What problem does it solve? Personal knowledge bases accumulate hundreds of notes, but the valuable non-obvious connections between them remain hidden. Manual cross-referencing does not scale, and naive LLM summarization only restates what single notes already say. This Skill discovers genuinely new insights by pairing notes, generating connection hypotheses, scoring them with an independent critic, and discarding anything below a quality threshold. ## Core Features & Use Cases - Generator-Critic Pipeline: Samples 20-50 recency-weighted note pairs, generates connection hypotheses in parallel on cheaper models, then scores each on novelty, utility, and grounding (1-10 scale). - Quality Threshold Filtering: Only insights averaging 7.0 or higher survive; weak connections are discarded by default so quality always beats quantity. - Traceable Insight Notes: Every persisted insight links back to its source pair, critic score, and a suggested next action, then rolls up into a digest with top insights first. - Use Case: Ask to find hidden connections between your notes on two research topics; the Skill samples pairs, proposes links like a shared underlying mechanism, filters out shallow ones, and delivers a digest of scored insight notes with suggested actions. ## Quick Start Ask the assistant to find non-obvious connections between your notes on two topics and produce a digest of the highest-scoring insights.
