I’ll still continue using barely anything other than ChatGPT though.
Just wanted to express genuine acknowledgement about being aware of playing on the loser’s side.
I still didn’t really eval or personally use any Claude this year.
First time I tried they instantly banned me (rightfully so), second time I gave it $ 15 in credits and it got fully drained with like a third of the agentic session done.
New chatgpt.com is slightly uglier than before and more normieesque, but I love the new tool call summaries.
They weren’t a thing at all just an hour ago.
Probably not an issue though.
The titles don’t seem to appear in the agent’s context, so the small model’s incoherent speech won’t increase the main model’s perplexity.
