I don’t know about “good”, but yeah apparently the labs are now training models in tandem with the harnesses, so if you use a model with another harness you’re probably burning more tokens and getting worse output.
I’d have a hard time describing the Claude harness and the rest of its associated software as anything less than “good,” but that’s just my personal opinion.
Especially when we are talking about non-technical business users. The things that Claude desktop can do with things like Outlook, Jira, and Confluence are impressive.
A bit confusing. You can definitely purchase credit for Claude and use it in pi or omp. I have done this.
They have a separate plan that gives you vertically-integrated usage through Claude code and they have a pay-for-tokens plan but as far as I know the two are separate (though they’re going to add some usage for the latter if you pay for the former).
You can definitely purchase credit and use it in pi. If you don’t want to, that’s fine. People stop using services all the time and that’s fine.
6 comments
[ 0.28 ms ] story [ 22.7 ms ] threadEspecially when we are talking about non-technical business users. The things that Claude desktop can do with things like Outlook, Jira, and Confluence are impressive.
They have a separate plan that gives you vertically-integrated usage through Claude code and they have a pay-for-tokens plan but as far as I know the two are separate (though they’re going to add some usage for the latter if you pay for the former).
You can definitely purchase credit and use it in pi. If you don’t want to, that’s fine. People stop using services all the time and that’s fine.