@VraserX Totally agree. Whats interesting is this the same situation with Opus 5.5 and Fable. I guess scaling laws and distillation works!
Definitely think 6.1 was meant to be Astra Minor and distilled from Astra.
@RonHendersonGTR@ah20im I figured I would have to route it codex for anything complicated. The fact that is does count for usuage makes me think it's Astra low.
@SimonasLTU1 https://t.co/RWTvJhbIQx -This is my site, with my harness, opus is actually worse than Astra and 6.1 Sol at design. So I'm assuming Sonnets okay. But were running different tests.
I am hoping we can change the model and reasoning effort of dots, that would really set it apart. But it sounds like we wont be able too since the usage is unlimited.
@VraserX Yea, it's pretty cool but you can set an agent to run 24.7 in codex now on any task. And you can direct it with a main thread. So, I guess all of that is just sort of built in now.
Totally. Hopefully the agent capabilities are payment, passwords, and passkeys are all able to be used well in codex and chatgpt. Since I have a macbook, I have the issue of the agent always needing me to approve my passkey (which I could probably solve on my own). it's all a bit clunky at the moment anyways.
@AIMelGibson Do you think we'll get access to a preview version of bel or just a demo of capabilities? Definitely think we'll hear about the millennium problems...I bet the hedge conjecture is solved and will be released.