Research velocity is rising, but the gates are doing what we expected

The latest run of Hadto’s ontology evolution loop says something simple: discovery is moving faster, and the editing gates are stopping ontology changes that are not ready to carry real work.

CQ coverage remains near maximum at 99.8 percent, and discovery-backed questions are still growing across multiple verticals. Dental added new research contexts around manual intervention escalation and payer policy bulletins. Home services, professional services, and franchise operations also showed CQ growth.

Research velocity is up. Readiness is now the constraint.

The queue looks slower for a reason

The same cycle produced almost no accepted maintenance proposals. In plain English: the system is still finding new grounded work, but it is refusing to convert weak maintenance ideas into ontology edits.

The delta counters support that claim. The dental cycle added 2 new competency questions and generated 0 proposals. Maintenance generation accepted 0 proposals and rejected 3 through invalid_answer_path. Research discovery accepted 2 proposals with no rejections there.

That behavior is healthy. A maintenance proposal can sound sensible and still fail the handoff to execution. For example, a proposed edit tied to payer policy bulletins should not pass unless the answer path shows exactly how the source evidence maps to the competency question and then to the ontology change. If that chain is missing or inconsistent, rejection is progress. It keeps a plausible but weak edit out of the production ontology.

Watch readiness, not raw proposal count

Two trends matter. Research pressure is expanding across four verticals, and governance is acting as the bottleneck by design.

That can look like lost throughput if the headline is proposal volume alone. The better external reading is stricter than that: Hadto is learning faster than it is granting itself permission to edit the ontology. That is the right failure mode at this stage because rejected proposals keep ontology debt from being promoted into the platform.

The practical work now is straightforward. Maintenance candidates need explicit, machine-checkable answer mapping. Proposal-generation filters need to stay separate from proposal-readiness checks. Answer-path failures need clear reasons. Authors need enough visibility to repair those failures before the next run.

The gates belong in the scaling story

This run also reinforces other ontology-quality work already on the table. ONT-006 and ONT-007 remain open reminders that logical consistency is not the same as ontological soundness. RBox and taxonomy governance belong in the same readiness story as answer-path validation. Any future scaling plan should treat those checks as standard preconditions for cross-vertical rollout.

The operating standard is clear: discovery can accelerate, but the ontology does not change until evidence, answer paths, and governance all line up. That is not drag on the system. It is the system doing its job.

---

Source evidence used in this note: Keet Chapter 5 study tracker (2026-03-31-keet-ontology-engineering-progress-tracker.md), the daily research report (evolution/daily_report.md), delta stats (evolution/delta_report.json), and open governance issues ONT-007/ONT-006 context.