@j34nsh33@charlieholtz yo @charlieholtz can you answer this one? I haven't seen this working tbh and would love to know if its being tackled by the team.
@charlieholtz YO DAWG, might I suggest some sort of hot-swappable model configuration so you don't have to upload every time a new model comes out? Not really sure how this works with claude/codex/etc but might be kind of cool.
@conductor_build cc @mattyp , just updated to the most recent version of Conductor and I'm getting this from GPT 5.6 Sol:
It’s still failing with the identical missing binary:
~/Library/Application Support/com.conductor.app/agent-binaries/codex/0.144.0/codex-code-mode-host
@conductor_build and @mattyp can you help me figure out how to stop Conductor from changing the port on app restart (updates for instance)? I use the ports against Docker containers holding databases, and I have to do a bunch of janky stuff to have to sync ports after updates.
@0xJackPhantom@sama I’d say one of the #1 things here is a really good visual feedback loop. That part of the models today is really struggling…with it screenshotting gets more powerful.
@willchen500 I know enough about Harvey/Legora architecture to tell you unequivocally, that you are very wrong. You also have built not even 1/1000 of the app they have. My DMs are open if you want more insight.
@Shpigford any issues today with Opus 4.6 just not "continuining"...been running into issues where it does like a single agent call and then never continues (after like sitting for 5 min)
@pxue@opencode Would love a technical explanation on how this was fixed, I was experiencing some weird tool calling issues and behavior when using Kimi K2.5 via Fireworks and Vercels AI SDK (v5) just last night.
@theo@skeptrune@theo knows something. Either their is a good model coming down the pipe like Composer-2 and he's seen it...oorrrrrr someone is building an IDE that hasn't shown it yet. Or a variant of those two.