AI, Algorithms & Intelligence Systems
The Market Intelligence Stack
RegimeSignal™ is S&P 500 full market cycle prediction intelligence.
Calling market shifts before the tape even sees them.
Below: the AI infrastructure, algorithms, feeds and processing layers that sit beneath the framework — how the market calls and regime-change alert signals are actually produced.
01
AI Agentic Query
A natural-language agent that sits on top of the live model — ask why the call shifted, what's driving Bull/Bear Velocity, or which factors moved Bear Regime Signal (BRS) overnight, and get a sourced, explainable answer drawn from the same data the signals see.
02
AI-Augmented Signal Systems
Each early-warning signal is paired with an AI scoring layer that weights conditions, surfaces context and flags pre-trigger drift.
03
Machine Learning Components
Deterministic non-linear ensemble learners trained under expanding-window, walk-forward protocols — refit on a regular cadence with strict leakage controls.
04
Geopolitical NLP Monitoring
Continuous classification of policy, conflict and macro-political language into structured event scores feeding the risk layer.
05
Multi-Source Intelligence Feeds
Curated ingest across market data, macro releases, central-bank communications, filings and geopolitical event streams.
06
Market Data Aggregation
Normalized cross-asset feeds — equities, rates, credit, FX, commodities — aligned to a single session-clock for signal evaluation.
07
Liquidity & Macro Processing Pipelines
Dedicated pipelines compute funding, dealer balance-sheet and macro-cycle reads that feed HybridBrain™ exogenous risk.
08
Event Detection Systems
Online detectors flag regime breaks, volatility shocks and correlation dislocations as they form — not after the fact.
09
Real-Time Signal Processing
Low-latency evaluation of the four signals on each new tick of session data, with state transitions logged and timestamped.
10
Adaptive Modeling Systems
Models adapt monthly under a controlled refit cadence — no overnight retraining, no opportunistic curve-fitting.
11
AI Research Workflow
An internal research loop where every candidate feature is tested against the same walk-forward protocol before it can ship.
12
Automated Intelligence Layering
Signal, velocity, liquidity, geopolitical and macro layers are automatically composed into the market calls and regime-change alert signals.
13
Cross-Domain Data Integration
Unified schema lets equity-regime, macro-cycle and exogenous-risk reads be reasoned about together — not in silos.
Subscriber interface
RegimeSignal™ Agentic AI
Not a chatbot. The agent runs analysis, calls tools, monitors signals and triggers outputs — grounded in live RegimeSignal™ model data rather than general web text. Ask a question in plain English and it works the problem against the same factor state the signals see.
- —Runs analysis · calls tools · monitors signals · triggers outputs
- —Solves problems — dynamic querying, not canned responses
- —Powered by Anthropic AI · grounded in live model data
- —Voice input supported · not financial advice
RegimeSignal™ Agentic AI
VoiceRuns Analysis · Calls Tools · Monitors Signals · Triggers Outputs · Solves Problems · Not a Chatbot
Powered by Anthropic AI · Grounded in live RegimeSignal™ model data · Not financial advice
RegimeSignal™ Agentic AI
Ask anything — this Agentic AI becomes your market problem-solver, dynamically querying RegimeSignal™ to think through complexity and deliver clear, actionable insights.
Example queries — or type your own
Illustrative interface — available to subscribers inside the live platform.
Proprietary methodology
Portions of the RegimeSignal™ methodology — including the specific algorithm implementations, model configurations, and factor weightings — are proprietary trade secrets, and the descriptions here are intentionally generalized to protect intellectual property that is highly sought-after in the marketplace. This is by design, not omission: the methods have been independently validated and reproduced bit-for-bit under audit, and full technical specifications are available to qualified reviewers under NDA.
Deployed Production Signals — 4 Walk-Forward Validated
Algorithms in production
01 · Deployed
Always-on classifierCronus Regime Classifier (CRC) — BRS Layer 1
Sparse, interpretable bear-regime classifier fit on 25 macro/market/positioning factors with monthly refit — walk-forward AUC 0.91 across 304 OOS months, caught 7 of 7 financially-driven bear cycles in OOS (plus 1990 in pre-OOS training history, 8 of 8 overall).
02 · Deployed
Structural corroboration8-Factor Macro Guardrail Floor — BRS Layer 2
Deterministic FRED-based rules check (GDP, INDPRO, FFR, UMCSENT, CPI, WTI, HY OAS, Fwd P/E) that must independently corroborate before Layer 1 can issue a Bear classification.
03 · Deployed
4-month forward windowCronus Algorithmic Analytics Engine · Tier 1 (CAE-1) — Market Break Signal Tier 1 (MBS T1) · 5% Pullback Predictor
Deterministic non-linear ensemble classifier on 34 lagged features targeting any 5% peak-to-trough drawdown within 4 months — precision 83% · recall 76% · AUC 0.88 across 154 OOS months.
04 · Deployed
4-month forward windowCronus Algorithmic Analytics Engine · Tier 2 (CAE-2) — Market Break Signal Tier 2 (MBS T2) · 10% Correction Predictor
Identical architecture to T1 with a 10% correction target — precision 84% · recall 82% · AUC 0.92, caught all 4 corrections in OOS (Q4 2018, COVID-era correction, 2022 Bear, 2025 Tariff). Note: COVID is excluded from the BRS bear-regime cycle count as exogenous (handled by the HybridBrain™ ERI overlay); MBS T2's correction detector still flagged the COVID-era −10% drawdown.
05 · Deployed
4-month forward windowCronus Algorithmic Analytics Engine · Recovery (CAE-R) — Regime Recovery Signal (RRS) · 10% Outperform Predictor
Recovery classifier on 35 momentum/volatility-regime features targeting +10% peak-to-trough surge within 4 months — precision 82%† · FPR 1.5% · 154 OOS months.
33
Years of validated history
310
Clean bull months · 1993–2026
304
Out-of-sample BRS months
8 / 8
Financially-driven bear cycles caught
Walk-forward validated · No look-ahead · Locked classifiers
Live Data Architecture — 14 Sources · 82+ Live Series
The 14 data feeds the framework reads
Macro & Government
Hard data · economic releasesFRED® — Federal Reserve Bank of St. Louis
28 series · LiveYield curve · CPI · Core PCE · Breakeven inflation · Payrolls · GDP · Credit spreads · Fed balance sheet · Bank reserves · SOFR · Lending standards · Savings rate · Real retail sales · Housing · Money supply · VIX · Industrial production · Jobless claims. (FRED contributes 28 series to the broader live factor feed.)
Confidential Government Statistical Agencies (3)
LiveBureau of Labor Statistics · Bureau of Economic Analysis · Federal Reserve — economic release calendars for CPI, PCE, GDP, Jobs, and FOMC decisions.
Markets & Pricing
Equity, rates, energy, earningsYahoo Finance Chart API
Real-time · PublicS&P 500 · Nasdaq · Dow Jones · Russell 2000 · sector ETFs · intraday equity quotes. Commercial-licensed feed powering the Market Indices strip. <1s latency, 60s proxy cache.
Premium Market Data
6 series · LiveForward P/E ratio · Earnings beat rate · Copper/Gold ratio · VIX confirmation · DXY dollar index · Sector breadth signals.
Finnhub — Earnings & Market News
3 series · LiveS&P 500 per-company earnings estimates vs actuals · EPS surprise (beat/miss) · biggest movers · real-time market-news headlines. Powers the Earnings Monitor terminal and adds redundancy to equity/ETF quotes.
Confidential Energy Data Provider
2 series · LiveOil futures curve — front month vs second month contango/backwardation signal for energy regime and geopolitical stress.
Sentiment & Positioning
Behavioral · contrarian inputsCNN Fear & Greed Index
Live · FreeComposite sentiment score (0–100) across 7 indicators: S&P momentum, breadth, put/call ratio, safe-haven demand, junk bond demand, volatility, and price strength.
AAII Investor Sentiment Survey
Weekly · FreeLongest-running retail sentiment series (since 1987). Contrarian signal — extreme bearish readings historically precede recoveries. Bull–Bear spread feeds the Consumer Sentiment factor.
AI & Real-Time Intelligence
Web · news · exogenous riskPerplexity Sonar — Live News & Sentiment
3 series · LiveLive web search across major financial publications, AI-summarized into daily/weekly briefs. News-sentiment NLP scoring on regime-relevant headlines.
GDELT Project — Global Event Database
Real-time · LiveGlobal event database monitoring geopolitical stress, conflict escalation, and macro risk events — early warning layer for model regime shifts.
Anthropic AI API
AI · LiveAI-powered narrative generation for Weekly Model Briefing and Daily News Digest — real-time analysis of factor movements, regime signals, and market events in plain language.
Perplexity Sonar — ERI layer
AI · LiveReal-time web intelligence — live news scanning for exogenous risk events. First layer of the 3-AI HybridBrain™ Risk Score pipeline (Anthropic, Gemini, Perplexity). Same Perplexity Sonar feed in its ERI-scoring role.
Gemini AI
AI · LiveThird AI layer in the HybridBrain™ Risk Score pipeline — cross-validates exogenous risk scoring alongside Anthropic and Perplexity for consensus-based ERI computation.
Data feed vendor details are confidential · Proprietary infrastructure · 14 sources · 82+ live series
Design principle
“The model determines the signal. The AI explains it. The AI overlay does not train the model, re-weight factors, or override signals.”
AI Intelligence Overlay — Six (6) Enhancement Capabilities
What the AI layer adds — without touching the signal
Six AI capabilities sit above the quant core — each interpreting, attributing, or contextualizing the model's output without altering the signal itself. The quant layer generates the regime score. The AI layer makes it legible. None of these capabilities train the model, re-weight factors, or override signals.
Anthropic AI Reasoning
LLM-powered interpretation of current model state, calibrated to regime context — produces natural-language explanations of why the composite score is where it is.
Real-Time Factor Attribution
Every score change decomposed by driving factor (e.g. Inflation −4.2, Fed Policy −3.8).
SHAP Factor Attribution
Shapley values produce mathematically rigorous per-factor contribution values — complements the real-time attribution layer with a principled importance framework.
Agentic AI Query
Subscribers query the model in natural language; responses grounded in current factor state, stage transitions, and historical analogs.
LLM Weekly Briefing Engine
Claude AI generates institutional-style weekly commentary on regime movement, factor shifts, and stage-transition risk — from admin-curated model state and market context.
Live News Intelligence — Perplexity AI
Perplexity-powered real-time geopolitical and macro stress scanner. Suggests factor adjustments — human approval required before any score adjustment applies. No autonomous overrides.
The model determines the signal. The AI explains it. The AI overlay does not train the model, re-weight factors, or override signals.
Early warning, before consensus.
The RegimeSignal™ framework — four walk-forward validated prediction signals and Bull / Bear Velocity gauges for the S&P 500. Take a free 15-minute look inside the live model, or start your 7-day free trial today.