5 pages tagged
#agent-training
meta-harness
concept
Outer-loop optimizer that rewrites harness *programs*; same-model 6× spread.
chroma
entity
vector-DB company; Context-1 trains `prune_chunks` as a learned retrieval-side policy.
cursor
entity
AI-first IDE; Composer 2 real-time RL + summary-as-graded-subtask; file-backed MCP A/B test.
moonshot-ai
entity
Chinese AI lab (Kimi); Kimi K2.5 PARL — orchestrator-only Agent training, annealed `r_parallel`.
2026-04-27-llm-training-principles-paths-practices
source
Tw93's third essay: post-training, eval/reward, harness, distillation — the pipeline's back half.
esc