Think-Reason-Learn

What it is

Think-Reason-Learn

An MIT-licensed Python library that fuses LLMs with interpretable machine learning. The model does not just produce a score; it produces the reasoning that led to the score, in a form a human can read, edit, and verify.

Think of Think-Reason-Learn as scikit-learn for problems that used to be considered unquantifiable: strategic choices, qualitative assessments, expert decisions. We built it for venture capital, but it generalizes to any domain where humans make high-stakes judgment calls with incomplete information.

How it works

In four parts

1. I

   GPTree

   LLM-powered decision trees: decisions are made by walking a tree whose branches are natural-language questions.

2. II

   Random Rule Forest

   Interpretable ensembles of LLM-generated YES/NO rules combined via voting.

3. III

   Reasoned Rule Mining

   Bayesian, precision-optimized classification with stage-wise evidence calibration and log-odds fusion.

4. IV

   Policy Induction

   Memory-augmented in-context learning with an editable natural-language policy. No gradient-based training.

In numbers

19%

precision on founder success prediction

10x

the 1.9% US unicorn base rate

fit / predict

the same pattern as scikit-learn