On September 23, 2026 Anthropic published how it made claude.ai and the Claude desktop app about three times faster in a two-week sprint. Claude found the bottlenecks and wrote the fixes, and the team merged more than three thousand changes without a customer-facing incident or rollback. Most of what it found are ordinary web-app bugs, which is why they are worth checking for in your own code.
Anthropic's post, How we made claude.ai 3x faster in two weeks: https://claude.dev/blog/how-we-made-claude-ai-faster/
The numbers
| Measured at the 75th percentile | Before | After |
|---|---|---|
| A fresh claude.ai load, until the page is typeable | 3.1 s | 0.55 s |
| Starting a new Claude Code session | 0.8 s | 0.3 s |
| Loading a Claude Cowork cloud session | 2.6 s | 0.73 s |
These are Anthropic's measurements of its own apps. They show what the fixes did there, not what the same changes will do to a different codebase.
The em dash that froze the page
If a reply's markdown contained any character outside Latin-1, such as an em dash or a curly quote, V8 (the JavaScript engine in Chrome) stored the whole string as UTF-16. That pushed the syntax-highlighting regexes onto a slower path, and highlighting a finished code block could freeze the page for about a second. Claude fixed it with a twenty-line change that copies each code block into a one-byte string before highlighting it.
The em dash is the character AI models reach for constantly, so any app that highlights or searches model output is a candidate for the same slowdown.
Free, for your email
Get the full setup steps + the copy-paste setup prompt
The rest of this guide (4 more sections, 1 copy-paste prompt) opens right here. The setup prompt, one paste into Claude that sets it all up, comes by email only. Use your real address.
Unlocks: The reload nobody could see · The rest of the list · Check your own app with the FIX skill · Measure before you believe it
50,000+ follow where these guides come from. No spam, reply “unsubscribe” any time.