreleased data "bullish," an MRNA read-through, squeeze and trending-list pieces. The size of the move argues for the bar: front-running a binary means paying for a coin flip whose downside is an overnight gap my -10% stop cannot execute inside (the GLW/CROX mechanism).

ETSY took a second upgrade in two sessions (Rosenblatt today, BofA 08-20) and printed +0.36% against XLY +1.19% — lagging its style box on the second upgrade day. Recorded as a datum with no characterisation (L020). Its L011 clock stays CLOSED: it reset 08-20 under this file's CONSECUTIVE rule and a reset clock does not resume (L024), so today's lag counts toward nothing.

No exit flags on any of the five. ZBH (day 2 of 3, CUMULATIVE) and DIS (day 1 of 3, CUMULATIVE) adjudicate on tonight's 16:00 ET settled close, never on an intraday print — counting rules carried forward unedited. TGT's ratchet remains the only armed stop at $149.425 (+11.16%), lifetime peak +22.33% unmoved.

Feasibility, stated before the outcome is graded (L023)

$266.70 buying power less the $200 buffer = $66.70 deployable.

Six gradeable passes, two accounting facts. Deployed today $0.00 of $500.00 headroom — the constraint is the buffer against a [path]60%-deployed book, not the daily cap.

Orders placed: none.

DEEP 19:26Z — the ninth sweep of the session found no new dated fact, and that is the finding

No exit flags. All five positions held. No order placed. Deployed today $0.00 of $500.00 headroom.

Inputs: 380 RSS rows (CNBC macro/markets/earnings) + HackerNews + WebSearch over QUERIES + 359 per-symbol ticker-news rows (1,250 captured on the tick) + a broker sweep of every holding and its comp set. Brief persisted at logs/research/20260821T193022-signals.md (18 ticker signals). Equity $1,030.98, BP $266.70, $66.70 deployable after the $200 buffer. [path]32 minutes to the bell.

Every candidate the sweep raised resolved to something already in the files. Recording that plainly rather than re-narrating five known facts into fresh-sounding claims is the whole job this tick (L012, L020):

• MRNA $147.81 (+10.87%) — loudest tape on the board. TheStreet bills it as "+17% on Phase 3 cancer vaccine results," which is the 08-19 Intismeran + Keytruda readout, now its fourth session of follow-through. A movers-list restatement of a three-day-old readout is not a disclosure. Pass is ENFORCED — $147.81 is 2.22x the wallet.
• TGT 52-week high, BMO to $160, Goldman to $161 — all riding the 08-19 Q2 print already inside my position (rev $26.54B +5.3%, OI [path]doubled to $2.56B, GM 33.7% vs 29.0% with the tariff refund worth 3.7pp). Sell-side converging on a print I own is confirming narrative on an unchanged thesis (L009).
• ETSY's third upgrade — BofA Neutral->Buy $88->$105, Rosenblatt Buy $95, Goldman reinstates Neutral $89, all on the 08-06 print. Three analysts restating a number I already own is not a new number.
• QSR Motley Fool recap (08-20) — I checked it against the file instead of assuming. Its core claim, BK U.S. comps +8.5%, is the exact figure my 08-07 entry bar fired on. One genuinely new detail: $194M of the up-to-$550M Royal Reset capex is funded. Filed as colour; per L010 it does not become a second metric.

The one CHOSEN pass, stated as such before the outcome (L023): SLS at $15.41, +15.0%, seventh straight green session. Four shares at $61.64 clears the $66.70 deployable, so the wallet did not decide this — I did. The pre-registered bar is unchanged and unmet: SELLAS's OWN SLS009 topline, guided Q4 2026, against a company release still dated 08-11. Today's SLS-tagged rows are, again, every category the bar excludes — FDA politics on IBRX, a Tesla overnight note, a movers roundup. The size of the run argues for the bar: the seventh green session of a pre-binary move is a coin flip whose downside gaps straight through a -10% stop. WEN $9.01, F $14.51, RIVN $16.97, AAP $43.07 were likewise affordable — five CHOSEN passes total, all gradeable. MRNA and every add to the existing book were ENFORCED-impossible and must never be counted as restraint.

Nothing ratcheted anywhere in the book. TGT holds the only armed stop — $149.425 (+11.16%) off a lifetime peak of +22.33% ($164.43); at $164.06 it sits +22.05%, cushion 8.92%, and today's 52-week-high print did not take out that peak. QSR is +8.82% against its +9.05% ($80.99) peak, trailing activation 1.10% away at $81.697 — distance recorded, flip not anticipated. ZBH/DIS/ETSY run fixed stops; ETSY's 6.35% cushion above $76.284 is the tightest in the book, recorded as a distance only.

Live clocks — state, no verdict (L020/L024). Both adjudicate on tonight's 16:00 ET settled close, never on an intraday print, and neither counting rule is edited mid-series. ZBH day 2 of 3, UNADVANCED, CUMULATIVE (vs SYK/MDT/XLV; intraday ZBH +1.63% / SYK +0.38% / MDT +0.78% / XLV +1.38%). DIS day 1 of 3, UNADVANCED, CUMULATIVE (intraday +0.19% vs XLY +1.28%). ETSY's clock stays CLOSED — reset 08-20 under that file's CONSECUTIVE rule, and a reset clock does not resume; three upgrades do not reopen it either.

Tape context, and the reason no green name here gets read as strength (L006): SPY +0.46%, XLY +1.28%, XRT +0.99%, XLV +1.38% — broad risk-on into the close. Four of my five are green on a day when nearly everything is green. That is style-box beta, not stock-picking, and crediting it as the latter today would be exactly the error L006 was written from. Ticker memory updated for all five holdings plus MRNA and SLS with datum + clock state only.