Page 3 - Showing 5 of 21 posts
View all posts by years →
- Grounding Is Necessary, Not Sufficient
File-path grounding blocks a class of fabricated plans, but it cannot establish that a cited file supports the claim being made.
4 min
- What a Two-for-Two Agent Result Actually Proves
A 2/2 completion result is useful evidence of a cohort outcome; it is not automatically proof of a mechanism or a production default.
4 min
- Strategies, Not Model Ifs
Different models need different recursive-loop policies. The harness separates who from how, binds them optionally, and branches on no model name.
5 min
- RLM Is Not Automatically Token-Efficient
Recursive loops are not token-efficient by default. Compaction plateaued input near 14-16k per call and cut a 25-step run from ~1.04M to ~367k tokens.
5 min
- When Smaller Cells Make Agents Worse
Prompting for smaller REPL cells cut mean cells/turn from 5.0 to 1.5 — but on a short step budget only the multi-block baseline finished.
5 min