A controlled ablation on a production science agent found model choice dominates topology, and a PPO policy nearly matched it for free.

On a verifiable protein-function characterization task routed across tools, model choice swamped federation topology, RL-versus-LLM harness, and prompt expertise: Opus at roughly 92 to 94%, o4-mini at 40 to 50%. Federation across institutional boundaries cost almost nothing (arXiv 2608.25215). The result I'd act on: a PPO policy hit 88% at zero token cost with the fastest latency and perfect consistency, but no reasoning trace. For routine verifiable tasks, a cheap deterministic policy sits close to frontier. Prompt dependence was largest exactly when the task was hardest.