🎛 Mission Control

The home base for tracking and grading forecasts. This Week's Picks are your own selected tickers, tracked with your own directional hypothesis. Blind is a random control group the engine predicts on its own, zero human input — so Picks' accuracy can be judged against a genuine baseline, not just against "feeling right."

AUDIT 🧭 Scanner Audit 📜 History 📊 Insights 📝 Tuning Notes

ENGINE 📏 Calibration ⊕ ∇P Training

PILLARS ⚛️ Physics 🏥 Fund Health 📊 Indicators 🎯 Probability 🌐 Global QA

+ Add

+ Add

↻ Refresh

Full technical release history, admin-only. The public About modal shows a trimmed bio and headline-only recent highlights — the formulas, weights, and guard counts below stay internal.

Origin story (full)

Every line of code was engineered in partnership with Claude (Anthropic) — designing the architecture, writing the logic, debugging, and iterating in real time through natural conversation. No traditional development team. No agency. Just a trader who dabbles in math and physics, with a vision — and an AI that could build it. Brian serves as lead product designer, domain expert, and beta tester: driving every feature decision, validating the physics models against real market behavior, and stress-testing every release before it ships. Version 3.0 introduced the Market Physics engine — applying Hamiltonian mechanics, wave-particle duality, and quantum orbital theory to market structure. RC 1.0 marked the first feature-complete release. RC 2.0 deepened the data layer, added the solo price chart canvas, and expanded the Desktop companion app. RC 3.0 continued to push the platform forward — adding free SEC EDGAR Form 4 insider scanning, a Super Investors hub for all major fund managers, EMA overlays on the price chart, a redesigned Gov Trades interface, and deeper Electron sidebar integration including cross-app ticker sync. RC 3.5 added five Claude AI features across the app. RC 3.75 enriched the earnings calendar and added the Admin Roadmap. RC 4.5 introduced a centralized P&L engine, admin-only error monitoring, a full suite of data-integrity fixes, and a complete 2D/3D/4D Market Physics visual overhaul. RC 5.5 extends the unified physics conviction engine to Mid Cap, Large Cap, and 2:30 PM scans — all scanners now run the same x/12 alignment system with proper delete/star/NEW/STALE logic, align-based scoring replacing the micro-cap unicorn score, and decimal precision across all score displays. RC 6 focuses on UX polish and new data layers — intraday charting, smart physics pre-filtering, Schwab portfolio redesign, sector rotation in Morning Brief, mindset stat boxes in the Trade Log, and a full screen header cleanup across all screens. RC 7.0 introduced the ∇²H Hessian Wave Field — a live retina-sharp canvas of Hessian curvature across phase space, integrated directly into the Probability tab alongside the NEP forecast. API sessionStorage caching eliminated redundant startup calls. RC 8.0 completes the Physics Canvas Unified Overlay system — every canvas now carries the −∇H restoring force arrow, phase angle φ arc, von Neumann entropy label, and Wigner W⁺ squeezed state, all data-driven and hover-annotated. Navigation cleanup promoted Scenario and BT to top-level views. RC 6.25 introduces the Heim 12-Dimensional Probability Framework — a unified quantum field model that reads the market across 9 active dimensions before price moves manifest. Ψ₅₋₆ measures organizational coherence (Fund Health x/14: Altman Z, Piotroski, Revenue Growth, Short Float, FCF Growth). Ψ₇₋₈ reads the information field — archetype classification (DISRUPTION · GROWTH · VALUE · TURNAROUND · DISTRESS), analyst consensus crystallization, and PT dispersion coherence. Ψ₉ encodes the trader’s own intention field — Schwab position, Trade Log history, watchlist engagement, journal depth, emotion quality (7 states with quality weighting), and setup source. All four tiers collapse into a single Heim 12D Probability on every Deep Dive. Quantum eigenstates ψₙ (ψ₀–ψ₄) surface Hermite polynomial wavefunction nodes as physics-derived S/R levels with excitation and relaxation price targets. The 3-stage Smart Scan pipeline completes the scanning workflow: Gainers/Active → Quick Filter → Heim Physics → Top 3 by Heim 12D P. Scan cards gain the Ω Heim strip and dynamic hover-overs across all indicators. RC 8.0 introduces the 🌅 9:45 AM Morning Scan — a dedicated opening-range scanner with 13 criteria tuned for gap confirmation, holding open, and early volume conviction, feeding the same scan → candidates → watchlist → Physics pipeline. Physics Alignment Filter is now fully configurable in Settings (default 5/12) via a new ⚛️ Physics Alignment Filter section, with a system-wide getMinAlignScore() replacing all hardcoded thresholds. Restore Defaults button added to the Settings footer. Schwab credentials moved to ~/.schwab/ outside the repo for improved security. RC 8.25 was engineered with Claude Fable 5, Anthropic's latest model, in a single working session: the ⏰ Timed Scans tab unifies the 9:45 AM and 2:30 PM scanners, a batch pre-filter cuts Long-Term and ETF scan API usage by ~85%, AI Briefing extends to every scan tab with per-scan analyst framing, and the Deep Dive gets cleaner — Synthesis and Heim 12D now live inside Physics Details, standard active-pill states throughout, and hover tooltips restored across the overview card and every physics canvas. RC 8.5 is the mathematics release — a full audit of the non-Euclidean probability engine with Claude Fable 5. The H² heat kernel recovered the 25–35% of probability mass a numerical singularity was silently losing, and the missing 2π of the hyperbolic area element was restored — NEP bands now carry their honest stated coverage. On top of the corrected foundation, three genuine pieces of differential geometry: SABR leverage skew (ρ̂ measured per ticker from Parkinson high/low vol — bands widen below, tighten above, the real equity skew), measured curvature (κ̂ = ν̂² from vol-of-vol replaces the hand-set slider), and the showpiece — the hyperbolic Ornstein–Uhlenbeck bound/free phase transition: the spring constant now lives inside the probability geometry itself, with curvature pushing the particle out and the spring pulling it in. Bound stocks' probability bands stop widening; free particles keep the full hyperbolic fan. Escape radius d* = κ̂/λ on every Deep Dive. RC 8.55 renders the new geometry and hardens the pipeline: the Probability chart gains a probability fog — the full radial distribution drawn as a density fan with the ρ̂-skew and spring saturation visible — plus d* spring-horizon lines and a BOUND/FREE phase badge; the 2D orbit gets a curvature-vs-spring net radial force gauge and d* escape ring (mirrored on the 3D LIVE plane); the ℳ GEOMETRY strip joins Ω₇/₈ and Ω₉ on the Deep Dive with θ, d*, ρ̂ and κ̂ measured at fetch time; and a deployment audit fixed a redirect loop, repointed releases at the true production environment, and removed legacy deployments. RC 9.75 closes the loop from physics to money: the 📏 Calibration Ledger auto-logs every directional forecast and grades it against what the market actually did — Brier score and predicted-vs-observed buckets turn "the physics feels right" into measured evidence — and the 💰 Kelly Position Sizer converts the audited probability and the ρ̂-skewed payoff ratio into a risk-bounded share count, with fractional Kelly, a hard 25% account cap, and the liquidity boundary rule (shares ≤ 1% of average daily volume) built in. Measured geometry → calibrated probability → position size: the platform now reasons all the way from curvature to share count. The final wave closed the circle on Brian's own coursework: 📐 Random Matrix factor analysis (Marchenko–Pastur eigenvalues — how many REAL factors move the watchlist today, with a diversification-illusion warning), ⇄ transfer-entropy information flow (who leads whom, in bits), the ⏳ ergodicity check on every position (time-average growth vs ensemble EV — positive expectation can still shrink an account), the 🌡 κ̂ regime thermometer, exact first-passage (absorbing-boundary Fokker–Planck replacing the last approximation in the engine), the ⚖ Geometry-vs-Options gauge (VIX implied vs measured vol), ⚛ phase-transition alerts reaching native macOS through the Electron sidebar — with 🔔 Phase Watch polling the watchlist on a timer — and the science record made durable: cloud-synced, exportable, and backed up three ways. "The market is finite, random — but we are finding a hidden order from objective data." The mathematics that found it now stands guard over it. RC 9.85 turned the platform into a daily companion — the 🦄 Today strip spine, the one-tap Weekly Close ritual, and a unified All · Stocks · ETFs · Crypto watchlist index. RC 9.9 made the science record bulletproof: every cloud-sync table was audited and repaired with its schema captured as code so it can never silently drift again, stock dividends now fold into a position's total return, and the Trade Log was polished so nothing hides off-screen. RC 10.0 brings the Mindset foundation to life — a full Academy of evidence-grounded mindset tools and the Biological Communication Filter, whose Filter Check measures the clarity of the observer and stamps it onto the Calibration Ledger — and locks the physics engine itself behind a golden-vector guard so the math can never drift unseen. RC 10.5 opens the Academy's classroom — guided lessons for Trading Overview and the 6 Mindstates, each card check-tracked, gated True/False quizzes that unlock once a lesson is complete, and a Walkthrough tour — over the new Mind Engine, the 4th pillar's own behavioral math (Markov state-chain, calibration-by-clarity, bias fingerprint) standing behind its own golden-vector guard like the physics. RC 10.75 brings the Academy's media to life — every Mindspace video and Sound-Frequency track now plays from a clean tap-to-play thumbnail card, the Frequencies room is reorganized into two collapsible columns (Sound & Solfeggio · Brain Waves), and three new TEDx talks join Goals, Affirmations and Mindset. RC 10.85 adds the ⚡ Day Trading scan — a dedicated penny day-trade scanner (Regular · Premarket · 2:45 Bell Close · Individual) built as a fully self-contained, walled-off module that reads only shared plumbing and never touches another scanner's state. A staged, call-efficient pipeline (a free batch-quote gate, then per-symbol float/RSI/liquidity, with VWAP computed last on only the finalists) screens sub-$5 low-float momentum names, and a weighted Day-Trade x/12 score — VWAP, volume and float weighted heaviest, the way intraday trading actually rewards them — ranks them behind a 60% qualifying cutoff, each result diving into the same Deep Dive every scan shares. RC 11.5 took the platform server-side — FMP and Anthropic keys moved out of the browser and behind a Supabase edge-function proxy, and a runaway API-call loop was traced and sealed — while sharpening the daily tools: the Day Trading scan gained a single-line card and the ⚡ Ham Phys opening-range sweep read, and the admin Categories view became a sector heat-grid (Stock Themes · ETF Categories) where today's move, relative strength vs SPY, and the 5-day trend make a hot sector readable at a glance. RC 12.0 adds the Gen-2 Real-Yield Scanner to the Crypto tab — a Saylor scan (CoinGecko top movers by volume and 24h move, 4h STALE) and a Real-Yield scan scoring seven DeFi tokens /12 via DefiLlama cash-flow data (HYPE · DYDX · INJ · GMX · GNS · AAVE · MKR) with regime reads (COMPOUNDING · NEUTRAL · DILUTING), localStorage persistence, and automatic calibration-ledger logging on every scan. RC 12.1 adds the ∇P Gradient Probability Engine — a physics-informed logistic model trained browser-side on five years of watchlist OHLCV: six learned weights (φ phase proximity · ρ̂ leverage skew · align momentum · θ± signed regime · σ volatility) computed via Brier-loss gradient descent on 60-day rolling windows with 80/20 walk-forward validation, blessed weights stored in localStorage, and a live ∇P probability strip in every Deep Dive below ℳ GEOMETRY with a divergence alert when ∇P and Heim 12D disagree by more than 15%. RC 12.1 also adds watchlist source tagging — every ticker added from a scan card is stamped with its origin (2:30 PM · 9:45 AM · Day Trade · Mid Cap · Large Cap · Long Term · Stocks · Gov Trades · Manual) and displayed as a color-coded pill on the watchlist card. RC 12.5 completes the ∇P engine with Phase 4 Continuous Learning — each graded Calibration Ledger entry with cached features triggers one Brier gradient step on a live weight copy separate from blessed weights; a full sweep of the 23K-sample Phase 1 historical dataset runs on demand; drift is reviewed before a human+AI Promote copies live weights to blessed — the model improves with every graded trade while physics governance stays intact. RC 12.5 also hardens the Admin and Settings modals — both the header and tab bar now stay correctly pinned when scrolling (Safari sticky fix via overflow-x:clip), and the Admin header is fully opaque so content cannot bleed through. RC 12.75 is the day-trade release — a full methodology overhaul of the ⚡ Day Trading scan, co-worked from a line-by-line code review. RSI moved to the intraday 5-minute series and became a scored factor rather than a hard gate, so momentum runners are no longer filtered out; missing-data factors now earn nothing and drop from the /12 denominator, so the score reflects only what is actually known; relative volume is time-of-day adjusted (today's pace projected to a full day); the VWAP-hold gate fails closed; and the $1-reclaim factor is proximity-weighted for sub-$1 names with a prior >$1 print. Every qualified finalist is auto-logged to the Calibration Ledger as a one-day forecast. The universe went wide — the FMP company-screener now feeds a ~300-name penny universe across all whole-market modes, behind a Wide day-trade universe setting that conserves API on smaller plans. Bell Close became a genuine power-hour scan (breakaway relative volume, EMA 9/21 trend, and a MACD + EMA200 Drive confirmation); a CMF conflict resolver stands the trader aside when continuation and fade signals disagree and money flow cannot break the tie; and a 🎯 Today's Pick card surfaces the single best setup — or says sit out — for the one-trade-a-day discipline. Built in collaboration with Claude Opus 4.8. July 2026. RC 12.8 sharpens the crypto scanner and the research cards. The Gen-2 Saylor score is now normalized to the factors actually available, so a coin whose momentum data was rate-limited to null is no longer silently deflated against one that fetched cleanly — a ⚠ partial badge marks the difference, and gentler CoinGecko batching means fewer coins go partial at all. On every candidate card's Research tab, the chart pill is now 🕯 Graph, the on-demand quality read is renamed 📊 Signals to match the Signals scan, and a new 📈 Trending pill reads the ticker's multi-timeframe momentum on click — all research, no API until you ask for it. Built in collaboration with Claude Opus 4.8. July 2026. RC 12.9 hardens the account layer and fixes a call-volume spike. Firestore access for account registration, activity logging, beta-expiry, and session enforcement — previously blocked outright by a deny-all rule — now runs through a proper Firebase custom-token bridge tied to each user's real Supabase identity, with per-user Firestore rules restricting every record to its owner. Separately, an FMP call-volume spike was traced to a gap in the RC 11.1 startApp re-entry guard: a sign-out/sign-in cycle was silently stacking a second full set of polling intervals on top of the first, compounding every cycle — now fixed by clearing each timer before it's re-registered, the same pattern already used safely elsewhere in the app. A second, compounding contributor was found and fixed the same way: the Trade Log's live-price fetch was pulling a quote for every symbol ever traded, open or closed, instead of just current positions — inflating a 300+-symbol request that failed FMP's batch-quote endpoint outright and fell back to hundreds of individual per-symbol calls every cycle. Built in collaboration with Claude Sonnet 5. July 2026. RC 13.5 adds a Watchlist Clear button (with confirmation) and reworks the UCC Lab (formerly Physics Lab) admin tool: the Schwab section is now an open-ended Picks list instead of a fixed 5, and the Blind test group is now genuinely bias-free — a random 5-ticker sample and its directional hypothesis are both derived entirely from the engine's own data, with zero manual ticker or prediction input. Day Trade's Today's Pick card gained a one-click 🔬 UCC Lab pill to send the day's top setup straight into a Pick slot. Built in collaboration with Claude Sonnet 5. July 2026. RC 13.55 adds a third UCC Lab section, 🎯 #1 Suggested — auto-populated only by a scan's "UCC Lab" pill (readonly, no manual ticker or hypothesis entry), so a scan's own top recommendation gets tracked and graded separately from both your judgment (Picks) and the random control group (Blind). All three sections are now mutually exclusive — a ticker already tracked in one won't be duplicated into another the same week. Built in collaboration with Claude Sonnet 5. July 2026. RC 13.6 reorganizes QA inside UCC Lab into 🎛 Mission Control (one shared ticker list, tested identically across every pillar) and ⚛️ Physics Pillar (the existing blind Claude review, now reading from that shared list). Mission Control lays the groundwork for testing all 4 Deep Dive pillars — Fund Health, Indicators, and Probability pillars are planned next — plus a future Global QA pass that flags where pillars disagree on the same ticker; both are visible now as inactive placeholders. Built in collaboration with Claude Sonnet 5. July 2026. RC 13.65 merges Mission Control's ticker-list content into the ⚛️ Physics tab for visual beta-testing before the final rename — the Mission Control tab remains in place untouched as a placeholder pointing to Physics, nothing removed yet. Built in collaboration with Claude Sonnet 5. July 2026. RC 13.67 completes Mission Control's redesign — the merged content is now organized into five mutually-exclusive pills (Scanner Audit, History, Insights, Tuning Notes, Config), the Schwab quick-scan box was removed as redundant with Auto-Fill Schwab, This Week's Picks and Blind now carry a plain-language overview and a rationale blurb explaining why Blind exists as a control group, Picks rows gained a ⋮ actions menu (Modify / Delete) matching the Trade Log's pattern, and a working 🧹 Clear button resets the screen alongside an inert 📤 Commit to Supabase placeholder for a future bulk-save. Built in collaboration with Claude Sonnet 5. July 2026. RC 13.68 retires Group A/B and the old Claude-blind-review QA test entirely — weaker than the two validation layers already in place (the golden-vector guard for math correctness, and Picks/Blind/Suggested's real graded backtest), so it no longer earned its place. The Config pill and the deprecated "Old QA" tab are both removed; QA Pillars and Global QA live on as inert placeholders inside Scanner Audit; and Physics Pillar is now a plain read-only placeholder, consistent with the still-unbuilt Fund Health, Indicators, and Probability pillars. Built in collaboration with Claude Sonnet 5. July 2026. RC 14.0F brings the Probability Insights pillar under the golden-vector guard — the Track B Bayesian posterior, signal-entropy confidence, phase gate, the Heim 12D full Ψ chain, the Fund Health x/14 formula, and the ∇P feature math are now pure shared functions frozen behind 30 blessed golden vectors (up from 12), so no model or session can drift the pillar math unseen. The Deep Dive card, the Smart Scan ranking, and the coming UCC Lab 4-pillar QA all read the same formulas — one source of truth from screen to test. Built in collaboration with Claude Fable 5. July 2026. RC 14.1F delivers that QA: UCC Lab's Scanner Audit now captures all four Deep Dive pillars — one Deep-Dive-grade fetch per ticker snapshots Physics Conviction, Fund Health, Indicators Alignment, and the Probability posterior (with its observer context: prior, BCF weight, phase, ∇P weight provenance) through the very functions the Deep Dive card calls, and Evaluate grades every pillar's directional read against the actual price move — a pillar that made no claim scores ⊘ No Signal and stays out of the accuracy math. Sections run 🎯 #1 Suggested · This Week's Picks · Blind; a wrong-way reversal no longer grades a momentum call correct; mutual exclusion is enforced in the database, not just on screen; and the physics_lab schema is captured as a migration so it can never silently vanish again. Built in collaboration with Claude Fable 5. July 2026. RC 14.2F makes Mission Control true to its name: the 📏 Calibration Ledger and ⊕ ∇P Training moved from the Admin tab bar into Mission Control as pills beside Tuning Notes — the full loop of testing the calls, grading the engine, and retraining it now runs from one room. Early Check became preview-only (grades commit only at target date), the ∇P panel's feature list was corrected to all 8 features, and the shared Physics Conviction formula joined the golden guard — 32 blessed vectors standing watch. Built in collaboration with Claude Fable 5. July 2026. RC 14.25F completes the consolidation: the Lab's sub-tab bar is retired and UCC Lab simply is Mission Control — three labeled pill groups (AUDIT · ENGINE · PILLARS) holding the weekly call-grading, the engine tools, and each pillar's coming read-only track record in one room. A design decided together, as partners. Built in collaboration with Claude Fable 5. July 2026. RC 14.3F lights up the first of them: the ⚛️ Physics Pillar dashboard — hit rate, pending count, every capture's conviction and particle call graded against the actual price move, and a conviction-vs-outcome read that asks the honest question: do winners really carry higher conviction at entry than losers? The other three pillars follow the same pattern. Built in collaboration with Claude Fable 5. July 2026. RC 14.35F finishes the set: Fund Health, Indicators, and Probability dashboards go live on one shared renderer — each pillar's directional calls graded against the market, no-claim reads honestly excluded, and every Probability row carrying the observer context it was computed under. Four pillars, four track records, one Mission Control. Built in collaboration with Claude Fable 5. July 2026. RC 14.4F points Aumann at the market itself: the Aumann gap now measures your posterior against an options-flow-implied probability — real unusual-options premium, dollar-weighted and sentiment-aware, shrunk toward ½ so thin flow can't shout — with both gaps, market and engine, on every Deep Dive and stored in every QA capture. The golden guard grows to 35 blessed vectors. Built in collaboration with Claude Fable 5. July 2026. RC 14.75F builds the business backbone: a Stripe subscription backend — hosted checkout and a billing portal behind Supabase edge functions, a signature-verified webhook (fails closed) mirroring every subscription's life into a service-role-only table, and a server-side paywall gate on the keyed API services that ships dormant behind a single secret, admin-bypassed and ready to flip at launch. Card data never touches UCC — only Stripe's hosted page ever sees it. Proven live end-to-end in both directions: payment grants access, cancellation revokes it. Built in collaboration with Claude Fable 5. July 2026. RC 14.8F puts the switch in the room: a 💳 Stripe section in Admin → 🔑 Keys holds both the paywall toggle — authorized by the same publish secret as ∇P, stored server-side in app_config, arming or disarming enforcement within 60 seconds, no terminal required — and a subscription status card that opens Stripe's hosted checkout or billing portal in one tap (it moves to user-facing Settings at launch). Every proxy call answers a paywall block with a plain "Subscription required." RC 14.85F is the first fix from the co-reviewed physics & probability audit: Fund Health stops crediting missing data — the x/14 score is renormalized to the factors actually known (⚠ marks a partial read), under half the scale known honestly reads NO DATA and abstains from the Heim chain, and the QA ledger grades those as no-claim. A blessed co-decision replacing the original midpoint-credit design; the golden guard grew to carry the new abstention vectors, 36/36 green. Built in collaboration with Claude Fable 5. July 2026. RC 14.86F lands audit fixes #4 and #5 in tandem: the Probability pillar's uncertainty read now normalizes cross-signal disagreement by its exact bound (mean·(1−mean), not the flat 0.25 that under-penalized conflict among lopsided signals), and the Aumann flow comparator never averages dollars with contract counts — premium-weighting wins whenever any premium exists, volume only as a whole-set fallback. Co-decided, re-blessed, 36/36 green. Built in collaboration with Claude Fable 5. July 2026. RC 14.87F closes the audit's read-only pair with the 📐 Model Audit card in Mission Control's Insights: stated-probability-vs-reality coverage buckets with the posterior's running Brier score, and a fusion-constants fitness report that re-fuses every graded capture's stored observer context under candidate weights to ask whether the hand-set constants (Heim .4 · ∇P .3 · phase gate .6/1.1) are earning their values — report only, adoption stays a co-decision. Both tiles unlock as the graded record grows; the record began today. Built in collaboration with Claude Fable 5. July 2026. RC 14.9F completes the audit's structural fix: the Heim 12D modifier chain moves into log-odds space — each legacy probability multiplier m becomes the logit shift 2·ln(m), the unique translation matching the old behavior at P≈½ — so ψ₅₆ · ψ₇₈ · ψ₉ · Hessian evidence now moves the odds symmetrically instead of distorting near the tails, and fund health is recentered so strength genuinely lifts where the multiplicative chain could only shrink. The same update rule as the Track B posterior: one Bayesian grammar across the pillar. Co-decided, re-blessed, 36/36 green. Built in collaboration with Claude Fable 5. July 2026. RC 14.95F adds the audit's first modern-math addition: adaptive conformal band coverage (Gibbs–Candès ACI). Every NEP forecast now stamps its stated 90% band onto the Calibration Ledger, every graded outcome replays through a guarded pure corrector — each miss widens the outer band, each hit tightens it — and long-run coverage converges to the stated level regardless of model misspecification: the physics gives the band its shape, conformal calibration guarantees its honesty. The Model Audit card reports the running adapted level. Guard grows to 37 vectors, 37/37 green. Built in collaboration with Claude Fable 5. July 2026. RC 14.96F acts on a two-agent adversarial review: the conformal loop's stated band is now buffered so it attaches regardless of render/log order (a first-view ordering bug had silently kept coverage pinned at nominal); the Stripe return re-reads real status with retries instead of trusting the URL; the AI-overload retry is bounded; the paywall toggle reconciles with the server before flipping; subscription status won't paint across a user switch; and defensive NaN/undefined guards were added across the probability math. No free-access path was found — enforcement is genuinely server-side. Built in collaboration with Claude Fable 5. July 2026. RC 14.97F adds the audit's second modern-math addition: an anytime-valid calibration e-value (a betting-martingale e-process averaged over a λ-grid) in the Model Audit card — a running measure of evidence that the posterior is miscalibrated that stays statistically valid no matter how often you check it, unlike a p-value (e≥20 rejects calibration at 0.05 by Ville's inequality, at any stopping time). Guard grows to 38 vectors, 38/38 green. Built in collaboration with Claude Fable 5. July 2026. RC 14.98F–15.0F close the session with polish: a UCC Lab review guards the last unhandled NO-DATA Fund Health display, the manual physics-hypothesis dropdown is removed from This Week's Picks — now that all four pillars are graded and Physics derives its own call from the engine, every slot is equally bias-free — and the ET clock moves up beside the RC version as a small box, leaving the 🦄 TODAY strip's five status pills symmetric. Built in collaboration with Claude Fable 5. July 2026. RC 15.1F lights up the last inert placeholder in UCC Lab: 🌐 Global QA — the whole-board honesty check that lays all four pillars' directional calls for each ticker side by side and flags where they conflict. A disagreement isn't a bug; it's the platform admitting its own lenses point different ways, and the cue to look closer before trusting any single read. Every UCC Lab placeholder is now a live view. Built in collaboration with Claude Fable 5. July 2026. RC 15.2F adds the audit's third modern-math addition — path signatures (rough-path theory): two depth-2 log-signature Lévy areas now frozen under the golden guard and flowing into every ∇P dataset sample — tpArea (was the move front- or back-loaded — early exhaustion vs fresh late momentum) and pvArea (does volume lead price, or chase it). They encode how price moved, not just where it ended — genuinely new information beyond the current eight features. Collected and frozen now; folding them into the live model is a Track A retrain + re-bless co-decision, never a silent swap. Guard grows to 39 vectors. Built in collaboration with Claude Fable 5. July 2026. RC 15.3F closes the audit's addition list with a burst-aware Aumann comparator — the honest realization of the Hawkes idea. A true temporal Hawkes needs a timestamped options-print stream the snapshot data doesn't carry, so rather than dress that up, the market-implied probability now shrinks toward ½ by EFFECTIVE sample size (inverse-Herfindahl of the premium weights), not raw dollar premium: a few dominant prints — one coordinated sweep — collapse the effective count toward one and can no longer scream conviction, while genuinely dispersed flow keeps its full voice. Co-decided and re-blessed. Built in collaboration with Claude Fable 5. July 2026. RC 15.4F brings the Quantum Eigenstate math under the golden-vector guard — the Hermite-node S/R levels, the ψₙ₊₁/ψₙ₋₁ turning-point targets, and the harmonic-oscillator transition probabilities were correct but unguarded; they are now frozen like the rest of the physics core so they can never drift unseen. No behavior change — purely protective. Guard grows to 40 vectors. Built in collaboration with Claude Fable 5. July 2026. RC 15.6F completes the physics & probability audit — the last substantive item, estimator uncertainty in the NEP bands. ρ̂, κ̂=ν̂², ν̂ and λ are measured from finite history: a scale estimated off M bars carries ~1/√(2M) relative error, and κ̂ doubles it, so short or dirty histories carried a band scale the engine didn't actually know — yet drew it as exact. Now the band SPREAD (never the drift) is widened by the predictive factor √(1+6/N) on the effective sample behind the geometry: about +1% at a full year, +14% at twenty bars, →1 as history grows. It composes with the conformal loop — this is the honest model-based prior width, conformal empirically corrects the residual — the completion of the RC 8.5 program. Co-decided (analytic over bootstrap: deterministic, freezable, no per-render latency), re-blessed, guard 42 vectors. Built in collaboration with Claude Fable 5. July 2026. RC 15.7F reimagines the 4D phase-space view: the fourth dimension is the future. The 3D ribbon already flows through time (deviation × momentum × time, history → now); 4D now extends the time axis forward from the LIVE plane as a forward probability cone — the honest ±1.6σ deviation band widening as √(bars ahead) and by the #6 estimator-uncertainty factor, fading as certainty falls. Past = the solid ribbon behind; future = the probabilistic cone ahead. It puts all the band work (NEP · #6 · conformal) into the flagship visual and renders the path-signature idea literally — the ribbon's swept area is the Lévy area. The oscillating-tilt, holographic-echo and comet gimmicks are retired; the cone is the fourth dimension. (It is the price/deviation marginal forward in time, labelled as such.) Built in collaboration with Claude Fable 5. July 2026. RC 15.8F closes the loop on the forecast itself: the tab formerly mislabelled "Poisson" is now the honest Non-Euclidean Forecast, its fog richer (16 shells, an always-on expected-value spine), geometry now measured on a full year of history to match the pillar — and the headline, the realized-outcome backtest: the chart draws the 90% band the model WOULD have drawn T bars ago (σ/μ from data up to then, no lookahead) with the real price path landing inside it (✓ held) or breaking out (⚠ miss). The forecast grades itself on screen — the calibration soul of the platform, made visible. Built in collaboration with Claude Fable 5. July 2026. RC 15.9F adds a new quantum-information read grounded in real physics: Mandel-Q number statistics. A coherent state has Poissonian event statistics; the Mandel Q parameter (Fano − 1) says whether the market's discrete events are independent (Q≈0 COHERENT), steadier than random (Q<0 REGULAR, sub-Poissonian), or clustered and bursty (Q>0 BUNCHED, super-Poissonian — the self-exciting/Hawkes regime). Two channels — volume bursts (participation clustering) and >1σ moves (volatility clustering) — each read over 15 windows, only labelled once Q clears 1.5× the Poisson-null noise so thin data can't fake a verdict. It's the discrete-count partner to the κ̂ vol-of-vol thermometer and the complement to the von Neumann state-purity read, shown as a ⊙ Coherence tile in Market Physics. The estimator is frozen under the golden guard (43 vectors). No QKD, no POVM cosplay — just the honest projection that transfers. Built in collaboration with Claude Fable 5. July 2026. RC 15.94F makes the Hessian a real Hessian. A review found the classification badge was computing empirical time-curvatures of the price path — not the Hessian of the Hamiltonian — while the field visualization already drew the true one. Now `_computeHessian` computes the genuine Hessian of H = ½kx² + p²/2m with the momentum-coupled cross term: ∂²H/∂x² = k · ∂²H/∂p² = 1/m · ∂²H/∂x∂p = γ·p, evaluated at the live (x,p) — so det = k/m − (γp)² and the surface saddles exactly when the momentum coupling overtakes the spring/mass curvature (|p| > ω/|γ|). Badge and field now agree. A consequence of the honest physics: because k>0 and 1/m>0 the trace is always positive, so LOCAL MAX is impossible — the old method's 'exhaustion' reading was an artifact, and it's gone. The classification (which feeds the Heim modifier) is guarded, re-blessed, 45 vectors. Built in collaboration with Claude Fable 5. July 2026. RC 16.0F closes the Hessian review's last open item: D_target measures its gradient honestly — the two-point 6-bar difference is replaced by the ordinary least-squares slope of the last 10 log-closes, so every point votes instead of two endpoints (~6× less gradient noise, same per-bar units, same tanh scale and Aligned/Opposed thresholds). Bundled cleanup: the Heim chain's −5% LOCAL MAX modifier — dead code since the true Hessian made LOCAL MAX impossible — is deleted rather than faked back in; the golden guard now freezes the FLAT default branch instead. Re-blessed, 44 vectors. Built in collaboration with Claude Fable 5. July 2026. RC 16.1F composes the Coherence toolkit into the Aumann market comparator (stage 6): when a ticker's own event stream reads significantly BUNCHED (super-Poissonian volume bursts, Mandel-Q z ≥ 1.5), clustered options prints are echoes of the same information, not independent evidence — so the flow-implied P(up) shrinks its effective sample by the standard overdispersion Fano factor F = 1 + Q before the toward-½ shrinkage. Down-weight only: REGULAR flow never up-weights, and coherent or thin streams are untouched. The gate itself is frozen under the guard (an insignificant Q must reproduce the unadjusted read exactly). Re-blessed, 46 vectors. Built in collaboration with Claude Fable 5. July 2026. RC 16.11F turns the review lens on the ∿ Hamiltonian Surface itself: the x_turn/p_turn turning-point markers now compute their height with the exact same formula as the surface cells — resting on the bowl wall where the physics puts them, instead of floating near the floor on a mismatched scale — and the E1–E4 iso-energy rings clamp symmetrically on both lobes, so a high-energy ring can no longer squash on one side while overshooting the other. Render-only fixes; no guarded math touched. Built in collaboration with Claude Fable 5. July 2026. RC 16.12F finishes the job with one unit system for the whole well: everything on the surface now derives from the engine's own energy formulas (PE = 50k·x², KE = 50p²/m). The z axis is genuinely canonical momentum p = m·v — so the particle's plotted height equals its true H for every mass, the CLASSICAL MECHANICS panel's turning points shed a 10× overstatement, the E1–E4 rings widen 10× to actually wrap the bowl wall at their stated energies, and the predicted orbit finally passes through the particle instead of orbiting 100× beneath it. Built in collaboration with Claude Fable 5. July 2026. RC 16.13F is the well's performance pass: surface heights and colors are computed once per fetch and only re-projected per frame (previously ~4,900 height samples and 1,225 color strings every frame), the stats strip updates once per result instead of at animation rate, paints are capped at 30fps (the slow rotation looks identical at half the cost), and the canvas gains Retina backing for a sharper surface. Built in collaboration with Claude Fable 5. July 2026. RC 16.14F teaches the well's shape: Brian noticed AAPL draws a narrow bowl and NVDA a wide one — that IS the spring constant made visible, since the walls rise as 50k·x² and the surface normalizes to the ticker's own energy, making the width the ratio H/k. The ENERGY LANDSCAPE legend now carries the read (narrow = stiff spring, reverts fast · wide = soft spring or big H, trends roam) with the live k, and hovering the box explains the full interpretation with the symbol's own k, ω, mass, and rim deviation. Built in collaboration with Claude Fable 5. July 2026. RC 16.15F makes the 3D and 4D phase-space views depth-honest: every energy ring and orbit-ribbon segment now paints back-to-front by its true projected depth, so near geometry correctly occludes far geometry at any rotation angle, with ring labels drawn on a final pass so text always stays legible. Render-only; no guarded math touched. Built in collaboration with Claude Sonnet 5 under Claude Fable 5's review. July 2026. RC 16.16F reshapes the daily entry points: the Morning Brief trades its one long scroll for a two-column grid of boxes — each section scrolling inside its own card — and the home Quick Links bar grows to six pills in one row, adding ⚡ Day Trade and 🌱 Long-Term buttons that jump straight to their Scanner tabs. Built in collaboration with Claude Sonnet 5 under Claude Fable 5's review. July 2026. RC 16.17F brings the Academy's own pattern to the Mindspace: the 20 Inspiration Quotes leave their long static stack for the Trading Overview lesson-box treatment — a responsive grid of numbered author cards, each with the ☑ viewed check and a Viewed n / 20 progress line — the first step of the quotes becoming structured lesson plans. Built in collaboration with Claude Sonnet 5 under Claude Fable 5's review. July 2026. RC 16.18F names the doors after the rooms: the card Tools buttons and every scan result's dive action now say Pillar Lens — the app's established name for the four-pillar Deep Dive — and the crypto card's physics view becomes Coin Orbit, on the button and its panel title alike. Built in collaboration with Claude Fable 5. July 2026. RC 16.19F catches the one that hid: the Market Physics Scanner's own results row still said Dive — spotted by Brian on a live AAPL scan — and now reads Pillar Lens like the rest. Built in collaboration with Claude Fable 5. July 2026. RC 16.20F opens the Phase Cone review with its honesty pass: the future rings' tooltip no longer claims a Lyapunov damping the drawing never applied — it now says plainly that θ is held constant forward — and the cyan dashed ring is labelled the illustrative boundary it is, not a measured negativity region. The words now match the picture. Built in collaboration with Claude Fable 5. July 2026. RC 16.21F completes the Phase Cone review's mechanical items: the stats bar updates once per result instead of thirty times a second, the zoom slider can finally reach its own default again, the cone canvas gains Retina backing like its siblings, and the ring and Wigner geometry is computed once per result and merely re-projected as the view rotates. Built in collaboration with Claude Sonnet 5 under Claude Fable 5's review. July 2026. RC 16.22F is a whole-day polish wave: the Phase Cone's four stat cards (including the new PREDICTED ORBIT card, one source of truth with its graph) move above the controls and the cone canvas becomes a true square; the Watchlist Highlights finally refresh the moment a ticker is removed; the entire News tab group becomes two-across boxes and the ticker search actually searches the ticker you typed; the Day Trading scan's run buttons adopt the centered pill row with the sharp red softened to an accent — calm chrome, loud warnings; and the Trading Framework becomes the Alignment Checklist, now the default tab of its screen. Built in collaboration with Claude Sonnet 5 under Claude Fable 5's review. July 2026. RC 16.23F closes the Phase Cone review: the future rings now decay at the engine's own measured Lyapunov rate — θ·e^(−γ·m), the first co-decided math addition since the coherence toolkit, frozen into the guard as it grows to 48 vectors — the holographic-echo and comet-tail gimmicks retire from the cone as they did from the 4D view, and the last four Dive doors are renamed Pillar Lens. Built in collaboration with Claude Fable 5. July 2026.

Version history

🖥 Mac Mini Security Stack

PENDING

RECOMMENDED

Last reviewed: 2026-06-08

📖 Physics Defs

🐛 Error Log