Ran minimal agent loop (Ollama /api/chat + read_file/write_file/run_bash) on
steel141 3090 Ti against 3 models on a broken-median-function task:
- gemma4:31b-it-q4_K_M: PASS (8 iters, 1 write, 44s) — textbook trace
- qwen3-coder:30b: PASS (15 iters, 1 write, 22s) — correct but chatty
- gemma4:26b: FAIL (6 iters, 0 writes) — silently stops with eval=4
after reading source. Reproduced on second run. One-shot probe
confirms 26b CAN produce the correct fix — failure is specifically
at the write_file tool-call argument boundary.
Updates GOTCHAS with a new HIGH-severity entry, SYNTHESIS model-selection
table, CORPUS_cli_coding_agent.md empirical-follow-up pointer, and adds
docs/reference/bakeoff-2026-04-18.md with the full writeup.
Fills the gap between existing chat-agent (Simon) and pipeline
(AI_Visualizer) patterns. Covers openclaw/open code/pi/hermes first-party
agents from the HF launch blog, honest positioning vs qwen3-coder:30b,
CLI-agent-specific gotchas (safety filter on security code, long-JSON
weakness, no code exec), and a concrete homelab bakeoff plan pointed at
CT 166 openclaw2 → CT 105 Ollama on pve197.
Key research finding: Google published LiveCodeBench + Codeforces but
NOT SWE-bench or Aider polyglot. The "autonomous agents" claim is
plausible but unproven for multi-file repo-scale coding specifically.
Mortdecai