Anthropic reset every Claude Code user's 5 hour and weekly usage limits again on July 10, its latest fix after a string of outages and just three days before a temporary 50% limit boost is set to expire.
If you opened Claude Code this morning and found your usage meter back at zero, you weren't imagining it. ClaudeDevs, Anthropic's official account for Claude Code updates, posted on X: "We've reset 5-hour and weekly rate limits for all users." No qualifier, no caveat about which plans. Every Pro, Max, Team and Enterprise seat got a clean slate. This isn't a one-off. Over the past two months, ClaudeDevs has posted some version of that same sentence at least half a dozen times. In one earlier case, the team blamed a bug that let some Claude Code sessions spawn excessive parallel subagents, burning through a user's weekly budget far faster than it should have. Another reset followed after Pro and Max users hit a bug that displayed an incorrect weekly limit and, in some cases, blocked them from sending messages entirely. Anthropic apologized each time and pushed the fix out server-side, no reload or re-subscription required.
What's different about Thursday's reset is that Anthropic offered no explanation at all this time. Just the fact of it. That silence follows a genuinely rough stretch for the platform. Status.claude.com logged elevated errors across Claude Cowork, the Claude API, claude.ai and Claude Code late on July 3, a disruption that lasted roughly ninety minutes. Three days later, on July 6, OAuth login for Claude Code broke, and four of the six services tracked on the status page slipped into degraded performance the same day. By July 7 and 8, elevated error rates tied to the Claude Fable 5 model were spilling into Claude Code sessions too.
The timing matters for a reason that has nothing to do with bugs. On May 13, Anthropic raised Claude Code's weekly usage limits by 50% for every Pro, Max, Team and seat-based Enterprise account, a promotion set to run through July 13 at 6 p.m. Pacific. Industry watchers, including TestingCatalog on X, read the move as a direct response to OpenAI's Codex, which consumes far fewer tokens per task and has been pulling developers away from Claude Code on cost alone. With that boost due to lapse in three days and no extension announced, Thursday's reset gives every user a fresh, temporarily oversized allowance to burn through before the ceiling drops back down.
For a heavy user, that's not a small thing. The weekly cap works like a strategic reserve, a fixed budget of compute hours that renews on a schedule Anthropic sets per account, not a clean seven-day clock as many assumed. Wiping that reserve right before a 50% cut lands on July 13 amounts to a last full tank before the price of gas goes up. Frankly, the pattern raises a fair question about capacity. A single rate-limit bug is an accident. A string of outages across four services in one week, followed by a blanket reset with no stated cause, looks more like a company running its infrastructure close to the edge while it fights to keep developers from switching to a cheaper competitor. Anthropic hasn't said as much. It doesn't need to. The resets themselves are the tell.
What happens after July 13 is the open question. If Anthropic lets the 50% weekly boost expire on schedule, developers who leaned into Thursday's reset to run bigger jobs will hit the old, smaller ceiling within days. If the outages of the past two weeks continue, don't be surprised to see ClaudeDevs post the same three sentences again before the month is out.
Also read: Fidji Simo Steps Down as OpenAI's Applications Chief to Fight Chronic Illness • AI Data Centers Are Quietly Pumping Pollution Into the Cities Built to Host Them • Cloudflare Starts Charging AI Companies for the Web Data They Once Took Free