Ability to set base branch for repos and also enforce signed commits instead of circumventing the issue if Bitwarden/1password vault is locked.
Also clarity on which features work with api/proxy in the app. Using it with my personal subscription it’s clear but when trying to use it at work behind an ai gateway I see odd behavior like the following issue.
https://t.co/jx31tNlNbf
@thsottiaux Creating a worktree option seems to disappear occasionally (Using api). Also, the settings in the bottom left don’t display the normal menu.
@LLMJunky It feels all right. It's still not good at front-end work and also I don't recommend enabling the 1 million context window. It just gets lost in its own context.
At least I can say I tried gpt-5.4 with the 1m ctx window, but it becomes unusable past a certain threshold. It starts getting lost in its own context and referencing older info that's no longer relevant.
The codex memory system is neat. Being able to recall past bugs I’ve hit and fix them in a similar way feels like building on prior debugging experience instead of starting from scratch.
I just don’t see humans becoming “dead weight” in a year. You’re not going to spin up complex AWS/GCP infra and real production systems with a magic wand. There’s architecture, networking, security, trade-offs.
What I can see is sales discovering AI coding agents and showing up with a “quick prototype” that’s somehow already in a customer deck before engineering even knew it existed… with comments like these 😅.
This looks promising, but one recurring pain point for me: many tools treat a workspace/project as a single repo.
A lot of teams aren’t monorepo based. I often want a workspace that groups multiple repos and lets me create worktrees across them together.
Repo centric workspaces break down fast in that world.
We’re excited to launch the Codex app, a command center for building with agents.
It gives you a focused space to manage multiple agents at once, run work in parallel, and collaborate with agents over long-running tasks.
https://t.co/ldE9k0uL5z
@zebassembly@opencode I hit the same issues yesterday in opencode. It also ended up in some weird loop where it just kept repeating the same thing over and over.
But it feels like a good model I might start using more.
Not sure if it had to do with having the new subagents and collaboration mode features enabled in the codex cli, but it felt like I was going in circles with it today.
Been having good luck so far with Kimi K2.5 in OpenCode this evening. The model did get stuck on some JSON parsing issues and resorted to running Python, but other than that it’s fast and gets the job done for incremental frontend improvements.
@xmarioapps I’m in this process right now and it’s painful especially if you plan to release to playstore as well. So many little things to understand to make it all work.