active
library
the-learning-problem
Essays on induction, inference, and the search for useful representations
Resources & Distribution
The Learning Problem
Essays on induction, inference, and the search for useful representations.
Posts
| Post | Topic |
|---|---|
2024-09-10-the-policy/ | When optimization becomes existential threat |
2024-09-30-universal-bayes/ | All induction is the same induction |
2024-10-15-latent-reasoning-traces/ | Memory as learned prior |
2024-12-01-mcts-reasoning/ | Tree search for LLM reasoning |
2025-12-19-incomputability-of-simple-learning/ | Why simple learning is impossible |
2026-01-18-value-functions-reasoning-traces/ | Value functions over reasoning traces |
2026-01-rational-agents-llms/ | From A* to GPT |
The Theme
Learning is constrained search, and the constraints shape what gets learned.
- All induction is Bayesian inference with different knobs
- The simplest learning is incomputable, forcing approximations
- Those approximations—priors, architectures, objectives—shape intelligence
Documentation
make docs-serve # Preview at localhost:8000
Blog
These posts also appear on metafunctor.com.
Author
Alex Towell — metafunctor.com — queelius@gmail.com
License
MIT