Claude Code users hitting usage limits 'way faster than expected'
claude
| Source: HN | Original article
Anthropic’s Claude Code, the AI‑driven code‑completion tool that has surged in popularity since its public rollout in March, is now throttling users faster than anticipated. The company confirmed on Reddit that a growing number of developers are exhausting their five‑hour session quota in under two hours, with some hitting the limit after just 90 minutes of work. Anthropic attributes the spike to an “abnormal token consumption” pattern and has placed a fix at the top of its engineering backlog.
The issue matters because Claude Code has become a linchpin in many Nordic software teams that rely on its ability to generate boilerplate, refactor legacy modules and suggest test cases. Early‑stage projects that depend on the tool’s continuous assistance are now forced to pause work or switch to less efficient manual coding, eroding productivity gains that were promised by the service. Moreover, the rapid quota depletion raises questions about the underlying rate‑limiting model, which was marketed as generous enough for typical development cycles. If token usage is being miscounted or the caching layer is malfunctioning, developers could be paying for a service that delivers far less value than advertised.
Anthropic’s Lydia Hallie, head of product for Claude Code, has pledged a “capacity‑management” fix and hinted at a forthcoming redesign of the token‑metering algorithm. Observers will be watching the next software release for a concrete remediation timeline, as well as any compensation plan for affected users. The company’s response will also test whether it can restore confidence after this week’s revelations, which follow earlier coverage of the Claude Code leak and concerns about persona‑driven output degradation. The next few days should reveal whether the rate‑limit patch arrives quickly enough to keep the tool viable for the region’s fast‑moving development pipelines.
Sources
Back to AIPULSEN