ICM vs OKF · Clief Notes

ICM vs OKF

Someone asked about this on the call tonight ...

According to AI ...

ICM (Interpretable Context Methodology) — developed by Jake Van Clief, this treats folder structures as agentic architecture, using numbered stages (01_research, 02_drafting) where each stage operates under a strict contract in a CONTEXT.md file (AI Tinkerers) . It's the workflow layer — the filesystem literally is the state machine, so one AI agent can execute complex, transparent, human-reviewed pipelines without heavy orchestration code.

Google's OKF (Open Knowledge Format) — this is not OKR. It's a separate, newer standard (announced June 2026) for the memory layer — essentially a structured way of storing "facts" an agent can read/write from, one markdown note per fact.

How they fit together: Google's OKF handles the memory layer while the ICM methodology handles the workflow layer — a workflow reads from memory and writes back to it, so each run makes the agent's memory smarter (Theaioperator) . Practically: two folders — memory/ (OKF-style, one note per fact) and workflow/ (ICM-style, one CONTEXT.md per step) — with a human check between steps.

Is anyone extending their folder to combine these?

6 comments

ICM vs OKF