Claude Code users hitting usage limits faster than expected, bugs suspected

Claude Code quota issues disrupting developer workflows
Anthropic has confirmed that Claude Code users are hitting usage limits "way faster than expected," causing disruptions to automated workflows and daily coding. The company states this is their "top priority" and they're actively investigating.
User reports and specific complaints
Multiple users across different subscription tiers are affected:
- Claude Pro subscription ($200 annually): "It's maxed out every Monday and resets at Saturday... out of 30 days I get to use Claude 12."
- Max 5 plan ($100/month): "I used up Max 5 in 1 hour of working, before I could work 8 hours."
Potential causes and technical details
Several factors may be contributing to the increased token usage:
- Quota reductions during peak hours: Anthropic recently reduced quotas during peak hours, affecting approximately 7% of users.
- End of promotion: March 28 was the last day of a Claude promotion that doubled usage limits outside a six-hour peak window.
- Suspected bugs in prompt caching: A user claims to have found "two independent bugs that cause prompt cache to break, silently inflating costs by 10-20x" after reverse engineering the Claude Code binary.
Workarounds and cache optimization
Some users have found temporary solutions:
- Downgrading to older version: "Downgrading to 2.1.34 made a very noticeable difference."
- Prompt cache limitations: The cache has only a five-minute lifetime by default, meaning short breaks result in higher costs on resumption.
- Extended cache option: Developers can upgrade cache lifetime to one hour, but "1-hour cache write tokens are 2 times the base input tokens price." Cache read tokens cost 0.1 times the base price.
Transparency issues and broader context
Anthropic doesn't state exact usage limits for its plans:
- Pro plan: "at least five times the usage per session compared to our free service"
- Standard Team plan: "1.25x more usage per session than the Pro plan"
Users need to check their dashboard to monitor quota consumption. Similar issues were reported earlier this month with Google Antigravity.
Impact on automated workflows
For developers running Claude Code in automated workflows: "Rate-limit errors need to be caught explicitly – they look like generic failures and will silently trigger retries. One session in a loop can drain your daily budget in minutes."
📖 Read the full source: HN AI Agents
👀 See Also

Claude Opus 4.8 Released: Faster, Cheaper Fast Mode, Dynamic Workflows, and Honesty Improvements
Anthropic upgrades Claude Opus to 4.8 with benchmark improvements, dynamic workflows in Claude Code, 2.5x speed fast mode at 3x cheaper pricing, and honesty training.

Developer Replaces $25/hr Virtual Assistant with AI Agents, Confronts Ethical Implications
A developer replaced a $25/hour virtual assistant with AI agents that handle follow-ups, scheduling, lead tracking, and CRM updates. The AI setup costs about $1,000/month and performs tasks faster and more consistently than the human assistant.

Spotify Developers Leveraging AI for Code-Free Contributions
Spotify's key developers have not written code since December due to AI, notably through their internal 'Honk' system that facilitates remote, real-time code deployments using Claude Code.

Deterministic vs Probabilistic Code Generation: Why Bun's Vibe-Coded Rust Conversion Raises Red Flags
Noah Hall argues vibe-coded 1M-line repo changes (like Bun's Zig-to-Rust) are dangerous. Contrasts deterministic transpilers vs. probabilistic LLM output. Tests aren't enough.