Question: blind rederivation spot check week ending 2026-10-16 (cadence loop-34; method .brief/correctness-rederivation-audit.md "Cadence").

Verdict: 2,592 numbers checked, 2,409 MATCH, 5 MISMATCH, 178 UNVERIFIABLE. All five mismatches reached a reader (SHIPPED). They group into two mechanisms, and each gets one proposed DEFECTS row (below). None of the five is price math:

• One is a ticker. A generated brief section printed the Frankfurt listing 2GB.DE as "DE", which is Deere's symbol.
• Four are stale negatives. The three 10-09 scans say nothing dated October 9 is on file, but the same publication shipped 156 captures dated that day, including the copper cause that the same night's brief printed.

Every tape figure re-derives:

• the machine-written lazarus scan;
• all 711 rendered bound cells;
• the brief's rotation, lane-tape and desk-call columns;
• the three desk calls.

What we're asking

Do the numbers a reader saw this week survive an independent recomputation from L1? L1 here means:

• the on-disk daily bars (data/stocks/<SYM>/ohlc/max.json through python/src/bars.py read_daily, trimmed to each artifact's data date);
• EDGAR company-facts;
• the FRED series under data/macro/;
• the artifact's own cited on-disk sources.

Method

• Seed and picks. The candidates were the reader-facing files dated 2026-10-09 to 2026-10-15:

  • research/issues/*-market-brief.md: 5 files (10-09, 10-12, 10-13, 10-14 and 10-15);
  • research/deep-dives/: 3 (BE and GNRC on 10-09, ARGX on 10-12);
  • research/company-profiles/: none (all 53 are dated 2026-09-28);
  • research/scans/: 49;
  • research/signals/: 1.

  That is 58 candidates. No 10-16 files were included. None carried audience: internal; scans carry no audience field, so they default to reader. random.Random(20261016).shuffle(sorted(candidates)) (seed 20261016) took the first five:

  1. research/scans/2026-10-14-lazarus-scan.md
  2. research/issues/2026-10-13-market-brief.md
  3. research/scans/2026-10-09-food-security.md
  4. research/scans/2026-10-09-optical-supply-chain.md
  5. research/scans/2026-10-09-macro-commodities.md

• Desk calls (M=3). research/api-data/desk-calls.json has no conviction field, so the ranking is last week's:

  • First, open calls with stance buy. That gives CRDO (2026-07-08) and STX (2026-07-05), the only two, and the same pair as last week.
  • Then the most recent open call. Two are dated 2026-10-09: ASTS (pass) and CEG (watch). The tie went to CEG, because a watch carries an entry condition and a pass does not. File order would have given ASTS.
  • The open avoid (HAL, 07-31) ranks below both.

• Math. Code did all of it. The scratch scripts are tape.py, lazarus.py, calls.py, bound.py, lanes.py, rot.py, fund.py, shelf.ts and tally.py, in the session scratchpad blind-1016/. They mirror the engine's definitions:

  • Wilder RSI over the full daily series.
  • Calendar-day change1d/7d/30d/3m from the newest bar on or before the target date.
  • change1y from the nearest bar within 10 days.
  • 52-week high = maximum intraday high in the trailing year (370 days for lazarus, as its code does).
  • relVolume against the prior 20 bars.
  • The trend label from price against the 50- and 200-day.
  • Lane medians as buildLaneTape computes them: summary-file membership, rows on the session, and n≥5 for the 90% counts.
  • Rotation RS as lib/rotation-map.ts computes it: median 30-day change of the fresh rows minus SPY's 30-day change, plus acceleration, 7-day RS and divergence.
  • Desk-call returns: the close on or after the call date and the close on or after +N calendar days, against SPY.
  • Support shelves and base pivots: the engine's own volumeProfileScalars and detectMomentumBase, run read-only on bars trimmed to the date.
  • Revenue growth from EDGAR company-facts, with the latest filing per period (restated comparatives). For the oil names, contract revenue (RevenueFromContractWithCustomerExcludingAssessedTax), because total Revenues carries hedging gains and losses.

• Helper check. Before trusting the helper:

  • It reproduced 1,408 of 1,409 fields on today's summary rows (118 symbols across food-security, optical, macro, lazarus, focus and neoclouds, at their 2026-10-16 latestDate). The one miss is CTVA's relVolume, a post-spin series.
  • It reproduced every one of the 10 lazarus rows dated 10-14 exactly: score, the five checks, price, RSI, from-high, and the SPY 21- and 63-bar benchmark.
  • It reproduced the brief's 50 movers-table cells.
  • It reproduced 52 of 52 desk-call 30-day excess values.
  • The shelf script matched today's summary shelves and pivots for NTR, ADM and DE.

• Tolerances. These count as MATCH:

  • a value inside its printed rounding;
  • x.x5 boundaries;
  • double rounding (a 2dp summary value rounded again to 1dp in prose: MDLZ 3.153 → 3.15 → "3.1");
  • JS against Python half-rounding (5/8 = 62.5% printed as 63%).

• Bound cells. Every summary-bound cell is checked at the binding level: right symbol, right field, a date label present. It is also checked against the rendered text in the shipped projection: 670 numeric cells and 41 latestDate labels, each equal to bars at the artifact's date. The binding resolver now pins cells to the artifact date. The 10-13 brief's projection, regenerated 2026-10-16T21:08Z, still renders 2026-10-13 values under "As of Tuesday's settled close, 2026-10-13". Last week's row D invariant held across this sample.

• Caveat (A746). Summaries are overwritten in place. Bars were rewritten 2026-10-16T21:00Z, and Massive has been the primary OHLC source since 10-06. Every dated tape value here is therefore consistent with on-disk bars, not proven equal to the row the author read. Provider and basis revisions since publication count as UNVERIFIABLE, not MISMATCH.

What we found

Per-artifact table

What each artifact's counts cover:

• Lazarus 10-14. For all ten rows, every cell matches 10-14 bars: score, status, the five checks, buy zone, price, RSI and from-high. So do SPY's +1.6% over 21 bars and +2.5% over 63, and the waking-up list (ISRG and ZS at 4/5).

• Brief 10-13 (desk-written sections plus the generated block).

  • The tape figures in the Read, Pulse, Hatch, Focus List (26 names), Sector Scorecard (21 lanes) and What I'd Tell a Friend all match 10-13 bars. That covers prices, 1D and 7D moves, RSI, distances from the 200-day, years, and distances from the high.
  • So do the named levels:
    • MSFT's $542.74 pivot (10-13) and $537.22 (10-12);
    • MU's $965.55 shelf (10-12), $921.02 next shelf and $981.98 50-day;
    • AAPL's $345.34 pivot on a 14-week base;
    • AMZN's $254.90, OUST's $33.18 and ARGX's $790.26 shelves;
    • SMH's $580.10 and NBIS's $224.70 50-day averages;
    • UNH's $397.63 open against a $378.52 prior close.
  • NBIS's "$219.70 shelf" was the engine's nearest support shelf through 10-07. It is carried as a named level and is consistent with the bars.
  • All 25 lanes the brief names match: median, up/n and RS. So do the 125-lane count, "1 closed at least 90% up" and "10 at least 90% down" (under the builder's n≥5 rule; see L1), and the neocloud lane's last place on the month (RS −19.0).
  • The rotation paragraph: 130 of 130 parsed numbers match (RS, % strong-up, today's move, acceleration, % up, 30d/7d/divergence), plus the SPY +1.8% benchmark and the leadership shifts against 10-12.
  • Regime counts are priceState 246/371/230/148, with 41 flips, 0 entries, 14 exits and the six named flips.
  • FRED readings and dates, the 12 climate-trigger levels and latest values, and the VIX gate and credit tape all match.
  • Crowd odds match against the files' own history.
  • EPS surprises match: UNH +13.3, JPM −3.2, GS +30.0, JNJ +2.7, C +7.9, WFC +0.8 and DPZ +4.8 all reproduce from the provider's unrounded estimates.
  • Upcoming dates match except DE (item 1).
  • The other generated tables all match: alarm prices, 52 desk-call 30-day excess values, the origination queue's 7d/30d, targets in zone with their percentages and trend labels, the movers tables and the derivative exclusions.
  • Fundamentals-wire revenue growth: 12 names, 24 figures, all reproduced from EDGAR. OXY's +37%/−4% only reproduces on OxyChem-restated comparatives (7,053/5,163 and 5,569/5,813). ABCL's receivables are −45.1% and revenue −76.3%.

• Food security 10-09. All prose tape figures match 10-09 and 10-02 bars:

  • "10 of 11 red", "six of nine below the 200-day (from five)", "five negative-alpha (from four)";
  • Deere's −9.62% week and its cushion shrinking from 16.9% to 4.8%;
  • Deere −3.8% and CNH −6.1% on 10-07;
  • the shelves (DE 607.13, ADM 78.38, BG 94.21, NTR 65.83);
  • ADM's 22-week base with an 88.75 pivot;
  • the report dates;
  • CTVA and VYLR's seven post-spin bars and their week's changes.

  The 87 bound cells all render 10-09 values with a 10-09 as-of column.

• Optical 10-09. Every count matches:

  • 5 of 28 up on the week, 15 of 28 green on Friday, 0 of 28 on 10-08 (median −5.76%);
  • 20 of 27 above the 200-day (seven below, from six), four negative-alpha (from one), ten under the 20-day (from one), 25 of 28 green on 10-02.

  So do the leaders and laggards, the 40–45 band and the shelves (LITE 914.94, CIEN 425.66, AXTI 64.88, AEHR 80.37). Also matching:

  • GLW's 7.4-week base and 171.3 pivot;
  • SNDK's 1582.50 50-day;
  • AXT's three straight losing sessions;
  • nine report dates.

  The 252 bound cells all match.

• Macro commodities 10-09. Every count and level matches:

  • EWZ +14.01%, RSI 76.2;
  • UUP at RSI 69.9, 0.21% from its high;
  • 18 of 37 above the 200-day;
  • the oversold tier of 11 names in RSI order;
  • EWJ basing → uptrend and AA collapse → downtrend (regime files);
  • GLD −1.67% on 10-07;
  • DGS10 5.28% and DFII10 2.92% on 10-07;
  • Teck's 10-07 and 10-08 6-Ks on disk;
  • no stock events from 10-03 to 10-09 on the board.

  The 311 bound cells match, including the latestDate header.

• Desk calls.

  • CRDO:

    • entry and invalidation: 33.3% operating margin on $1.335B FY26 revenue (10-K 2026-06-15); the ~$255.50 20-day and "flagging under it" on 07-07; the ~$160 200-day;
    • the 09-14 flag ($150.09, −14.7%, $175.89, RSI 30.2, −42.3%);
    • the 10-09 note ($216.55, RSI 57.0, strong-up, +29.0%, 93 days);
    • returns: −3.40/3.74/−7.14 and −14.64/4.52/−19.16 (a −19.155 boundary), to date −34.40;
    • the current 200-day, 180.18.

  • STX:

    • the 23% → 32% four-quarter rise in total-company GAAP operating margin (EDGAR FQ4'25–FQ3'26: 23.2/26.4/29.8/32.1);
    • the 10-09 note;
    • the returns.

    Last week's row E is fixed: the call now says "company-wide GAAP operating margin … Seagate reports no separate nearline margin", and its invalidation is computable.

  • CEG: the $286.14 200-day and $298.07 entry close at 10-09, the $300.40 close on 10-06, the 11-09 report event, and −8.96% to date.

Every MISMATCH, with evidence

1. issues/2026-10-13-market-brief.md:318, Upcoming Prints: "2026-10-15 — … time unknown: DE".
   • The only event behind it is research/stock-events/2026-10-15-2gb.de-earnings.md. Its sole declared ticker is 2GB.DE, a Frankfurt (XETRA) listing in the ai-bottlenecks watchlist. Deere's next event is research/stock-events/2026-11-25-de-earnings.md.
   • research/api-data/calendar.json carries "tickers": ["DE"] for the 2GB.DE event. That holds in the committed versions in 319da94ec4 and 347b5a906d.
   • Root: research/build-terminal-projections.ts:712. The declaredTickers filter ^([A-Z][A-Z0-9.\-]{0,9}|\d{4,6}\.[A-Z]{1,3})$ rejects 2GB.DE (a digit, then letters). calendarTickers (:720) then falls through to symbolsFromText, which finds "DE".
   • The comment at :707–711 records the same failure for 005930.KS → "KS" on 2026-07-30. That fix only admitted 4–6-digit prefixes.
   • Live-corpus sweep (declared tickers that all fail the filter): 1 artifact (this event) and 1 watchlist symbol (2GB.DE).
   • The same line printed in the 10-12, 10-14 and 10-15 briefs (outside the sample) and reaches /calendar through calendar.json. SHIPPED.
2. scans/2026-10-09-food-security.md:43 (repeated at :47 and :145): "No articles or posts dated October 9 are on file."
   • The EOW commit that shipped this scan (6d07511e14, 2026-10-09 23:53 PT) also shipped 156 captures dated 2026-10-09. The 10-09 market brief, written that evening, cites them.
   • The claim is false at publication.
   • The scan's substantive point stands: none of those captures concerns fertilizer, Deere, grain or farm equipment, so "no cause on file" for the ag selloff holds. SHIPPED.
3. scans/2026-10-09-optical-supply-chain.md:60 (and :178): "no articles or posts dated October 9 are on file."
   • This is the same false-at-publication count.
   • The single-name causes stay unknown. The 10-09 tech captures are generic market wraps. SHIPPED.
4. scans/2026-10-09-macro-commodities.md:58 (and :300): "no October 9 captures or news pulls are in the corpus" and "No October 9 article, tweet, Reddit or filing captures exist." False at publication, as in items 2–3. SHIPPED.
5. …macro-commodities.md:73 (and :85, :276, :289): "Friday's metals and copper bounce has no dated cause: the newest captures are October 8", listed as a "Missing receipts" watch item.
   • research/capture/articles/2026-10-09-bloomberg-copper-clinches-weekly-gain-as-chinese-demand-signals-streng-tmm5qht9njls00.md credits "improving Chinese demand signals and a rebound in global equities" for copper's recovery on 10-09.
   • The 10-09 brief prints exactly that at :77 ("Bloomberg crediting strengthening Chinese demand signals for copper's weekly gain"), and again at :98, :166 and :209.
   • So two pages from one publication disagree on whether a cause is on file.
   • Writing order cannot be established. On this checkout the capture files were born in a two-second batch (13:54:33–35 PT), which looks like a sync, not authorship. The scans were written 13:55–14:01 PT, and their projections were generated at 14:28 PT.
   • What is established: the publication carried both. SHIPPED.

Proposed DEFECTS.md rows (shipped; file BEFORE any fix)

• A — The calendar resolver prints the Frankfurt listing 2GB.DE as Deere (DE) (item 1).

  • Inject: a stock event whose only declared ticker is an exchange-qualified symbol that starts with a digit followed by letters (2GB.DE).

  • Scream: the event's calendar.json tickers equal its declared ticker (or the row holds). Upcoming Prints never names a listed company the event does not declare.

  • Caught by: blind re-derivation spot check, loop-34, 2026-10-16. No gate flagged it across four briefs.

  • Class: unvalidated-security-identifier, in its text-scavenger-fallback form. It has the same root as the 2026-07-30 "KS" case in the code comment.

  • What would stop it:

    • admit declared symbols by membership in the identity authority (watchlists plus _securities.json) rather than by a shape regex;
    • hold the row when a declared ticker is refused, instead of scavenging the text.

    The replay fixture is the 2GB.DE frontmatter. The live sweep is 1 artifact and 1 symbol, a finite invariant.

• B — The 10-09 scans shipped "nothing dated October 9 on file" while the same publication held 156 such captures, including the copper cause the same night's brief printed (items 2–5).
  • Inject: scan prose drafted before the evening capture ingest and published after it, unchanged.
  • Scream: an "absence" sentence ("no <date> captures / no dated cause on file") is true against the corpus at publish time, and two pages in one publication do not disagree about it.
  • Caught by: spot check.
  • Class: silence-as-absence, in its stale-negative (timing) form.
  • What would stop it: make absence claims values computed at publish (a bound count of captures dated D, or a publish-step re-run of the cause check) rather than typed prose.

LATENT findings (not reader-wrong values; belong in TASKS-AUDIT)

• L1 — The Lane Tape's 90% counts hide their n≥5 floor.
  • The brief prints "1 closed at least 90% up", while its own table shows Specialty Electronic Gases at 3/3.
  • The count is right under LANE_UNANIMOUS_MIN_N = 5 (research/build-daily-brief.ts:1665), but the label never says so.
• L2 — The fundamentals wire led MU with an older quarter.
  • On 10-13 it printed "+346% YoY (FQ3) vs +196% YoY (FQ2)". MU's FQ4 10-K was on EDGAR from 2026-10-09: revenue $54.2B, +379.3% YoY.
  • The labels are true, and the series is still accelerating. The Massive statement path lags EDGAR.
• L3 — Dated volume ratios recompute higher, never lower.
  • Eight printed relVolume figures recompute higher on today's rewritten bars: NTR 1.83→1.86, CF 1.74→1.89, DE 1.47→1.52, AAOI 2.42→2.45, CIEN 2.47→2.48, EWZ 1.47→1.49, CIFR 3.87→3.88 and JNJ 2.41→2.42.
  • None recomputes lower.
  • This fits a settled-close fetch (13:34 PT) that reads a same-day volume which is later finalized. Test it by comparing tonight's summary volume to Monday's re-fetched bar.
• L4 — CRDO's buy call is still status: open, behind a 09-14 "Met as written … Void" flag. This is last week's L4, unchanged.
• L5 — bars.py returns CTVA's pre-spin prices unadjusted.
  • The raw 30-day change at 10-13 is −84%. The engine nulls windows that cross the 10-01 distribution; "the one reader" does not.
  • This has the same shape as last week's L2 (ATRO).
  • It made this check special-case CTVA to reproduce Food Security's RS of −13.0. Without the special case the result is −13.34.
• L6 — A third 10-09 event entered with the scans.
  • research/stock-events/2026-10-09-6506.t-earnings.md entered the corpus in the same EOW commit as the scans that say the 10-09 calendar lists only DAL and ORCL.
  • The order is unknown, so this is counted as unverifiable. It has the same shape as row B.

UNVERIFIABLE (178), by reason

• Overwritten in place, no snapshot (A746): 60.
  • Convergence leaders: 51. signal-convergence.json is regenerated daily.
  • News-receipt dates: 9. Later pulls rewrote news/latest.json for the named names.
• Filing rows: 19. These are the 12 "NEW" filings, the 188-more and 1,254-coverage counts, and four 8-K dates in the insight wire. The on-disk EDGAR submission snapshots predate 10-13, and filing-watch.json is overwritten. PACB's 10-08 Form 4 does match.
• Ragged universe at build time: 21.
  • At build time the brief's universe had 18 rows still dated 10-12 and 15 undated. It has since grown from 1,231 to 1,290 symbols with the 10-14/10-15 monster-discovery adds.
  • Affected figures: breadth (493/697/21 of 1,211), coverage 2,818/2,851, row-date counts, the derivative count, and the AI Bottlenecks lane (−0.7%, 76/199; it recomputes to −0.8%, 79/231) with its −0.9% basket.
• Volume re-stamped (bars rewritten 2026-10-16): 11. CIFR 3.87x (three places) and JNJ 2.41x (two) in the brief, and six in the scans (see L3).
• Revised series: 1. NFCI −0.49 "as of 2026-10-02" now reads −0.507; the Chicago Fed revises NFCI history weekly.
• Corpus timing in the scans: 8.
  • News-receipt dates: food's "pulled October 8" set and "DE/FMC early September" (DE's was refreshed 10-09 16:19 PT, carrying nothing on the move); optical's "October 8 or earlier"; macro's three receipt statements. Later pulls have since rewritten these files.
  • "The 10-09 calendar lists only DAL and ORCL", in food and optical (L6).
• External event facts, not in L1: 58. These are attributed to news, wire or company sources:
  • the September CPI expectations (0.6%/3.6%/3.4%);
  • Cipher's $850M 0.00% 2032 convertible and SentinelOne's $1.25B 2031 notes;
  • Volta's $5B deal;
  • UNH's $20.50 guide (from $19.50–$20.00);
  • JNJ's $0.09 GAAP EPS and the $11.10/$11.29 single-source figures;
  • Nurix's $673.1M, $4.73, $559.5M and runway into 2029, and the ~$0.60 consensus;
  • Paradis's 95%;
  • Santoli's 16% drawdown;
  • "16 months" and "since July";
  • Citi's $28 target and BAC's $1.2B fine;
  • FRMI "pre-revenue";
  • OpenAI's ~$50B run-rate;
  • Aehr's $6M orders;
  • "all 19 officials", the October 25 runoff, the 17-month euro low, 39 tonnes of central-bank gold, 21% heating-oil bills and the October 14 CPI date.

Judgment calls counted as MATCH:

• macro :186 "the 10-year above 5% since mid-September". DGS10 first reached 5.00 on 09-15, dipped to 4.94–4.96 three times through 09-22, and has been above 5 on every observation since 09-23.
• macro :96 the dollar eased "from just over 70": RSI was 72.6 on 10-02.

Excluded from the count, and said here so the scope is plain:

• "From the Feeds" (curated external picks, attributed quotes);
• Tripwires, provenance and overdue cadences (internal state);
• the Exit-Check Advisories (paper-book positions; out of scope and book-blind).

Verdict + reasoning

Where the errors are. For a second week, the price math holds:

• every machine table and every bound cell re-derives;
• so does every lane statistic, rotation figure and desk-call return;
• so does nearly every agent-typed tape figure, across about 2,400 matches.

The escapes are both about what a page says is or is not on file:

• an identity bug in a generated section, which turns a Frankfurt listing into Deere (A);
• a timing failure in typed absence claims, which outlived the evening ingest and contradicted the same night's brief (B).

Row B is the D035 class again. The difference is that the answer arrived after the sentence was written, not before.

What is not wrong. Six inspected anomalies did not hold up:

• OXY's revenue acceleration is right on restated comparatives;
• NBIS's $219.70 is a carried engine shelf;
• Food Security's RS is right once CTVA's spin is nulled;
• the 90% lane counts are right under the n≥5 rule;
• C, JNJ and WFC's surprise percentages come from the provider's unrounded estimates;
• DE's 10-09 news pull carries nothing on the move.

What has improved since last week.

• Row D's invariant held in this sample:
  • the bound tables carry a latestDate label or render the artifact date's values;
  • the brief's gauge table stayed pinned to 10-13 after today's regeneration.
• Row E's basis label on the STX call is fixed.

Next step. Recommend filing rows A and B in DEFECTS.md before any fix, and latent findings L1–L6 in TASKS-AUDIT. This note edits neither file.

Sources

• L1 bars: research/market-engine/data/stocks/<SYM>/ohlc/max.json via research/market-engine/python/src/bars.py (rewritten 2026-10-16T21:00Z).
• Filings: research/market-engine/data/stocks/{CRDO,STX,MTDR,OXY,CRC,CVX,MUR,KLIC,KYMR,WAT,HIMS,MU,ONTO,ABCL}/edgar/company-facts.json; …/<SYM>/edgar/submissions.json.
• Macro and state: research/market-engine/data/macro/*.json; research/market-engine/data/state/regimes/2026-10-{02,09,12,13}.json; research/market-engine/config/climate-triggers.json; research/market-engine/data/prediction-markets/*.json.
• Engine definitions mirrored:
  • research/market-engine/cli/chart/precompute-summaries.ts (calculateChangeDetail, computeMultiYearReturns, computePctFrom52wkHigh, detectMomentumBase);
  • research/market-engine/lib/indicators.ts (computeRSI, computeVolumeMetrics, computeTrend);
  • research/market-engine/lib/volume-profile.ts;
  • research/market-engine/cli/synthesis/lazarus-scan.ts (gradeSymbol);
  • research/market-engine/lib/rotation-map.ts;
  • research/build-daily-brief.ts (buildLaneTape, regimeSection);
  • research/market-engine/lib/forward-return.ts;
  • research/market-engine/lib/fundamentals-wire.ts (detectRevenueAccel).
• Shipped projections: research/api-data/events/2026-10-09-{food-security,optical-supply-chain,macro-commodities}.json (generated 2026-10-09T21:28Z), research/api-data/events/2026-10-13-market-brief.json (2026-10-16T21:08Z), research/api-data/desk-calls.json (2026-10-16T22:01Z), research/api-data/regimes.json, research/api-data/calendar.json (committed 319da94ec4, 347b5a906d).
• Mechanism A: research/build-terminal-projections.ts:691-739; research/stock-events/2026-10-15-2gb.de-earnings.md; research/market-engine/watchlists/ai-bottlenecks.json.
• Mechanism B: commit 6d07511e14; research/capture/articles/2026-10-09-bloomberg-copper-clinches-weekly-gain-as-chinese-demand-signals-streng-tmm5qht9njls00.md; research/issues/2026-10-09-market-brief.md:77.