Vol. I  ·  Issue 0414  ·  TUE, APR 14, 2026
The stock market memory for AI agents

24 million chart patterns.
One API call.

Historical pattern intelligence for AI research, trading, and risk agents — and a visual search on this page for humans. Condition on regime, sector, liquidity. Get the full distribution of what happened next.

Free · No signup required·24M+ patterns indexed·
Try:
62%
Direction-correct, last 90 days
16,424
Tracked predictions
9 of 9
Profitable years backtested
2016—today
Ten years of history

How it works

01

Type a ticker

Or upload a chart screenshot. Or pass a date ('NVDA 2024-06-15').

02

We find ten matches

Nearest neighbours in a 384-dimensional pattern embedding, ranked by similarity.

03

See what happened

Forward returns from realised daily bars — not predictions.

The primitive · live on prod

One API call, one conditional distribution.

How to call this from your agent →

The same POST /api/v1/cohort endpoint every agent calls. 500 historical analogs, filtered by regime, sector, or liquidity. Path percentiles, MAE/MFE, realized vol, and a survivorship flag — sub-2-second response.

Cohort · Conditional Distribution
loading…
5-DAY
no data
10-DAY
no data
Same primitive exposed to agents via POST /api/v1/cohort and the get_cohort_distribution MCP tool.
Colophon

Set in Source Serif 4, Inter, and JetBrains Mono. Twenty-four million chart patterns, indexed from public market data (2016–present). Headline returns are bias-adjusted against the model’s recent prediction error. Read the methodology. Educational use; not financial advice.