You can now prompt Codex to spin up new threads inside Codex.

Additionally codex can read all threads here's a useful prompt:

Cool updates x3

@_mauriciorubio Ohhh, I missed this, very cool!

the future of work is apparently an ai having a very productive internal monologue.

Can we make one chat session to talk to another chat session? One chat is for frontend one for backend and they can actively chat

I love this, and I love just asking for reminders - it creates automations for you now too with reminders. Codex has come so damn far... kudos @sama and the Codex team.

the self-managing thread is just the 2026 version of a cron job that emails itself. took 40 years but we got here.

yeah this is sick

Before you run the super prompt, ask ChatGPT to improve it first. Tell it to make it better and add anything useful. Then use the improved prompt it gives you.

So I can pin 1 chat, which will be my assistant. The assistant can manage everything else. Meaning I just need 1 agent to talk to to manage multiple projects. Interesting, have to give this a try

this is big!

not enough. i should be able to tell codex to make me a billionaire and it should be able to. after that it should guide me to become a dictator of a small african nation without me even asking.

You seem unimpressed with pretty much everything. Do you work for them? They promised absolute wonders, blamed a bug, and in the end delivered almost nothing relevant. Yet they still get praised and hyped up?

The useful shift is scope control, not just history search. New threads let you branch intent cleanly without dragging the whole prior context window behind every task.

damn codex doing gimmicks now

the interesting part is this flips the orchestration pattern. normally a controller decides when to delegate. here the agent itself decides. feels right for deep refactors where subproblems only surface once youre in the weeds. scary version is a model that spawns threads too eagerly and burns through context

the real win isn't spawning threads — it's being able to kill them. i burn so much context on dead-end explorations that could've been separate threads. finally

the pace of this is insane

Finally!!! This is awesome 🤩

isn't that just adding more noise? splitting focus can kill productivity.

I've also started creating folders from chats and then storing a complete transcript of that chat in a documents folder as well as a handoff document in case I wanna roll it into a project.

Still can't create a new codex project right? Can only spin up new threads inside a project or in a one-off chat. Would be cool to create an actual codex project.

Great update

nc

Nested thread spawning is the step that makes it feel like a system. Not just a chat window.

Are you able to control a thread on a remote computer? Codex tells me it should be able to control threads on my Mac mini from my MacBook Air, and it can send messages to the thread ID, but it can seem to receive messages back. It doesn’t see the remote thread in the thread list.

im building a keyword research/niche or idea finder since two days. am i doing it wrong? i see people build fast and make money. i feel left out. should i be less controlling and start building faster? took me a month to build ann app template and a nextjs website template too.

Can new threads pull context from the main session?

Love it

spawning threads from inside codex is most interesting when the child thread returns a decision instead of more text.

Very cool

Before I was telling it to use Apple scripts to get it done

Houston we have arrived

I heard you like agents. So here’s an agent to spawn agents to spawn subagents. Honestly this is my favorite thing since subagents.