Open questions
Research in progress, in the open. Each question states a hypothesis, the evidence, the experiments that would settle it, results so far, and our current belief — never presenting an open question as a settled result. Consolidated from the Octoryn Research notebook.
- RQ-001State: HypothesisConfidence: Low–mediumCan Apple GPUs support persistent-kernel global synchronisation?
A forward-progress-safe persistent kernel may be possible only if the runtime can bound residency, barriers and workgroup scheduling under realistic contention.
- Sovereign and Local AI
- Model Evaluation
- RQ-002State: Evidence gatheringConfidence: MediumWhich DAWM Phase 7 regressions are architectural rather than tuning noise?
The recurring regressions likely come from state-abstraction pressure, not a single optimiser or dataset setting.
- World Models
- Model Evaluation
- RQ-003State: Design noteConfidence: Medium–lowWhat governance primitive belongs inside an agent runtime rather than above it?
Governance should own reversible execution boundaries, not just policy prompts or external approval flows.
- Governed AI Systems
- Evidence, Audit and Replay