@Dimillian Awesome

@Presidentlin gotta channelise the big dawg energy

@gustavo_pch For now you can't, I'm playing with that ideas, but it's a looooot of stuff for a remote/mobile app. so will see

@Dimillian Another feature missing for windows/android users

@Dimillian How to stage/unstage hunks? Need that on the Mac app as well to stop using git add/diff in the terminal

@reach_vb well done!

@Dimillian this is so useful

@reach_vb It’s that simple.

Don’t gave to provide a demigod model, just provide a good model people can actually use

@Dimillian android updates?

@Dimillian Diff sheet comments are the useful bit here. A phone review only works if the tiny note lands back in the composer before Codex keeps moving.

@reach_vb Nice!👍 create and manage /goal from mobile add inline review comments to changed files.

imo /goal is wildly underrated. I use it with my own simple loop harness (using the CLI) & it provides the best balance of running many tasks while not getting cognitive overload. 1/2

@Dimillian This is sweet!

@reach_vb Always for iOS Always for ios

@reach_vb The Android app needs this fix ASAP. Fonts in the codebox should be monospace.

The worktree-per-thread piece is the real unlock. In real agent workflows, parallel threads collide fast when they share one working tree, so giving each its own worktree is what makes running several at once practical. The usage stats matter for the same reason: in practice the binding constraint on agentic coding is cost-per-outcome visibility, not raw capability.

@Dimillian yes

@reach_vb Per-thread worktrees is the right call. What bit me running parallel coding agents was shared working state, two threads editing the same files and quietly corrupting each other. Isolating each run in its own worktree kills a class of flaky bugs you cannot reproduce later.

@reach_vb The apps (desktop+mobile) have become my workspace & /goal generator + master control panel while the CLI executes via my simple loop runner

It’s such a great combo

I do want to add some cheaper models into mix for easy tasks but codex apps + CLI with /goal is my winning combo

@Dimillian nice, very nice.. however i was hoping for 5.6

Vaibhav, if you remove the Spark model limit and give us a combined usage limit instead, I'd be happy with that. The Spark model isn't smart enough for my use cases, so I rarely use it. As a result, that quota goes to waste while I end up consuming my regular model limits much faster.