Anthropic's PR lead Lydia Hallie posted a thread this week explaining why Claude Code users are burning through their limits faster than before. The short version: peak-hour limits got tighter, context sessions got bigger. Here are some tips to use less. She was polite about it.

The response was still a disaster.

Not because Lydia said anything wrong. She said what she could say. The problem is structural: when users run into limits at scale, the right response isn't to teach them to be smaller. You can't fix users. What you can do is watch what they're actually doing and ask whether the system accommodates it.

Users burning through limits fast is signal. It tells you what they find valuable and what they can't put down. A thread that responds to that with "turn off extended thinking" and "start fresh sessions" isn't a product improvement. It's asking users to be different users. That's a deflection dressed as advice.

Don't blame the people. Blame the system.

Original post