Repository Issue Activity (beta)

#43 selectFrames(1, ...) drops the guaranteed wildcard ~50% of the time

#54 Offer the repo as a Claude Code marketplace listing

#53 Cross-LLM support — decorrelate generator/critic across vendors, not just Claude checkpoints

#52 MCP server wrapper — expose run() as a callable tool

#51 Live TUI tree visualizer for runs

#50 Frame-selection learning across runs (the "dreaming" feedback loop)

#49 Frame packs — installable domain-specific frame sets

#48 Chunked/pairwise scoring — critic context overload at high branch count

#47 Pluggable scorers — custom weights and trap detectors

#46 Cluster-level narrowing instead of idea-level deepening

#45 Recursive deepen — real tree-of-thought, not one level

#44 One failed LLM call aborts the whole run — no partial-failure resilience

#42 Offer the repo as a Claude Code marketplace

#38 One failed LLM call aborts the whole run — no partial-failure resilience

#37 selectFrames(1, ...) drops the guaranteed wildcard ~50% of the time

#32 Skils

#30 Paper titled "Attention Deficit is All You Need" published to the arxivs

#18 Run evals at K=10 and K=20 to bridge K=5 (ours) vs K=100 (literature) gap

#17 Distinguish ADHD frames from personas (A5) and from domain specialists (Han-style)

#16 Address A7 (CHI 2025) human-in-loop counter-evidence in paper §7

#15 Cluster-level narrowing instead of idea-level deepening

#14 Run head-to-head evals vs MoA, Self-Consistency, GPT-5 Pro, superpower-brainstorm

#11 Hyperfocus / flow-state companion skill (brain-model series)

#10 Frame-selection learning across runs (the "dreaming" feedback loop)

#7 Critic context overload at high branch count — pairwise/chunked scoring