gpt-5.6 models say 1m context in api docs but in codex pro lite they show up as 272k max context lol
so basically the model can do it, but the subscription route is capped for saving cost on rollout, but kind of annoying when comparing it to claude opus 1m context @OpenAI
https://t.co/x4aA9IhRFU
if you are not using hermes agent you're wasting your time @NousResearch
fully switched from using plain codex, I think this is the future of how llms will be used
LLMs are not just simple curve fitting models. They are complex non-linear state search libraries. If you lead/tune them right with skills/rules they will explore the space you desire. Value can only be extracted if lead with human intention to the right library section.