@TheRohanVarma generally just the ergonomics of a gui being far better than a tui. I loved opencode for a while, and I still use it occasionally, but my productivity increased in an app that actually felt like it could keep up with me
@thsottiaux how do you guys manage right now? I find myself going back to opus or other competent models because it's such a pain to get codex to make something that looks remotly good
@DataChaz I would be very skeptical of tools like this. evan already tested this and found increased token consumption, while I would also bet this also makes caching worse
https://t.co/1SF9dmPS4Q
I integrated this into Copilot and ran some evals this afternoon. Results were neutral at best and in most cases resulted in MORE token consumption. These results are consistent with our prior evaluation of RTK.
Compression ends up removing information that the agent needs, which triggers a re-read, ultimately resulting in larger costs and latency.
Beware of things that sounds too good to be true.
@_Evan_Boyle my other concern with a lot of this is agent performance. I can't imagine lobotimizing the agents in the way that many of these types of frameworks do is productive