batching

Checkpoint: CP-007 Status: QUEUED — not yet run. No results exist for this experiment.

Question

Does batching many jobs, or segmenting one model, reduce cost per inference / per token / per CU?

Why it matters

E001 found proving cost is almost entirely fixed per proof. That is precisely the condition under which batching should help enormously — or reveal that the fixed cost reappears per batch. Do not assume it helps; measure it.

Gate

Blocked on CP-002. This is the highest-priority follow-up implied by the E001 result.

---

This is a stub. It exists so the roadmap's structure is visible, not to imply work has been done. Nothing in ../../benchmarks/results/ refers to it yet.