makefaster

What problem does it solve? Improving website speed is usually guesswork: developers apply optimizations without measuring whether they actually help real users. This Skill runs a disciplined autoresearch loop that baselines a user-felt metric, tests one performance hypothesis per iteration, and keeps or reverts each change based on measured numbers. ## Core Features & Use Cases - Baseline profiling: Measures cold and warm page loads with managed Lighthouse (median of at least 3 runs) to establish a noise floor and a north-star metric like LCP. - Ranked checklist walk: Works through an imported improvement leaderboard in rank order, skipping non-viable categories with stated reasons and testing viable ones one at a time. - Keep-or-revert discipline: Every experiment is re-measured under identical conditions and kept only if it beats the noise floor by at least 5% or 20 ms; results are recorded in results.json after every iteration. - Use Case: Point the loop at a self-hosted app repository; it walks dozens of ranked techniques (compression, caching, lazy-loading, font payload reduction), keeps the ones that measurably improve LCP, and reports results for public leaderboards. ## Quick Start Run npx makefaster inside your site repository and let the agent baseline and optimize your page load performance. ## What problem does it solve? See above.