@thsottiaux being able to view images from the mobile app
for example, in my thread i ask for screenshot evidence of output, agent complies, but in the mobile app I just see broken image references.
@aiCrof i've got it hooked up to hermes, opencode, and pi. I'd really like to daily-drive it and replace my codex sub ($20/mo tier), but haven't found a UI agent orchestrator that I love (like i fell in love with codex). I find myself burning through codex & cursor limits like crazy.
@leerob is @cursor_ai 3 intentionally obscuring the question/answer tool when I'm past my usage limit? I get a notification about a need for answers but cant access them from glass, i wind up having to ask the agent to put them into the markdown plan and have me explicitly answer
@Zenoware@cheatyyyy@FireworksAI_HQ are you saying GLM-5 is available on fireworks pass, as in the same unlimited token usage as Kimi K2.5 turbo? or just regular token pricing?
All this talk about web layout models reminds me of css constraints and the cassowary algorithm https://t.co/Aeg1wjpNzo
I got into this for a while. Could probably vibecode a super efficient canvas implementation.
Probably shouldn't.
3 things are bringing me back to @cursor_ai today
1. Composer 2 (vs my codex credits that i burned through in a single day this week)
2. Glass. If anyone can get this right, I pray it's the @cursor_ai team.
3. @leerob is just a stand-up guy. His interactions are always kind.
you guys can be arm chair lawyers and open source philosophers, I'm just gonna see what I can get out of this model. It's not a one-shot or anything, but it rips.
Kudos @cursor_ai