Skip to main content
The live backlog is the issue tracker. Headline items:
  • Recursive deepen — multi-level Tree-of-Thought expansion, not just one level
  • Pluggable scorers — user-defined weights and custom trap detectors
  • Frame packs — security, ML, frontend, distsys, product
  • Memory across runs — learn which frames win for which problem shapes (issue #10)
  • Streaming output during divergence
  • Cross-LLM support — frames don’t depend on Claude
  • Cross-model judging for the eval suite (issue #6)
  • Hyperfocus / flow-state companion skill (issue #11)

​
Open research threads

Findings from the independent Han research review are tracked openly:
  • Issue #16 — human-in-the-loop applicability (CHI 2025 counter-evidence)
  • Issue #17 — frames vs personas: how persona-prompting findings do or don’t map
  • Issue #18 — bridging the K=5 vs K=100 scale gap in the diversity literature
  • Issue #13 — a controlled parallel-N vs single-prompt-list-N comparison
Want to push on any of these? Join the Discord — frame design, eval problems, trap-hunting, and neurodivergence-inspired research on reasoning architectures happen there in real time.