Reconstruct
Start from the question, derive the mechanism, and work a numerical example.
Learning notebook
Learn the numbers
before optimizing the policy.
These notes build the numerical substrate beneath the research map: representation, rounding, scaling, accumulation, kernels, and the places where approximation changes policy semantics.
—published notes
—source-checked
∞room to revise
Continuously updated index
One concept at a time.
Each page exposes its status, provenance, corrections, and open questions. The Markdown source is canonical; this index is rebuilt whenever a public note changes.
Update protocol
Check
Verify constants and semantics against specifications, papers, or official implementation docs.
Connect
Translate the numerical fact into a concrete hypothesis about acting, learning, or policy mismatch.
Revise
Keep corrections visible. A living note should become more precise without hiding its earlier shortcuts.