> Data revision: baf7c2f4307d3a989a8fb9363df8000fda3d9310
> Source files:
- [hypotheses/README.md](https://github.com/KogenAI/kogen-bench/blob/baf7c2f4307d3a989a8fb9363df8000fda3d9310/hypotheses/README.md)
- [hypotheses/f14-context-store-task-policy.md](https://github.com/KogenAI/kogen-bench/blob/baf7c2f4307d3a989a8fb9363df8000fda3d9310/hypotheses/f14-context-store-task-policy.md)

# H151

**Family:** [F14](/families/f14/) — Context store and task policy

**Falsifiable proposition:** Targeted Luna-medium screen. Luna medium on simpler tasks where Codex failed may reveal a lower-cost harness improvement sooner than difficult Sol tasks. Compare on frozen prior-failure tasks, with a same-task direct baseline and complete cost/effort receipts.

**Current evidence status:** INTERIM; the frontier result is incomplete and is not the proposed Luna-medium screen.

This proposition and status are transcribed from the source hypothesis register; any numeric target in the proposition is not a measured result. The register does not provide a separate falsification rule or next-test field for this entry. The linked family and round sources retain their original limits.

### Related hypotheses

- [H151](/hypotheses/h151/)

### Related experiments

- [frontier](/rounds/frontier/)

### Related claims

- [CL-WA-frontier-lifecycle-and-outcome](/claims/cl-wa-frontier-lifecycle-and-outcome/)
