Roadmap
The real roadmap belongs to HeepX itself, not to this client — full detail, exit criteria, and abandonment conditions live in workspace/ROADMAP.md. Summarized here, deliberately without dates: phases are ordered by dependency, not by a schedule nobody can commit to.
Establish that a capability can be written down and mechanically checked at all. Nine capability repositories, each with a specification, a deterministic scorer, three shipped baselines, and honest limits.
Find out whether the ecosystem's central claim is true. One controlled measurement so far: +0.208 at p=0.125 (not significant), against a prompt-length cost significant at p=0.008. Exits either by reaching significance on two capabilities, or by saying publicly that it doesn't.
Remove two single points of failure: one author, one runtime (jellyclaw). Entry requires Phase 1 to exit, whichever way.
Work at a scale where nobody has read everything. Entry requires more maintainers than the founder can review behind.
The organization outlives everyone who built it. No schedule; the test the earlier phases are for.
This client (unofficial)
Separately, and much smaller: this Python client's own maintenance continues independent of the phases above — bug fixes, the six frozen functions, and whatever the real ecosystem's evolving HXS-1 schema needs reflected in CapabilityMetadata. It does not add capabilities, and it does not speak for HeepX's own roadmap.
Track HeepX's own progress in workspace/BACKLOG.md and OPEN-QUESTIONS.md. For this client specifically, see its GitHub Discussions.