Why Prerendered LCP Reads 6.2s Without activationStart
Speculation Rules prerendering folds the whole pre-click wait into raw LCP. I measured a 6244ms vs 103.5ms gap on Chrome 150 and mapped where RUM needs fixing.
Tags
performance
8 posts
WebSockets and bfcache: Chrome 150 Still Said 'websocket'
Chrome 149 said active WebSockets no longer block bfcache. I re-measured on three Chrome 150 builds and notRestoredReasons still returned 'websocket'. Why?
One listener costs you the back button: six bfcache probes
Six pages, one bfcache blocker each, measured on real back navigations with pageshow.persisted and notRestoredReasons. Only unload blocked the restore.
MIT TLT: Doubling Reasoning LLM Training Speed
MIT researchers introduced TLT, accelerating reasoning LLM RL training by 70-210% through adaptive drafters and speculative decoding. Reduces training costs without additional hardware.
Claude Code with Local Models Triggers Full Prompt Reprocessing
Analyzing the full prompt reprocessing issue when running Claude Code with local LLMs. Learn about KV cache invalidation mechanics and developer tool design lessons.
Recommendation System V3: 100% Faster Builds
Removing the bloated recommendations.json and embedding recommendation data directly in frontmatter eliminates runtime file I/O by 100%. Here is our V3 system journey and results.
Recommendation System Optimization: From 78K Tokens to Zero
Eliminated 78K tokens and reduced execution time 99% using metadata caching and hash-based updates. Guide to building a near-zero cost recommendation system.
Automate Web Performance Optimization with Chrome DevTools MCP
Complete guide to Chrome DevTools MCP: AI assistants measure and optimize performance with real browser data. From Core Web Vitals automation to practical workflows.