生産性・インテグレーション
Auto Research — Universal Optimization Loop
Universal autonomous optimization loop based on Karpathy's auto research methodology.
Accepts any artifact (code, prompt, document, config, template) plus a metric and eval
criteria, then runs an iterative improve-measure-keep loop without human involvement.
Use when user says "auto research", "optimize this", "run the loop", "improve this
autonomously", "auto optimize", "karpathy loop", "iterative improvement", "run evals
on this", "make this better automatically", or wants to systematically improve any
artifact with measurable outcomes. Also trigger when user mentions "binary evals",
"pass rate", "optimization loop", "autonomous improvement", or "auto loop".
Works for code performance, website speed, document quality, prompt reliability,
config tuning, template optimization, and any domain with an objective metric.
For skill-specific optimization, prefer /skill-optimizer which wraps this methodology
with skill-aware eval infrastructure.