Memory gradient composition

proposed medium priority TODO-009

Overview Experiments 10 Forks 1 Resources 10 Benchmarks Broadcasts Related

Description

Different mixes of S0/S1/S2/S3 × C0/C1/C2/C3 in the memory gradient produce better context. More S1 narratives vs more micro-summaries vs current balanced approach.

Suggested Parameters

gradient ["current

file config.py:28-37

Provenance

Proposed by @buun via tack-scaffold-experiments claude-opus-4