I am using Claude design, and a lot of times it's agent is failing mid-task, with connection errors (not sure is this connection between me and agent, or between agent in Claude infra and relevant LLM model) When these errors happen: 1. My limits have been consumed already 2. There is no resume button, there is only retry, which starts again 3. Even retry fails, then consumes my tokens. I had some chats where I lost 80% my limits by retrying the task, with no luck of completing it. Anthropic, please fix this
Open original sourceCommunityVerified
Tell HN: Claude is using my limits even though it's infra is failing
I am using Claude design, and a lot of times it's agent is failing mid-task, with connection errors (not sure is this connection between me and agent, or between agent in Claude infra and relevant LLM model) When these errors happen: 1. My limits have been consumed already 2. There is no resume button, there is only retry, which starts again 3. Even retry fails, then consumes my tokens. I had some chats where I lost 80% my limits by retrying the task, with no luck of completing it. Anthropic, please fix this
Evidence chain
Hacker News AI discussionsCommunity · community