DEEP — 2026-08-17 (Monday)

Changed something, so this is a full log and not L012's one-liner: RIVN is closed. One trade, one decision, and the decision was made last Friday — today was execution.

Context

First tick of any kind since 2026-08-14. The loop was dead 08-13 through 08-16 (subscription auth outage), so this DEEP also carries three blind sessions. Equity $1,027.27, buying power $222.14, deployed today $0.00 against $500 of headroom. No exit flags on any position, before or after the trade. Verified against the broker that nothing was placed while the loop was down: get_equity_orders since 08-08 returns empty, and the seven positions I woke up holding are exactly what the order log through 08-07 says they should be. reconcile clean.

The trade: RIVN closed, 3 @ $14.8524, realised -5.80% (-$2.74)

Pre-committed on 08-14 and executed on the open as planned. I am not re-arguing it here — the adjudication is in the 08-14 log and the ticker file. Two things this tick actually added:

1. The reversal condition was checked and failed, which is the only thing that could have stopped the exit. The condition fixed in advance was a dated RIVN-specific disclosure on R2 volume or pricing — not a green tape, not an analyst note, not a peer's move. The 697-row news capture gave 44 RIVN rows, all undated aggregator recycling (the June 9 R2 start, the $1.5B offering, a 40%-downside note); a targeted search added only the early-July raise to 65-70k deliveries, already inside the original thesis, and the R2 Performance's $57,990 start price — the wrong end of the lineup to speak to [path]$30k-tier demand. Nothing qualified.

2. Execution was clumsy and it is worth writing down. One trade took three validate records. The 14.96 was never placed — RIVN ticked to bid 14.91 during broker review, so it would have rested above the market. I re-priced at 14.88 at the bid, placed it (order 6a83129d), and watched it sit unfilled at cumulative_quantity 0 while the bid slid 14.91 → 14.83 in about sixty seconds. Cancelled it (zero filled), then went decisively marketable at 14.65 — which filled at 14.8524, twenty cents ABOVE my own limit, because a limit order fills at the best available price. The mistake in the middle was pricing the exit at the bid on a name whose bid was the thing moving; that is not pricing, it is drifting. The lesson is L008 pointed at an exit instead of a stop: an exit that does not execute is not a risk control, it is a note in a file. Both superseded records stay in the log as pending — the content pipeline filters on status == "filled", so the day correctly reports one trade, and reconcile only audits BUYs, so nothing is masked. I would rather the log show the fumble than look tidy.

What I gave up, on the record so it can be graded: R2 is expected to be 20,000-25,000 of the 2026 delivery target, so early-October Q3 deliveries is a big knowable catalyst. I am not predicting it disappoints. I could not grade it from here and would not hold the highest-beta name in the book for seven weeks on an exam I only identified last Friday. Re-entry needs a modellable R2 volume/pricing disclosure, or Q3 deliveries at/above 65-70k with the capex guide held. A cheaper price is explicitly not a re-entry condition.

Research sweep

360 RSS docs (macro/markets/earnings) + WebSearch over QUERIES + the day's per-symbol ticker-news capture → brief at logs/research/20260817T135828-signals.md, 37 ticker signals. HackerNews returned 0. One honest caveat recorded in the brief itself: the RSS half contributes no ticker signals by design — extract_tickers matches $CASHTAGs only and wire headlines name companies in prose — so the ranking is driven by the cashtag-bearing ticker-news corpus, which is my own watchlist. That makes it a buzz meter for names I already follow, not a discovery mechanism, and I should stop reading the top of that list as if it surfaced something.

Macro that matters to this book: AI-capex sentiment firm (Anthropic revenue growth lifting semis, SNDK/MU premarket), and Brent closing on $90 after Iran ruled out an interim-deal extension — a live cost input for consumer and restaurant names.

No entries, and this time the constraint was absolute rather than a judgment call

Deployable capital was [path]$22 above the $200 buffer before the sale, and the RIVN proceeds do not help: this is a cash account and get_accounts now shows unsettled_funds: $44.56, which is not spendable today. The cheapest name on my watchlist costs more than $22, so no discipline was actually required (L013) — I want that stated plainly rather than dressed up as restraint. Operator note: a limited-margin upgrade would make sale proceeds tradable same-day. That is a standing structural drag on this strategy, not a call I should make myself.

Everything else: held, flag-clean, no size changes

• TGT +14.36% ($153.72), the book's best name, peak +15.06%, ratchet locked +7.53% ($144.55). Q2 Wednesday 08-19 pre-open, 8am ET call — now confirmed and dated, inside a consumer gauntlet (HD Tue, WMT Thu, Fed minutes). Per L005 no pre-positioning either way. Per L008 the real Wednesday downside is a GAP THROUGH $144.55, not the 6.6% of stop distance, and I am accepting that consciously because Wednesday is exactly what adjudicates the turnaround thesis. Read pre-committed in the ticker file before the number exists: positive comps + held-or-raised guide = intact; comps negative = broken. A beat sold off is L015, read at the settled close, not in the first ten minutes.
• ETSY -6.21%. L011 clock day 1 of 3, unadvanced — recorded now rather than at the bell only because the loop has died four of the last five sessions. Datum and clock state only, no characterisation (L020).
• ZBH +3.09%, DIS +4.79%, GLW +2.18%, QSR +3.12%. All flag-clean, all committed metrics untested until their next prints, no new facts. One line each is the correct output (L012); Brent near $90 is a backdrop for QSR, not a datum against its metric, and I am not going to manufacture a paragraph out of it.

DEEP — 2026-08-17 10:32 ET (second DEEP of the session)

Sweep run in full, nothing new to act on, no orders. Recording that as the result rather than manufacturing a page out of it (L012).

What the sweep covered. 360 RSS docs (macro/markets/earnings), 30 HackerNews docs, the 705 per-symbol ticker-news rows this tick captured, and WebSearch over QUERIES (news, twitter, reddit) plus a targeted GLW/optical check. Brief persisted to logs/research/20260817T143249-signals.md, which paces the next DEEP. HackerNews returned 30 docs and 0 ticker signals — all archive stories, useless for this purpose, as it was at 13:58Z.

Every position-relevant RSS item it surfaced was already in the ticker files, and dated. The Wendy's/Trian-Peltz takeover report (08-12, sitting in QSR.md as a live forward RISK to the share-gain runway, still no formal bid), Burger King overtaking Wendy's as the #2 US burger chain (08-07), Restaurant Brands' Q2 beat (08-06), the Disney D'Amaro CNBC parks interview (08-14), and the NVIDIA [path]$3.2B Corning optical-capacity partnership (in GLW.md since the 08-07 entry). Re-finding a fact I already own is not a new fact, and the temptation to bank the QSR ones as fresh confirmation is precisely the trap QSR.md already names — accepting a STALE story as fresh because it flatters a thesis I hold.

The one genuine addition, and it is small. The 14:23Z FAST note read GLW's move as sleeve beta rather than Corning alpha (COHR +4.87%, LITE +2.45% against GLW +2.32% on SPY -0.15%) but could not SOURCE what was moving the sleeve. This sweep sources it: an AI-capex sentiment bid off Anthropic's reported >$11.5B Q2 revenue, SMH +1.13%, QQQ +0.36%, against DIA -0.20% and IWM -0.20% — semis/AI style box up, industrials and consumer lagging. That confirms the L006 rotation read rather than overturning it, and it touches no committed metric. Logged to GLW.md; the other five position files got nothing, because they had nothing new — the 14:23Z FAST wrote all six nine minutes ago and no fact has landed since.

No entries, and headroom is not the reason. Daily headroom is the full $500. The binding constraint is the $200 cash buffer against $222.14 buying power — roughly $22 deployable. The RIVN proceeds ($44.56, filled 13:56Z) are unsettled funds in a CASH account and are not spendable today. Nothing in this sweep was close to worth selling a flag-clean position to fund.

Flag check — all six clean, exit_flags: []. TGT +13.49% (peak +15.06%, ratchet locked +7.53% / $144.55, the only live ratchet in the book), ZBH +3.14%, DIS +4.50%, GLW +3.59% (fresh lifetime peak), ETSY -6.17% (L011 clock day 1 of 3, unadvanced — advances on settled closes only), QSR +4.58%.

The dominant risk is unchanged and dated: TGT Q2 Wednesday 08-19 pre-open (consensus EPS $2.23, [path]6.9% implied move), inside a consumer gauntlet of HD Tuesday, WMT Thursday and Fed minutes. Per L005 there is no pre-positioning into a scheduled binary in either direction. Per L008 the downside I have explicitly accepted is a gap THROUGH $144.55, not a tag of it. The read was pre-committed at 14:00Z and I am not restating it here beyond the pointer: positive comps + held-or-raised guide = hold; comps back negative = broken, exit on my terms. Brent near $90 on Iran ruling out an interim-deal extension stays a backdrop to TGT and QSR margins, not a metric.

Operator flag, repeated from 14:24Z because it is unresolved and will recur: today's order log still carries two orphaned pending rows for the RIVN exit (13:52:59Z @ $14.96, 13:54:29Z @ $14.88) from re-pricing the limit down to a fill. Broker state is correct and the position is flat RIVN; reconcile reports clean. trade.py has no cancel/void subcommand, so I will not hand-edit terminal states into the audit trail. Every re-priced exit will leave rows like these.

DEEP — 2026-08-17 11:05 ET

Full sweep run, no orders, one ticker file touched. Third DEEP of the session (13:58Z, 14:29Z, now 15:05Z). Inputs: 360 RSS docs, 30 HN docs (0 ticker signals), 705 per-symbol news rows, WebSearch over QUERIES plus targeted TGT/ETSY checks. Brief written to logs/research/20260817T150812-signals.md.

The sweep found one new fact and it is calibration, not a signal. TGT's Q2 numbers two days out read consensus EPS $2.32 on revenue $26.12B with [path]90bp of gross-margin expansion, an implied move of 7.08% vs a 4.93% four-print average, a split Street (12/15/2) and BofA at Underperform with the target lifted to $124. I carried $2.23 and [path]6.9% since 08-12, so this is a real update — logged to TGT.md. It does not grade my committed metric (positive comps, FY guide as tiebreak, pre-committed at 14:00Z today before the number exists), and per L005 it is not a reason to pre-position in either direction. It does push the L008 arithmetic one notch worse: a one-sigma down print now lands nearer $142 than $144, i.e. further THROUGH the $144.55 ratchet.

Everything else in the sweep was already in the files. The position-relevant RSS is dated 08-03 to 08-14 — the DIS parks interview, Burger King passing Wendy's, the RBI Q2 beat, the GLW midday-mover item. Today's fresh RSS is off-book (Alibaba/Intel/Sandisk premarket, the Ukraine strike). ETSY's targeted check returned only the Depop close, the 08-03 restructuring and the Goldman Neutral, all filed.

• TGT +13.93% (peak +15.06%, ratchet +7.53% / $144.55, the only live one) | ZBH +2.53% | DIS +4.03% | GLW +5.88% (fresh lifetime peak) | QSR +3.76% — flag clean.
• ETSY -5.67% — L011 clock day 1 of 3, unadvanced; advances on settled closes only. No 08-17 disclosure.
• Tape confirms the L006 rotation read, unchanged from 14:32Z: SMH +1.13%, QQQ +0.36% vs DIA -0.20%, IWM -0.20% on the AI-capex bid. GLW's strength is sleeve beta, not Corning alpha.

No entries, and the reason is arithmetic rather than judgment: headroom is the full $500, but buying power $222.14 against the $200 cash buffer leaves [path]$22 deployable, and the RIVN proceeds ($44.56, filled 13:56Z at $14.8524) are unsettled in a cash account. The cheapest name I would size today costs more than that. I am recording the constraint honestly rather than writing a thesis I could not act on.

DEEP — 2026-08-17 11:44 ET

Fourth DEEP of the session. No orders, no exit flags, and the reason there are no entries is capital, not conviction. Buying power $222.14 against the $200 cash buffer leaves [path]$22 deployable — not a position in any name on today's tape. Daily headroom is the full $500 and is not the binding constraint. The RIVN proceeds ($44.56, filled 13:56Z @ $14.8524) are unsettled in a cash account and unspendable until tomorrow, which even then only takes deployable to roughly $66.

I ran this sweep as an L021 audit rather than a news trawl — hunting my own files for facts I logged and then waved off — and that reframing is the only reason it produced anything. Three of the four candidate "new" facts turned out to be facts this book already owns, and I recorded each failure rather than letting a re-found headline masquerade as a fresh one:

• GLW / "Apple Glass Shift" (three outlets today) → the all-glass iPhone cancellation is stamped 08-10 and already in GLW.md, dismissed there on content (wrong segment — display/specialty, not the Optical line I underwrote) rather than on form. That distinction is exactly what L021 turns on. Failed the audit.
• QSR / "BK comeback can't offset Tim Hortons and Popeyes" → looked like a direct hit, since it grades consolidated performance instead of my committed BK U.S. comps. But named voider (b), written at entry, is Popeyes -5.2% deteriorating into a drag large enough to swamp BK. The alternative exam was already scheduled. Failed the audit.
• Iran / Hormuz / oil → already logged in TGT.md at 15:08Z today.

The one genuine addition is calendar structure, and it changes how I GRADE Wednesday, not whether I hold. The broker's verified calendar adds that TJX and LOW also print Wednesday morning alongside TGT (my file had only HD Tue / TGT Wed / WMT Thu). So Wednesday's 9:30 tape is a simultaneous off-price + home-improvement + discount print, and a red TGT will not cleanly separate "the market disliked Target's comps" from "the whole complex re-rated at once." I am naming that confound before the event, the same way the 08-12 implied-move note named the gap risk in advance. The corollary is free: HD prints Tuesday morning, one session ahead of me — an L015 same-day read on a big-box consumer print, available before I sit the same exam. I cannot act on it (L005 pre-commits no pre-position, no add, no trim; and there is no capital) but being calibrated by it costs nothing. Also flagged, not resolved: the broker carries consensus EPS $2.24 against the $2.32 Street figure I logged at 15:08Z — I am recording the [path]4c gap rather than adopting whichever number later makes me look right.

One item I could not date, logged as an open question: a TradingView recap flagged an ESPN spinoff proposal on DIS, and the search behind it contradicts itself (reporting Disney shelving a spinoff under D'Amaro, next to analyst pieces proposing one, next to 2023 Iger articles). A structural separation would re-cut the segments my committed metric (total segment operating income) is computed over, so it is not dismissible — but it is not tradeable on an undated recap either, and this file already carries the arrived-vs-occurred error twice. TO-DO next DEEP: date it to a primary source and establish direction. No clock opens on an undated item.

Comp reads — datum and clock state only (L020). GLW +5.30% against COHR +9.34% / LITE +6.83%: Corning is the weakest name in its own ripping sleeve, which resolves the alpha-vs-beta question against me — a fresh lifetime peak here is L006 style-box beta and buys no extra size. ETSY -0.70% against XRT -1.32% / XLY -0.94% — L011 clock day 1 of 3, unadvanced (settled closes only); the interim print happens to point my way today and that earns no characterisation either, which is the entire content of L020. ZBH -1.79% vs XLV +0.16% / SYK -1.03%. QSR -0.72% vs MCD -2.30% (third read where I win via MCD falling, still unsourced). TGT -1.13% vs XRT -1.32%. DIS -1.81% vs XLY -0.94%.

Ticker files touched: TGT, DIS, GLW, QSR. Deliberately not ETSY or ZBH — their clock state and comp reads were recorded at 15:33Z and nothing landed since; an identical observation set re-typed inside fifteen minutes is the L020 thumb-on-the-scale, not memory discipline.

Operator note (unchanged, still the binding constraint): this wallet is a cash account, so every sale's proceeds are dead capital for a day. A limited-margin upgrade is the fix and it is not something I can action.

DEEP #5 — two things changed, no order was possible

This one earns a full note (L012): a carried open question closed, and a dated test got pre-registered. The three DEEPs before it correctly wrote "nothing new"; this one did not.

1. Closed: the DIS ESPN item is dated, and it is April

The loose end from 15:44Z — a weekly recap surfacing "ESPN spinoff proposal" that I refused to call either a signal or a dismissal because I could not date it — resolves to 2026-04-28/29: Disney shelved the spinoff under D'Amaro, framed explicitly as not permanent, revisitable over the following six to nine months ([path]Oct 2026–Jan 2027). Four months old, three months older than my entry, therefore inside the price I paid, and structural rather than a hit to the committed metric (DTC segment OI/margin, total segment OI vs guide). Filed as a date, not a headline. New separation talk with a fresh dateline in that Oct–Jan window would be a different exam and would earn a re-underwrite; today's recap is not that. DIS.md already carries two errors from reading a freshly-arrived headline as a freshly-occurred fact — this was the third one queued up.

2. Opened: FN prints tonight, and my read is on the record BEFORE it lands

Fabrinet (FN) reports FQ4 after today's close, broker-verified, est EPS $3.79 (FN +3.91% into it). FN is the contract optical-manufacturing arm of the same AI-datacenter connectivity demand my GLW metric measures. Pre-registered call: if FN guides optical demand UP and GLW fails to hold its bid tomorrow, the tape is telling me the sleeve is already priced — a mark against the beta, not the thesis. I cannot trade it (no capital; L005 forbids pre-positioning around someone else's print). Same structure as the HD→TGT free look, one sleeve over.

3. GLW's +6.4% is sleeve, not Corning — and the tape says it louder than the headlines

The dated 08-17 driver is a CEO tariff-immunity interview ([path]90% domestic U.S. revenue, [path]1% of U.S. sales from Chinese factories), not a new contract; the "AI deals" headlines resolve to the Feb 3 Meta, Nvidia three-plant and Amazon fiber deals already in the file. COHR +10.39%, LITE +7.31%, GLW +6.38% on QQQ +0.26%: I am the laggard of my own sleeve on a sleeve-wide melt-up. L006 style-box beta — no added conviction, no added size.

4. Wednesday's confound got worse, named in advance

Broker-verified: TGT Wed 8:00am ET est $2.24, with LOW, TJX, EL and ADI the same morning, HD Tuesday, WMT Thursday — and new: FOMC minutes Wednesday 2:00pm ET. Morning tape is retail-complex noise, afternoon tape is rates noise, so there is no clean window on Wednesday at all. Consequence stated now, not after: I grade the print against comps + the gross-margin line where a tariff refund shows up; any relative-strength read waits for settled closes. Broker $2.24 vs Street $2.32 still unreconciled — adopting neither. DA Davidson's PT raise to $170 recorded, not acted on, dismissed on content (a PT grades sentiment about a number that has not printed) rather than the form-reflex L021 warns about.

Positions — flags clean, nothing sold, nothing added

TGT +14.09% (peak +15.06%, trailing ACTIVE at +7.53%/$144.55 — the only live ratchet; downside accepted as a gap through per L008). GLW +6.54%, QSR +3.93%, DIS +3.91%, ZBH +2.48%, ETSY -5.97% (fixed stop $76.28). ETSY L011 clock day 1 of 3, unadvanced — today's interim datum (-0.16% vs XRT -1.31%) points my way and earns no characterisation (L020, written from exactly this mistake). QSR outperformed a fourth-straight falling MCD (-2.03%) with still no MCD-specific disclosure in my capture.

Capital — the reason there is no trade, stated plainly

Daily headroom is the full $500 and is not the binding constraint. Buying power $222.14 against the $200 buffer = [path]$22 deployable, under one share of anything in this book; Friday's RIVN proceeds are still settling in a cash account. Equity $1,034.52. No orders this tick.

Next catalysts in order: FN guide tonight → HD tape Tuesday am → TGT print Wednesday 8:00am + FOMC minutes 2:00pm.

DEEP — one new fact, and it changes what I am grading on Wednesday

This is the sixth DEEP of the session and the first one that found something. No order — no exit flag is live on any of the six positions, and $222.14 of buying power against the $200 cash buffer leaves [path]$22 deployable, which is not a share of anything. Daily headroom is the full $500 and is not the binding constraint.

The find: I had the wrong consensus number, and I had it for five sweeps. All day I have been tracking a discrepancy between the broker's $2.24 EPS estimate for TGT and a $2.32 Street figure, flagging it as an open item to be settled by Wednesday's print. But TGT's committed metric in my own file is comparable sales, not EPS — so I was two days out from an earnings event, carefully maintaining a disagreement about a variable I never underwrote, while not knowing the passing mark for the one I did. This sweep got it: Street expects comps +2.5% (Truist at or slightly above 2.5%), on revenue $26.15B and EPS $2.35 — a third EPS figure, recorded and not adopted, and now moot rather than resolved.

Pre-registered, written into TGT.md before the event (L010): Wednesday 8:00am I grade comps against +2.5%, secondarily the gross-margin line where the tariff refund shows up. At or above, the metric passes and the thesis holds regardless of the tape — the load-bearing clause, because I have already logged that Wednesday has no clean tape window (LOW/TJX/EL/ADI print the same morning, FOMC minutes at 2:00pm). Below +2.5% is a comps reversal, my first-listed exit trigger since entry, and the position is then mine to defend or exit on the number rather than the price. Options imply a 7.08% move; per L008 the downside into it is priced as a gap through the $144.55 ratchet, not a tag of it.

Two smaller appends, both filed as data rather than verdicts (L020). GLW: the sleeve lag widened across the session — GLW +5.95% against COHR +11.16% and LITE +7.11%, versus +6.38%/+10.39%/+7.31% at 16:23Z. A lifetime-peak print bought by sleeve beta still buys no size; the pre-registered FN test (reports tonight, +3.82% into it) is graded tomorrow, not tonight. ETSY: sourced the Q2 numbers to the shareholder letter itself, and recorded which line the L011 clock is really about — habitual and repeat cohorts grew for the first time since 2023 — so that the replacement number is sitting in the file before the clock ever runs to term (L021). Interim datum only: ETSY +0.11% vs XRT -1.27%, XLY -1.07%; clock day 1 of 3, UNADVANCED, adjudicated on settled closes, no characterisation in either direction.

Comp reads: TGT -0.74% (XRT -1.27%, WMT -0.29%, HD -0.27%) · ZBH -1.86% (XLV +0.13%, SYK -1.17%) · QSR -0.32% (MCD -2.14%, fifth straight read where MCD falls harder with no MCD-specific disclosure I can source it to) · DIS -2.14% (XLY -1.07%). Tape: SPY -0.19%, QQQ +0.17% on a semis rebound.

Next: FN's guide tonight -> HD's tape Tuesday am -> my own print Wednesday 8:00am, graded against +2.5% comps.

DEEP #7 — the longest-running open question in the book closed, and it closed against me as much as for me

This one earns a full note (L012). Six DEEPs today wrote "no MCD-specific disclosure in my capture." This one found it.

1. Closed: the MCD drift behind my QSR read is dated 2026-08-04 and is inside my entry price

Every relative read on QSR today — six of them — came out the same way: QSR beats MCD because MCD falls, not because QSR rises, and I could not source the fall. Now I can. McDonald's Q2 2026, reported 08-04: U.S. same-store sales +0.8%, U.S. comparable guest counts NEGATIVE, visits -4.6% y/y, global comps decelerating to 1.3% from 3.8%, revenue $7.10B vs $7.13B consensus, new U.S. president installed. MCD -13.3% YTD at two-year lows, -14.5% off its February [path]$319 peak.

Filed as a DATE, not a headline — the same handling the DIS/ESPN item got at 16:23Z. The print predates my 08-07 entry by three days, so it is already inside the $74.27 I paid. It resolves an observation. It does not create a catalyst, it moves no exit level, and it opens no clock on a ticker I do not own.

2. And it hands me my own falsifier — which is the half that matters

McDonald's CEO attributed roughly two-thirds of the U.S. traffic miss to a botched "$3 EDAP" value-menu rollout, in his words an "execution failure." My QSR falsifier, written at entry before I owned a share, is "BK U.S. comps rolling back toward flat, or MCD's U.S. business visibly stabilising." A self-inflicted promo bungle with a named cause is the most fixable miss there is — much easier to reverse than structural share loss.

So the document that finally explains the spread also says two-thirds of the gap I am long is MCD's unforced error, not Burger King's pull. The thesis is not broken and nothing is sold. What changes is the bar: November grades BK U.S. comps on their own merit, not the gap — and a widening MCD-vs-QSR spread between now and then is not confirmation. Written down today so the Q3 print is graded against a bar set in August (L010).

This is the L021 shape running in the honest direction. I have spent six notes correctly declining to characterise an unsourced drift, and the payoff for waiting was a fact that qualifies my position rather than flattering it.

3. The BK side got a mechanism — logged as colour, explicitly NOT promoted to a metric

New and dated: Whopper sales +20% since the February recipe-and-packaging rework, $700M committed to remodels and value offers through 2028, franchise disclosure showing [path]14% average sales uplift and [path]13% traffic uplift on remodeled units. Until today my thesis was one number (+8.5% U.S. comps) with no account of why; this is the why, and it is capex-and-menu driven rather than promotional, which is the durable kind. The committed metric does not change and is not supplemented — BURGER KING U.S. COMPARABLE SALES, Q3 print, early November. Bolting on a second metric because today's news was good is precisely the error L010 exists to prevent.

4. Everything else: datum and clock state only (L020)

QSR -0.28% vs MCD -2.22%, YUM -1.98%, XLY -1.31%. ETSY -0.54% vs XRT -1.52%, XLY -1.31% — L011 clock day 1 of 3, UNADVANCED; clocks adjudicate on settled closes and the bell has not rung. TGT -1.31% vs XRT -1.52%, WMT -0.72%, HD -0.59%. ZBH -2.14% vs XLV +0.01%, SYK -1.35%, MDT -0.27%. GLW +5.47% vs COHR +9.54%, LITE +6.52%, FN +3.54% into tonight's print — third consecutive sweep with the lag widening, L006 style-box verdict stands: no added conviction, no added size. Macro SPY -0.33%, QQQ -0.10%.

The Wednesday TGT pre-registration from 16:57Z is carried verbatim and not re-derived: grade comparable sales against +2.5% consensus at 8:00am (Street EPS $2.35, revenue $26.15B, 7.08% implied move), secondarily the gross-margin line where a tariff refund shows. Nothing in this sweep touches it.

5. Positions and capital

Flags clean on all six. TGT +13.41% carries the book's only live ratchet (peak +15.06%, stop +7.53% / $144.55); ETSY -6.24% is nearest its fixed stop at $76.28. Daily headroom is the full $500 and is not binding — buying power $222.14 against the $200 cash buffer leaves [path]$22 deployable, under one share of anything on this tape. Friday's RIVN proceeds are still settling in a cash account. Equity $1,031.27. No orders this tick.

Next: FN's guide tonight -> HD's tape Tuesday am -> TGT's print Wednesday 8:00am, graded against +2.5% comps.

DEEP #8 — nothing new landed, and this is the short note that says so (L012)

The sweep ran in full (360 RSS docs, 40 HN docs with 0 ticker signals, per-symbol ticker news across all six holdings, a comp-wide broker quote sweep, get_accounts, WebSearch over QUERIES) and produced no new fact that touches a position. Three candidates were audited and all three failed:

• U.S.-Iran ceasefire expiry / risk-off tape — the session's macro driver, already logged in TGT.md at 15:08Z and in the 15:44Z brief. Brent is $88.31 (-0.24%), below the "near $90" framing this morning's notes used.
• "Home Depot CEO taking temporary medical leave" (CNBC) — new to my capture, but HD is a consumer READ for tomorrow morning, not a position, and a CEO medical leave is governance noise against that read. HD -0.48%.
• "Burger King overtakes Wendy's as #2 US burger chain" — dated 08-07, in QSR.md since entry.

The one genuine addition is the capital constraint, now exact instead of estimated. get_accounts confirms unsettled_funds $44.56 — Friday's RIVN proceeds to the cent — sitting in a CASH account, therefore unspendable and excluded from buying power. $222.14 buying power against the $200 buffer leaves [path]$22 deployable, under one share of anything in the book or on the watchlist. Daily headroom is the full $500 and is not binding. Eight DEEPs, zero entries, and the reason is plumbing rather than conviction.

Datum and clock state only (L020). GLW +5.73% vs COHR +9.96%, LITE +6.44%, FN +3.99% — eighth same-direction read, L006 style-box verdict stands. ETSY -1.26% vs XRT -1.50%, XLY -1.30% — L011 clock day 1 of 3, UNADVANCED. TGT -1.35% vs XRT -1.50%, WMT -1.00%, HD -0.48%. ZBH -2.10% vs XLV +0.01%, SYK -1.38%, MDT -0.19%. QSR -0.61% vs MCD -2.37%, YUM -1.80%. DIS -2.39%. SPY -0.36%, QQQ -0.15%.

Carried verbatim, not re-derived: the FN guide tonight (graded on GLW's bid tomorrow), and TGT Wednesday 8:00am ET graded on comparable sales vs +2.5% consensus (BofA +2.5%, Truist at-or-above; EPS $2.35, revenue $26.15B, 7.08% implied move).

Flags clean on all six. TGT carries the only live ratchet (peak +15.06%, stop +7.53% / $144.55, downside accepted as a GAP THROUGH per L008); ETSY is nearest its fixed stop at $76.28. Ticker memory updated for the two names with live tests (ETSY's clock, GLW's pre-registered FN read); the other four were rewritten between 16:20Z and 17:32Z and are deliberately left alone rather than re-narrated. No orders this tick.

Next: FN's guide tonight -> HD's tape Tuesday morning -> TGT's print Wednesday 8:00am.

Ninth DEEP of the session. Flags clean on all six, no order placed — but unlike the last four sweeps this one is not a one-liner, because two genuinely new facts landed and L009 says re-analysis is earned by a new FACT.

1. Wolfe Research upgraded $ETSY to Outperform from Peer Perform, PT $100, today. Recorded, not acted on — the fourth analyst action on this name in three sessions (Goldman Neutral, Canaccord $92, Evercore $83). The dismissal is on CONTENT and not on form, per L021: a price target grades sentiment about numbers already printed, while the number this position rides is the cohort/retention line (habitual and repeat buyer cohorts growing for the first time since 2023) — which Wolfe is not marking, and which is already written into ETSY.md as the replacement number should the L011 clock run to term.

Naming the trap I walked up to and stepped around: an upgrade landed and the tape declined to pay for it (ETSY -1.24% vs SPY -0.37%). That is the L015 shape exactly. But L015 is an ENTRY-side rule by construction — its whole claim is that the look is free because nothing is at stake on a position you do not own. I own this one, so the adjudicator stays the L011 clock on settled closes, and a same-session tape reaction to a broker note is not evidence I get to promote. Clock: day 1 of 3, UNADVANCED.

2. The AI rally has a named driver, and it answers L006 against me. Chipmakers rallied on Anthropic's revenue surge bolstering the AI trade — the attribution for a move I have spent eight sweeps calling "sleeve-wide" without naming what moved the sleeve. Run L006's test honestly: GLW +5.51% against COHR +9.24%, LITE +5.83%, FN +4.45%, on a tape where SPY is -0.37%. The whole style box is up and GLW is mid-pack in it, lagging the leader by [path]3.7pp. That is beta in the winning box, not idiosyncratic confirmation. No added conviction, no added size — and the committed metric (Optical Communications y/y at/above [path]30%) is untouched, because a sector rally does not grade it.

3. No buys, and the reason is plumbing rather than conviction. get_accounts re-confirms unsettled_funds of exactly $44.56 — Friday's RIVN proceeds have not settled during the session — so buying power $222.14 against the $200 cash buffer leaves [path]$22.14 deployable, under one share of anything in this book or on the watchlist. Daily headroom is the full $500 and is not the binding constraint. Nine consecutive DEEPs, zero entries, same cause each time.

Comp reads, datum and clock state only (L020). TGT -1.58% vs XRT -1.50%, WMT -1.01%, HD -0.67%. ZBH -2.21% vs XLV -0.08%, SYK -1.69%, MDT -0.36%. QSR -0.43% vs MCD -2.53%, YUM -1.98%. DIS -2.66%.

Carried unchanged and deliberately NOT re-argued: TGT's ratchet (peak +15.06%, stop +7.53% / $144.55, downside accepted as a GAP THROUGH per L008) and its comps-vs-+2.5% pre-registration for Wednesday 8:00am ET; ETSY's fixed stop $76.28, 3.3% cushion, the tightest leash in the book; QSR's Q3 BK-US-comps test in November; ZBH's group-organic-4%-with-knees test in early November.

Next: FN's guide tonight -> HD's tape Tuesday morning -> TGT's print Wednesday 8:00am, graded against +2.5% comps.

DEEP #10 of the session. Flags clean on all six; deployed today $0.00; daily headroom $500 and not the binding constraint. No order placed, and none was possible: get_accounts re-confirms unsettled_funds of $44.56 in the agentic wallet, so buying power $222.14 against the $200 cash buffer leaves [path]$22.14 — under one share of anything in the book or on the watchlist. Ten DEEPs, zero entries, on plumbing rather than conviction. Brief written: logs/research/20260817T191752-signals.md.

The one thing this tick changed, and why it earned a full log rather than L012's one line. A new, dated fact landed directly on a pre-registration I made three hours ago. At 16:57Z I wrote down the passing mark for Wednesday's TGT print: comparable sales graded against Street +2.5%, secondarily the gross-margin line. This sweep surfaced (a) Wolfe's Spencer Hanus raising his Q2 same-store-sales estimate to 3% — Hanus being the author of the 06-23 Outperform/$162 note that is this position's entry catalyst — and (b) a sell-side mean price objective of $143.15 on 12 Buy / 15 Hold / 2 Sell, roughly 5.5% below the $151.48 spot.

I did not move the bar. A pre-registration edited 40 hours before the exam is not a pre-registration. What I recorded instead is the consequence, in advance: a print between +2.5% and +3.0% passes my committed metric and may still be sold, because the buy-side bar sits at the top of that range. That is the L011 "grading the wrong exam" shape caught BEFORE the event rather than six sessions after it, and it costs nothing to name now — whereas naming it Wednesday afternoon, with a price in front of me, would just be an excuse. It also sits cleanly inside what this file already says: Wednesday has no clean tape window (LOW/TJX/EL/ADI the same morning, FOMC minutes 2:00pm), so the tape cannot adjudicate either way; the number does. The $143.15 mean is recorded and not acted on, dismissed on CONTENT not form per L021 — a target mean grades sentiment about a number that has not printed, and my metric IS that number. It is not one of my exit triggers, and I held this name through Barclays at UW $91, so treating an aggregate opinion as binding now would be incoherent.

The read I deliberately declined to take. QSR is -0.44% vs MCD -2.61% and YUM -2.02% on an XLY -1.07% tape — [path]2.2pp of idiosyncratic-looking outperformance, arriving one sweep after the Burger-King-overtakes-Wendy's item. That is a very easy story to write and L009 forbids writing it: relative strength gets read ONCE, on settled closes, against the true comp set. My own TGT file from 07-28 is the standing evidence — a peer edge that decayed +1.74pp -> +0.80pp -> +0.40pp inside one session and forced a walk-back. Datum logged to the ticker file, no promotion, no size.

Carried unchanged and explicitly NOT re-argued (L009/L012/L020): ETSY's L011 clock day 1 of 3, unadvanced — today's intraday reading (-1.59% vs XRT -1.45%) has flipped sign versus the 18:42Z one, which is exactly why the mechanism counts settled closes and not prints, and neither reading is a tick. GLW +5.23% vs COHR +8.18% / LITE +4.99% / FN +4.68% answers the L006 factor test the same way it did at 18:42Z, in the unflattering direction: mid-pack in a hot box, beta not alpha, no added conviction; FN's FQ4 guide lands after tonight's close and the pre-registered read on it stands verbatim. ZBH's group-organic-4%-with-knees test in early November. QSR's BK-US-comps test in November.

Risk state. Equity $1,028.54. TGT +12.69% carries the book's only live ratchet (peak +15.06%, stop +7.53% / $144.55); per L008 Wednesday's 7.08% implied one-sigma lands in the low $140s — through, not at, that level, so the real downside into the print is a gap through, accepted at entry rather than hedged, and no pre-positioning or trim either way (L005). ETSY -7.23% is nearest a stop, fixed at $76.28, [path]3.1% of cushion. Next: FN's guide tonight -> HD's tape Tuesday morning -> TGT Wednesday 8:00am, graded at +2.5% with the 3% whisper on the record.

DEEP, 19:51Z — the eleventh and last DEEP of the session, run with eight minutes to the bell. Full sweep: RSS macro/markets/earnings (360 docs), HackerNews (30, zero ticker signals for the eleventh straight time), per-symbol ticker news across all six holdings (183 docs), WebSearch over QUERIES. Brief written to logs/research/20260817T195116-signals.md.

Result: no new fact grades any committed metric. No exit flag. No order, and none possible.

• TGT +12.68%, trailing ACTIVE, stop $144.55 (peak +15.06%). Q2 Wednesday 8:00am ET: consensus EPS $2.35 / rev $26.15B, options implying a 7.08% move, sell-side mean target $143.15 — below the $151.46 spot. My pre-registered passing mark stays verbatim at Street +2.5% comps; I am not editing a pre-registration 40 hours before the exam because a 3% whisper appeared. A one-sigma down move lands in the low $140s — through, not at, the ratchet — accepted as gap risk at entry per L008, not hedged.
• ETSY -7.54%, fixed stop $76.28, [path]2.7% cushion — the tightest leash in the book. L011 clock day 1 of 3, UNADVANCED: it adjudicates on settled closes and the bell has not rung. Today's Goldman reinstatement at Neutral is a rating, not a GMS number; recorded, not counted.
• GLW +5.74%, trailing off. Fabrinet FQ4 after tonight's close is the first read on the pre-registered sleeve test: if FN guides optical demand up and GLW does not hold its bid tomorrow, the sleeve was already priced — a mark against the beta (L006), not against the committed metric (Optical Communications y/y ≥ [path]30%, next tested at the late-October Q3).
• ZBH +1.77% (stop $86.83), DIS +2.86% (stop $90.91), QSR +4.25% (stop $66.84) — flags clean, nothing dated landed on any of the three. The ZBH item was a 13F share count; the DIS items were D23/ESPN-spinoff recaps; the QSR items were post-Q2 hold-or-sell explainers. All dismissed on content per L021 — none of them measures the number I committed to.

No buys, and the reason is plumbing, not conviction. $222.14 buying power against the $200 cash buffer leaves [path]$22.14 deployable; Friday's $44.56 of RIVN proceeds stay unsettled through the bell in a CASH account. Daily headroom sat untouched at the full $500 all day and was never the binding constraint. Eleven DEEPs, zero entries — I could not have bought one share of anything on the watchlist if the brief had screamed.

No ticker-memory writes this tick, deliberately. All six files were refreshed between 18:41Z and 19:16Z and the sweep returned nothing new to add to any of them. Appending six near-identical paragraphs whose only fresh content is a price that moved a few cents is precisely the L012 failure mode — a claim about an unchanged position manufactured out of noise. The read that actually adjudicates anything today (ETSY's clock, GLW's sleeve, the settled closes) arrives after 20:00Z, and that is when the files get written.

The session in one line: one order (the pre-committed RIVN exit at 13:56Z, 3 shares @ avg $14.85, executed at the open), then eleven sweeps confirming nothing else broke. Next: FN's guide tonight → HD's tape Tuesday 8:00am → TGT's print Wednesday 8:00am, graded against +2.5% comps.