I did not build this to sell a signal.
I built it because I was tired of market commentary that starts with a conclusion and retrieves a metaphor.
The pipeline is intentionally plain:
- Pull macro series (growth, inflation, liquidity, risk, labour).
- Retrieve research and historical context.
- Classify a regime into a closed set.
- Find analogues.
- Map to a posture, with persistence so a one-day blip does not become a philosophy.
The model runs locally. Outputs are structured. If the object does not validate, the run fails. That is the whole product idea.
Why regimes
A regime is a compression. Goldilocks, overheat, stagflation, disinflation, recovery, slowdown, crisis, reflation. You can argue with the list. You should. The alternative is a paragraph that cannot be compared to last quarter.
Analogues matter more than the label. "We have seen a version of this" is a better sentence than "the model is bullish."
What this is not
It is not a live book. It is not advice. The source is private. What I can show is the logic walkthrough — signals, retrieval, a closed regime set, then a posture.