Anthropic just quietly cut Claude Code's weekly limits by 17 per cent, and users are doing the maths

Anthropic just gave Claude Code users a "permanent increase" that works out to a cut, and developers who actually did the maths are not thrilled.

Since May, Anthropic had been running a temporary 50 per cent boost on top of Claude Code's weekly limits, repeatedly extending it through mid July, then July 19, then August 19, then August 31. That boost expired on September 13. The very next day, a new "permanent" allowance kicked in, pitched as a 25 per cent increase on the original pre boost baseline.

Run the numbers on a simple example. Say the starting weekly limit before May was 100 units. The temporary boost pushed that to 150 units, which is what users had gotten used to for four months. The new permanent allowance lands at 125 units. That is genuinely 25 per cent above where things started in May, and it is also 17 per cent below what people had been working with all winter.

Both numbers are true. Only one of them is the number anyone actually feels.

This is the real story here, not the raw cut itself. Nobody budgets their week against a baseline from four months ago, they budget against what they had yesterday. Framing a reduction as an increase by quietly picking the older, lower comparison point is technically accurate and still lands like spin the moment anyone checks the working.

It is also a familiar shape in subscription land generally: offer a generous temporary boost, let people build habits and workflows around the higher number, then dial it back while marketing the new baseline as a win.

Developers rationing Claude Code every 30 minutes ahead of the change were exactly the audience most likely to notice the sleight of hand, and they did. Anthropic is not raising prices here, it is just changing what the existing price buys, and choosing the most flattering way to describe that change.

Worth watching whether Anthropic adjusts the messaging or just lets the next model release bury this news cycle, because that particular move has worked before.