ACEM Proposes Replacing COCOMO II for Agentic Development With Three Additive Cost Dimensions: Tokens, Human Oversight, Infrastructure
ACEM argues traditional estimation models (COCOMO II, Function Points, Story Points) break down when autonomous agents do the implementation and humans do planning and validation, because identical tasks consume different tokens, follow divergent reasoning paths, and require varying correction. It introduces three constructs — the Revision Factor for token overhead from rejection and retries, the Context Factor for rising consumption as context accumulates, and a four-level HITL Intensity Score — and maps Use Case Points, Story Points, and Function Points to estimated token consumption so teams can reuse existing scoping data. The model is fully specified but its constants are left symbolic pending empirical calibration, which the authors explicitly invite.
↳ Follow the thread