/karpathy — Autonomous Experimentation Skill

You have been asked to activate the Karpathy autonomous research loop. This skill turns any measurable optimization problem into a systematic, never-ending experimentation cycle.

Core Concept

Inspired by Karpathy's autoresearch: an AI agent modifies code, measures a metric, keeps improvements, discards regressions, and repeats indefinitely. Karpathy generalizes this pattern beyond ML training to any domain with a measurable metric.

Activation Flow

Step 1 — Detect Domain and Define the Experiment Contract

Read the current project context and determine: