@DavidOndrej1 This only applies of you have the capacity to run local models. Just make use of what you have and do not get too attached to the models. They are just tools.
Currently using both Opus 4.8(via claude code) and GPT 5.5(via pi agent), for some reason GPT 5.5 on Pi agent still wins in terms of simplicity, speed and execution. Anyone also compared both setup?
@thsottiaux@badlogicgames i have been using pi agent more than codex app itself using plus plan. Hopefully codex can take the good parts from pi agent like context efficiency and minimal tooling. Maybe add a "minimal mode" toggle feature.