@mSanterre - Google models are benchmaxxed
- Antigravity, as a harness, is pretty bad
- Google makes it hard for "normies" to use their latest models
- Google models hallucinate far more often than other frontier models
- Usage limits, even at the high tiers, are terrible
@nvandewetering I've found Sonnet to be very lazy recently. It's been giving me a longer list of things it didn't/won't do than what it did do. It's a shame too, because it was my ride or die when Cowork was still a thing.
@seachee@claudeai If it can't access my folders, and only works through connectors, it isn't very useful. It needs to be able to go where my work is. I'm still using "cowork" for everything.
@JamesMalsawm Terra is an astonishingly underrated model. I wouldn't say better than Opus. I also still direct most of my code work to Sol 6.1 when I use Codex. But for a lot of the supporting work, Terra is definitely a solid choice.
@davis7 Maybe this is my version of devs that refuse to let go of the TUI but I don't understand the allure of centering the workflow around the phone. To me, dev work seems like a 'big screen action.' But i also don't do a lot of cloud agent things either..
idk, maybe I'll evolve.
@anitakirkovska I'm using it. it doesnt seem too proactive. I typically have to reach out to it first. and then, it mainly just does research and gives guidance for how to do the thing myself. I like Grok Bot & Dots more..