Comment by alecco
19 hours ago
> I had some issues getting it to understand that it wasn't meant to terminate, so I instead told it to write a basic bash script that calls claude with the -p flag and "please continue" whenever it detects its not running.
This is why we can't have nice things. Anthropic is placing more restrictive limits and now you risk being locked out for hours if you need to use it a bit more than usual (e.g. you have an impending deadline or presentation).
I wish Anthropic just banned these abusive accounts instead of placing draconian (and fuzzy) limits. The other day there was an idiot YouTube streamer actively looking to hit limits with as many concurrent Claude Code sessions as he could, doing nonsense projects.
I believe these are fundamentally two different types of abuse — the OP is engaging in a significantly less harmful version … the seriously harmful version is the account sharing / massively concurrent one which abuses whatever holes exist in the streaming API to allow Claude to “complete the completion and then stop because of limit” (which I think is there to make the UX better)
Just letting a single instance run all the time … is not that bad, seriously.