House Mneme — 2026-08-04

desk: house-mneme date: 2026-08-04 forecasts: mneme: p_trade: 0.60 direction: long conviction: med trace: p_trade: 0.35 direction: long conviction: low

Desk Plan — House Mneme — 2026-08-04

Briefing reference: /Users/dadbot/Desktop/ClaudeBod/projects/dadbrain/Analysis/briefings/2026-08-04-pm.md PM: MNEME

Shared Market Read

• Firm-wide status: GO — no Tier-1 today. NFP Friday Aug 7 (3 sessions out) is a hard-skip day for both stacks; JOLTS at 10:00 is the first labor read before it.
• Event risk today: MEDIUM (briefing rating). No Tier-1. The intraday hinge is JOLTS Job Openings (Jun) at 10:00 ET (consensus 7.4M vs prior 7.6M) — a hot print re-ignites the hawkish narrative three sessions before NFP and is the one catalyst that can flip this tape risk-off. Secondary: Trade Balance 8:30 (LOW), Factory Orders 10:00 (LOW). The after-close cluster is the bigger single-name risk — AMD (Q2, ~54x fwd) and SPCX's first-ever quarterly report (SI ~32% of float, ~17% straddle) — expect possible afternoon drift/pin as traders position. Heavy BMO earnings (CAT, MRK, MCD, PFE, BP, DUK, CMI, MPC, APO, ROK, GILD) shape the open; CAT's blowout beat (adj EPS $8.17 vs ~$6.20, Construction +57%) is a Dow constituent and a DIA-specific open tailwind.
• Session character expected: Trending-to-mixed, news-driven churn. Futures firmer (Nasdaq +0.81%, Dow +0.51%, S&P +0.20% — QQQ leading again). Overnight narrative flipped: Iran risk reversal (Trump threatens fresh strikes; crude +2% after Monday's >5% plunge) — a sector-level whipsaw, not a regime change. Both SPY and QQQ sit above their SMA20 (two-tier resolved Monday); uptrend resuming. Expect: firm open → churn around the 10:00 print → possible afternoon drift into the AMD/SPCX positioning. The 07-24 two-tier fade failure mode is conditional on pre-market QQQ/SPY direction divergence — today both point up, so that specific risk is reduced; the new late-session noise source is after-close-positioning drift in the 14:00–16:00 execution windows.
• VIX regime: 15.59 — LOW (<18), down from 15.99 prior. Both members' best backtested regime (MNEME VIX<20 PF 2.39, TRACE VIX<20 PF 2.46). Standard sizing, normal conviction.
• Key levels: SPY prior close 757.72, prior range 749.21–758.58, SMA20 745.98 (above — day 2 of the reclaim), ATR(14) 8.88; 758.58 is the first resistance (Monday's high, record-adjacent), 749.21 first support, then SMA20. QQQ prior close 700.10, prior range 686.02–701.55, SMA20 699.86 — the ~700 round-number + SMA20 confluence is the line in the sand for tech leadership. DIA is not in the briefing price block — Dow closed at a record Monday; desk's recent Phase 3 band ~$515–524 (last trade entry 07-27 $523.31); CAT's beat argues a firm Dow open; confirm at open.
• Macro backdrop: Earnings-season Tuesday (the week's heaviest cluster) + first labor read before Friday's NFP + Iran/oil whipsaw — a firm, data-aware tape rather than a clean straight-line day.

MNEME — SPY KNN (p_trade 0.60, long, med conviction)

No hard skip today — all three windows eligible (JOLTS is not Tier-1/Tier-2).

Setup present: Likely. Same skeleton as Monday, which produced a Window C win (+$71.05): VIX LOW (my best regime), SPY above SMA20 in an uptrend, no event skip, one MEDIUM-impact 10:00 data print. The 10:00 JOLTS again complicates placement, not viability — it sits exactly on Window A's entry.

• Window A (observe 9:30–10:00, enter 10:00, hard flat 12:00): worst-placed window, same as Monday's ISM. The 6-bar observation encodes pre-JOLTS positioning plus the 8:30 Trade Balance; entry at 10:00 is at the print. If A fires, demand a decisive vote (≥ 70%) before trusting it — the 10:00–12:00 segment is a data-reaction window, the thinnest part of my pattern library.
• Window B (observe 10:00–12:00, enter 12:00, hard flat 16:00): anchor window (holdout PF 3.76) and the best-placed observation — JOLTS fully absorbed into the 24 bars. Execution 14:00–16:00 is the character-shift window; the 07-24 failure mode is conditional on pre-market QQQ/SPY divergence, and today both are green (QQQ leading harder: +0.81% vs +0.20%) — reduced risk. New nuance: semis are soft ahead of AMD's print; if SPY fades after lunch on AMD/SPCX positioning while QQQ holds, that is the afternoon-chop signal — skip B.
• Window C (observe 12:00–14:00, enter 14:00, hard flat 15:30): flagged in holdout (PF 1.25) but live-hot — four straight HARD_FLAT wins (07-23/28/31/08-03), +$284.97 combined, and Monday fired at vote 66.7% / max_sim 0.7794 — below the plan's own exceptional-strength bar — and won anyway. Today C is the cleanest-placed window of the day: observation captures post-JOLTS structure and the AMD/SPCX positioning drift; entry at 14:00 is well past both data prints; hard flat at 15:30 exits before any after-close event. The desk's exceptional-strength stance (max_sim > 0.65 AND vote > 70%) stays as the planning bar, but Monday is the standing reminder that C fires-and-wins below it — plans are non-binding; the KNN's own gates decide.
• Sizing: standard 0.75% ($187.50). No Tier-1; every window hard-flats at 12:00/16:00/15:30 — nothing is ever held across an event; per-trade risk is fixed at 0.75% by the 1.5× ATR stop. No cut for MEDIUM event risk (same stance as Monday, which worked).
• SPRT: LLR +2.855, only +0.089 from CONSISTENT-WITH-BACKTEST (+2.944) — one strong trade likely crosses it. Sequential monitor, not a sizing input; it does not change stance either way today.
• Invalidation: SPY opens gap > 1.5% either direction (extreme auction — sit out A, reassess at B); VIX > 20; a hot JOLTS (headline ≥ ~7.8M or any upside surprise the market reads hawkish) flipping the tape risk-off within the first hour → stand down B and C; QQQ/SPY divergence (QQQ gives back its gap while SPY holds, or a semis-led fade) → skip B, reassess C.

TRACE — DIA KNN (p_trade 0.35, long, low conviction)

No hard skip today — all three windows eligible.

Setup present: Possible — gate-improbable, not structurally excluded. This is not a "No trade expected" day: no hard skip, VIX LOW (best regime, PF 2.46), the Dow at record highs, and CAT's blowout beat (Dow constituent, Construction +57%) is a DIA-specific open tailwind that wasn't there Monday. The binding constraint is unchanged and yesterday added a new wrinkle:

• The near-miss detail that matters (08-03): Window C resolved at vote 70% (≥ 62%, PASS) but max_sim 0.589 vs the 0.60 floor — a 0.011 similarity miss, the first sim-bound gate failure in my Phase 3 record. The neighbors were decisive yesterday; the closest one was just under the bar. This joins the four sessions that resolved at exactly 60% vote (07-15 A, 07-17 C, 07-21 A, 07-31 C) — the standing read holds: 62% sits at my natural vote-resolution point, regime not calibration, no threshold action. The sim-side grazing is one data point, not a cluster — but it says the gates are not far away.
• Window alignment is the real story: my live record is 6 of 7 trades in Window A — and A today is again the worst-placed window (entry exactly at the 10:00 JOLTS print). If A fires, discount the 10:00–12:00 segment as data-reaction.
• Window B (post-JOLTS observation 10:00–12:00) is the cleanest window of the day for DIA; rare live (1 of 7 trades) but quality placement. Window C remains flagged (recomputed holdout PF 1.125; rolling-20-trade PF < 1.2 suspension rule live) — exceptional-strength bar only (max_sim > 0.65 AND vote > 70%). Yesterday C was my closest window (70% vote / 0.589 sim) — if it fires today it needs the sim floor to clear.
• Why DIA is again the least-bad instrument: the Dow holds one mega-cap tech name — the SPY/QQQ two-tier and the semis-soft-ahead-of-AMD dynamic barely touch its microstructure; oil +2% is a modest CVX tailwind; CAT's beat is a direct industrial read-through. VIX < 20 is my best backtested regime.
• Sizing: standard 0.75% if a genuine signal fires. No cut, no hedge — see SPRT posture.
• SPRT posture: LLR -1.350, -1.595 from DEGRADED (-2.944). A qualifying loss likely breaches DEGRADED → desk PM review within 1 session, TEMPER conversation within 5. Pre-registered; a sequential monitor is not a sizing input and not a reason to trade defensively — stated plainly so the plan does not hedge around it.
• Invalidation: hot JOLTS flipping the tape risk-off in the first hour (DIA's industrial tilt — CAT, HON, MMM — takes it hardest among our instruments) → stand down B and C; SPY opens gap > 1.5% (my pattern library is thin at index extremes) → sit out the morning, reassess at B; VIX > 20.

Pooled note (Family 5)

If both stacks fire today, same-family same-day longs count as one clustered observation for significance (desk rule). Each sizes independently; there is no shared position budget, and one stack's fire is not evidence for the other's — a strong SPY setup implies nothing about DIA's independent signal.

GHOST — IWM (FROZEN — Phase 2, Decision D2)

Skipped entirely. No plan, no forecast, no plan file.

Plan Filed

• Filed: 2026-08-04 07:35 ET
• Frontmatter forecasts complete for every active member: yes
• GHOST: FROZEN (Phase 2) — not planned for
• Firm-wide status: GO

House Mneme Desk Reflection — 2026-08-04

Plan reference: desks/house-mneme/plans/2026-08-04-plan.md (period also covers 2026-08-03-plan.md) EOD briefing: /Users/dadbot/Desktop/ClaudeBod/projects/dadbrain/Analysis/briefings/2026-08-04-eod.md

What Happened

Period since last reflection (2026-07-31 → 2026-08-04)

Combined desk cumulative (Phase 3): +$469.11 across 21 trades (14 MNEME, 7 TRACE) — up from -$130.16 at the 07-31 reflection. First positive combined level since the desk pooled.

MNEME — traded both sessions, won both, and crossed its SPRT decision boundary

MNEME's binding gate this period was nothing — both non-event sessions produced qualifying signals with strong similarity (0.7794, 0.8238) and decisive votes (66.7%, 73.3%), and both won. 08-03: Window C long 757.64 (vote 66.7% / max_sim 0.7794), HARD_FLAT +$71.05 — notably below the desk plan's exceptional-strength bar (max_sim > 0.65 AND vote > 70%) yet won again, the standing reminder that plans are non-binding and the KNN's own gates decide. 08-04: Window A — the window the plan keeps calling "worst-placed" on data-print days — fired at 10:00 with vote 73.3% (≥ the plan's own 70% decisive-vote demand) and max_sim 0.8238; JOLTS printed soft, the tape rallied through the data block, and the 10:00–12:00 segment was a clean continuation leg on a 1.63× ATR record-setting day. HARD_FLAT 768.77 at 12:00 for +$528.22 — the biggest single trade of Phase 3, at standard 0.75% sizing. The plan's decisive-vote stance on A was validated, not violated.

Milestone: SPRT crossed CONSISTENT-WITH-BACKTEST at trade #14 (LLR +3.494 vs +2.944) — the sticky first crossing, the firm's first agent to reach a decision boundary. The plan forecast it ("one strong trade likely crosses it"); the largest trade of Phase 3 delivered it. The monitor continues informationally; this declares the live record statistically consistent with the cleared v5 backtest and changes nothing about gates or sizing.

Cumulative: 14 trades, 8W/6L (57.1%), +$818.32 net — the desk's profit engine.

TRACE — vote gate binding again; first-ever sim-side graze adds a wrinkle; verdict unchanged: regime, not calibration

TRACE logged no qualifying trade either session. The load-bearing question — binding gate, and regime or calibration? — has a clean but evolving answer:

• 08-04 (today): the vote gate was binding on all three windows. A 60%, B 40%, C 60% — all below the 62% floor — while similarity passed every window by wide margins (0.902 / 0.682 / 0.901 vs the 0.60 floor). Both A and C resolved at exactly 60% — the fifth distinct session at exactly 60% (07-15 A, 07-17 C, 07-21 A, 07-31 C, 08-04 A+C), never 61.
• 08-03 (Mon): the sim gate was binding on C — vote 70% (PASS) but max_sim 0.589 vs the 0.60 floor, a 0.011 miss. This is TRACE's first sim-side gate failure in Phase 3, and the first window where the vote passed while similarity failed (every prior session was the reverse: sim passed, vote failed). The desk plan's own 08-03 note — "C needs the sim floor to clear" — was prescient.
• Verdict: regime, not calibration — unchanged. The vote-side near-miss signature is a cluster (five sessions resolving at exactly 60%, never 61, with healthy sim margins) — the signature of 62% sitting at TRACE's natural vote-resolution point. The sim-side is a single 0.011 graze — one data point, not a cluster. New watch item: a second sim-side graze within a few sessions would turn the sim threshold into a repeated near-miss and reopen the regime-vs-calibration question on that gate. No threshold action on either gate; the pre-registered framework is the only mechanism.

SPRT: LLR -1.350 (n=7, W1/L6) — unchanged, no qualifying trade. -1.595 from DEGRADED (-2.944); the pre-registered DEGRADED protocol (desk PM review within 1 session, TEMPER conversation within 5) remains one qualifying loss away. The desk's most urgent risk flag carries forward.

Plan calibration since last reflection (WS4b rows)

\* Would-be values — pipeline gap flagged again: intelligence/plan-calibration.csv has no rows for 08-03/08-04 as of filing, the same gap class flagged on 07-31 for 07-28/07-29. The desk filed complete forecasts both days; plan scoring is not keeping up with the daily cadence. Worth a look by the calibration owner.

Direction accuracy: the desk has now forecast a direction on every traded day of Phase 3 and hit every one — 08-03 and 08-04 included (long, long, both correct).

Desk-level observations

• No same-day clustered observation this period — MNEME traded alone on both 08-03 and 08-04, TRACE never fired — so the pooled Family 5 record adds two independent observations, both wins.
• MNEME's Window A live record is 2W/0L (07-16 +$96.03, 08-04 +$528.22) — including today's data-print-day win, which validated the plan's "demand ≥ 70% vote before trusting A on a 10:00-print day" stance. Window A holdout PF 2.64 is the second-best window; live it is holding.
• MNEME's flagged Window C streak extended: four straight HARD_FLAT wins (07-23/28/31, 08-03), +$284.97 combined; since 07-20, C is 4W/2L (+$160.89). The holdout flag (PF 1.25 / 31.2% WR) is looking stale at n=6, but the pre-registered rolling 15-trade PF < 1.2 suspension rule remains the only mechanism — no change.
• Structure under-call, favorable bias: both desk plans expected churn/rotation ("trending-to-mixed, news-driven churn" on 08-04; "fade-and-recover" on 08-03); both sessions were clean trend-ups (1.33×, 1.63× ATR). The desk has now under-called trend structure two sessions running — but the bias direction (up) favored this long-only desk: MNEME won both. Worth logging as a plan-calibration observation: in a no-Tier-1 NFP build-up week, the plans keep over-weighting 10:00 data prints as potential tape-flippers; both printed benign and the tape trended.
• Plan accuracy by member (period): MNEME 4/5 (event risk and setup nailed; character call wrong); TRACE 4/5 (event risk and the "gate-improbable" read nailed; character call wrong, moot).

Paper P&L

Combined desk P&L today: +$528.22 MNEME cumulative Phase 3: 14 trades, +$818.32 net, 57.1% WR, SPRT CONSISTENT-WITH-BACKTEST (LLR +3.494, declared at trade #14, sticky) TRACE cumulative Phase 3: 7 trades, -$349.21 net, 14.3% WR, SPRT CONTINUE (LLR -1.350, -1.595 to DEGRADED) Combined desk cumulative: 21 trades, +$469.11 net, 42.9% WR

Event Risk vs. Expectation

• 08-03 (LOW rated): correct. ISM headline beat hard (55.6 vs 54.0, highest since May 2022) but Prices Paid cooled (71.1 vs 73.0, below consensus) — the flagged inflation risk did not materialize; the tape rallied through the print. MNEME's C long entered at 14:00, well after the 10:00 data dust settled.
• 08-04 (MEDIUM rated): correct. No Tier-1. JOLTS (7.359M vs 7.44M) printed soft — the flagged "hot JOLTS flips the tape risk-off" scenario did not materialize; Factory Orders missed but was LOW as rated. The after-close AMD/SPCX cluster was the day's real event risk — both beat, both reversed after hours (-7%/-6%) — but it never touched this desk: MNEME hard-flatted at 12:00 and TRACE held no position. The plan's flagged afternoon positioning drift was the only wrinkle (marginal -0.13%/-0.21% last-hour fade) and did not dent the trend.
• Invalidations: none triggered for either member on either session — no >1.5% gap either direction, VIX 15.86 → 16.50 (never >20), no first-hour risk-off flip.
• Session character: both sessions were "Trending-up" per the EOD briefings; both plans called for churn. The structure calls under-called again (third straight for the AM, second for the desk's own plans) — direction right, cleanliness under-estimated. Favorable for a long-only desk, but the character read needs calibration.
• Forward risk (next sessions): Wed 08-05 ADP (8:15) + ISM Services (10:00, consensus ~54.5) — mid-tier, no skip; Thu 08-06 SPCX lockup expiration (up to 20% of float) — single-name, no skip; Fri 08-07 NFP (Tier-1) — hard skip for both stacks by pre-registered rule. VIX 16.50 keeps both stacks in their best backtested regime (VIX < 20: MNEME PF 2.39, TRACE PF 2.46).

Reflection Filed

• Filed: 2026-08-04 19:05 ET
• Next session event risk (from EOD briefing): MEDIUM — ADP (8:15) + ISM Services (10:00), Wed Aug 5; NFP Fri Aug 7 (Tier-1, hard skip for both members).
• GHOST: frozen (Phase 2, Decision D2) — no reflection written.
• MNEME and TRACE per-member reflections filed.
• Flags embedded for learnings.md (desk-blind, own members): MNEME — SPRT declared CONSISTENT-WITH-BACKTEST at trade #14 (LLR +3.494, sticky first crossing; firm's first decision boundary), Window A data-print-day win validated the ≥70% decisive-vote stance, Window C extended to four straight HARD_FLAT wins; TRACE — vote-gate clustering at exactly 60% now spans five distinct sessions, plus the first-ever sim-side graze (08-03 C, 0.589 vs 0.60) with vote passing — one data point, sim-side now a watch item; verdict regime, not calibration, unchanged. Also flagged: plan-calibration.csv pipeline gap (08-03/08-04 rows missing, same class as 07-28/07-29) and the desk's two-session trend-structure under-call.