OpenAI is more honest with its users than Anthropic.
Claude users can only use the frontier model, Fable, for 50% of their quota. So they get plenty of usage, just not Fable-level quality.
OpenAI users get a real choice: burn through everything fast with Sol UltraFast, or use Luna Max / Terra / build a smart, self-improving orchestra and save a huge amount of tokens.
Go codex
@theo I have 2 modes in Codex. Classic one when GPT tell me what he is doing, when I want to control and just am curious what is he doing. And another one mode when he write me only super important steps, very useful to check on the go from iphone.
poteto poteto poteto
Thanks a lot for opportunity to try @cursor_ai. I'm installing Cursor CLI now to my server and will 100% share feedback with community.
Mine are the same. Not dramatically leaked.
I use SOL Ultra (hard infra tasks where I need brains) + orchestration with delivering tasks to agents with small models + infra kit where we store all routes in infra + code kit where we store code which we need to repeat in different places.
24 hours of non-stop SOL Ultra = 66% of usage left now. It seems to be okay for me now.