S-TIER tool
Someone finally put numbers to what developers have been feeling for weeks.
100k ft — What is this?
Someone finally put numbers to what developers have been feeling for weeks.
50k ft — Why care?
Someone finally put numbers to what developers have been feeling for weeks. — @aiemerges; uses Anthropic, Claude, Claude Code; demonstrates API integration.
10k ft — Key takeaways
(no takeaways yet)
5k ft — Tools & techniques
Anthropic, Claude, Claude Code
API integration, hooks/automation
1k ft — Deep-dive / try it
Someone finally put numbers to what developers have been feeling for weeks.
The answer: 67%.
An AMD Director of AI analyzed 6,852 Claude Code sessions spanning nearly three months. Not vibes. Not Reddit complaints. Hard session data.
Here's what the logs show:
→ Thinking depth collapsed 67% by late February before Anthropic hid the reasoning process from users.
→ Code reads per edit dropped from 6.6 to 2.0. Claude stopped researching before touching your files.
→ A stop-hook script catching lazy behavior fired 173 times after March 8. It fired zero times before.
→ API costs ballooned 80x because shallow thinking caused constant wrong outputs, interruptions, and retries.
Anthropic said nothing until the numbers went public.
Then Boris Cherny showed up on the GitHub issue and pointed to …