Anthropic: Claude Code users hitting usage limits 'way faster than expected'

‘I get Claude 12 days a month’: Coders say their AI assistant taps out before lunch

TLDR: Anthropic’s Claude Code helper is burning through people’s paid usage limits shockingly fast, sometimes in under an hour, leaving devs locked out for the rest of the day. Commenters are furious, joking about an AI that “needs a nap,” and some are vowing to ditch big tech tools for open‑source options.

Anthropic’s coding helper Claude Code was supposed to be the tireless robot intern — instead, devs say it’s the coworker who vanishes after one email. Forums and Reddit are packed with users stunned that their pricey subscriptions are hitting “you’re done for the day” limits after just an hour or two of work. One Pro user paying $200 a year claims they only really get to use Claude 12 days out of 30, while another says they burned through a whole “Max 5” plan in a single hour of coding.

Theories are flying. Some blame secret quota cuts during busy hours, others point to a user who dug into the app and claims there are bugs making Claude “eat” 10–20x more resources than it should. One developer joked that asking Claude to write an internet library “killed the whole day’s allowance in one sentence.” Meanwhile, a growing chorus is saying: forget big companies altogether. One commenter proudly refuses to pay Anthropic, OpenAI, or Google, sticking to open‑source tools for privacy and sanity.

The vibe? Half consumer revolt, half group therapy. Some are worried about broken workflows and mystery limits, others are turning it into a meme about “AI that needs a nap.” Everyone agrees on one thing: if your robot assistant disappears by lunchtime, it’s not much of an assistant at all.

Key Points

  • Anthropic acknowledges Claude Code users are hitting usage limits faster than expected and is investigating.
  • Recent changes include peak-hour quota reductions (affecting ~7% of users) and the end of a promotion that doubled off-peak usage limits.
  • A user claims bugs in Claude Code break prompt caching, inflating token costs by 10–20x; downgrading to version 2.1.34 reportedly helps.
  • Prompt caching has a five-minute default lifetime; extending to one hour doubles write token price, with reads at 0.1x base price.
  • Anthropic does not disclose exact plan limits, instead advertising relative usage (Pro ≥5x vs free; Standard Team 1.25x Pro), complicating planning.

Hottest takes

"Out of 30 days I get to use Claude 12" — Discord user on Pro plan
"I used up Max 5 in 1 hour of working, before I could work 8 hours" — unnamed Reddit developer
"I just refuse to use openai/google/anthropic subscriptions, I only use open source models" — lukewarm707
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.