The project began from a solo plan: a four-week core alpha and an eight-week MVP at 15 hours per week. Agent-assisted implementation can shorten coding work; it does not replace user validation or native compatibility evidence.
- Offline synthetic demo.
- Version-bounded fresh Codex discovery.
- Duplicate-preserving, evidence-aware comparison.
- Single-entry support bundles with explicit disclosure choices.
- Contract, privacy, process, and CLI tests.
- Local inspection of selected names and duplicate entry IDs without opening raw snapshots.
- Short tagged-source demo command and compiled GitHub release installation.
- An opt-in runner for real Codex fixture observations, separate from synthetic CI tests.
- Version-consistent packaging and a tested, checksummed GitHub prerelease workflow.
The compatibility allowlist remains narrow. A publishing pipeline and local native fixtures do not replace independent users trying the product.
- Talk with five maintainers and identify two recurring support cases.
- Observe three external pilots; at least two should reach a useful comparison without live assistance.
- Document two real cases in which a user confirms the report was helpful.
These are uncompleted validation goals, not adoption claims. If existing tools solve the same cases more simply, prefer an integration or upstream contribution.
- Expectation checks and a small CI wrapper.
- Additional native CLI/OS versions, then one additional host.
- Richer reports and resource fingerprints.
No scheduled promise for model evaluation, hosted dashboards, or automatic repair. Stars are an interest signal; repeat use, useful support cases, and maintainability are the product signals.