@JamesMalsawm it depends on the task, too much reading would drain faster, if it has to play with tools/tests that takes time it would drain less in the same period
@thsottiaux The desktop app needs improvements on the harness when running many agents it hangs.
Agents should be able to watch file changes and wake up on task like Claude does.
@andrew_n_carr In my experience so far, google is the worst at AI, also I wonder how your only talk is benchmarks when we can’t reliably use any of your models
@Google Google is losing in AI battle like they lost on social media against Facebook. You guys failure, all this hype I tried argon and the benchmarks were a scam. its worse than any other frontier model.
Bet your carbon is no difference, hype wont get you success
@ajambrosino I just hope you guys fix the codex desktop app when big harness is running 20+ agents on long tasks it hangs occasionally and I have to restart the app to recover.
I don’t want codex cli as this desktop harness is great for large projects as agents are correcting each other