Oh Claude why? WHY?
Claude managed to fix something very complex for me that I, as a human, thought was impossible for me to fix.
But it has taken days because Anthropic have weird processes and policies.
If you check out social media and the unofficial user groups, you can see that people hit their usage limits, even on the upper-tier plans, really quickly. It does depend on which model you use, your prompts, how long the chat has gone on for, etc etc, but it is super easy to keep hitting that wall.
And it is super frustrating.
The conclusion many people are coming to is ... just use Claude less.
That can't be good for the company.
They also cut off, at the system prompt level, a whole growing use case. You might have heard about thousands of people buying Mac Mini computers to run OpenClaw and other personal assistant agent systems like it. OpenClaw users have their bots essentially run their lives and businesses, and Anthropic doesn't like that.
Mention OpenClaw in your prompt or request headers, and they return an error. Not only that, but using the CLI, they try to steer you back to using it just for software development!
Of course, the solution to both issues is to pay Anthropic even more money. If you add to your pay-as-you-go budget these issues magically go away, but you also run up huge bills.
There is another workaround, and that is to use any other LLM tool for daily, routine stuff and only use Claude when the other can't/won't/isn't capable.
I prefer Cursor for coding tasks, mainly because I use the "guess the next word" feature in the IDE and sometimes the 'cloud' tool is amazingly handy. Codex from OpenAI is a good CLI tool for running stuff in the terminal too.
This leaves Claude with just the Cowork "run my browser for me logged in as my user and click these buttons" stuff ... which, of course, means the others will catch up and eat Claude's lunch.