Learn
Cohort intelligence — and the discipline behind it.
Read the canonical pages on what cohort intelligence is, the calibration and eval discipline that backs it, the practical guides for wiring it into your agent or research workflow, and how it fits alongside other finance tools you might already use.
Concepts
What we mean by cohort intelligence.
Canonical pages owning the methodology terms. Read these first if cohort intelligence is new to you.
Cohort intelligence →
The pillar page. What it is, why it beats point forecasts, the four engineering disciplines.
Calibrated stock forecasting →
Why distributions beat point estimates and how we validate calibration.
Symbol-disjoint evaluation →
The eval discipline that prevents leakage in financial ML.
Conformal prediction in finance →
The math behind the calibrated probability bands.
Methodology
How Chart Library works under the hood.
Eval results, honest negatives, and the discipline that backs the cohort numbers.
Full methodology →
End-to-end methodology page: embedding pipeline, eval procedure, validation results, honest negatives.
How it works →
Plain-English walkthrough of the retrieval + calibration steps for a single anchor.
Layer 3 intelligence →
The full Layer 3 cohort intelligence — outcome distribution, feature attribution, regime stratification, risk profile.
Guides
Practical bridges from your problem to the cohort primitive.
Working code, agent-builder integrations, quant-research recipes.
MCP server for finance →
Wire Chart Library into Claude Desktop, Cursor, or any MCP-aware agent.
Build an AI trading agent with Claude →
End-to-end walkthrough of the cohort + memory loop architecture.
k-NN regression for stock returns →
What works, what fails, and the calibration that makes it honest.
Vector search for stock chart patterns with pgvector →
Production schema, IVFFlat tuning, gotchas at 25M embeddings scale.
Backtest a chart pattern honestly →
The eval discipline that separates honest backtests from inflated ones.
Compare
How Chart Library fits alongside other tools.
Most of these are complements, not competitors. Use both.