OpenAI's Jason Liu recirculates a satirical post about integrating ChatGPT and Codex in an endless recursive loop
The viral exchange is purely a joke, not an actual roadmap.
Positive users are excited about OpenAI's recursive ChatGPT and Codex integration as a game changer that improves workflows, while negative users object because it risks combining separate usage limits.
No Digg Deeper questions have been answered for this story yet.
Most Activity
Yo dawg we heard you like Codex in ChatGPT
@mark_k @OpenAI It's pretty simple. We are bringing ChatGPT to Codex so that Codex can be in ChatGPT in Codex in ChatGPT.
Telemetry shows sharp spikes in (reasoning_output_tokens) at exact values like 516, 1034, 1552 and so on. This happens mainly on high and xhigh reasoning, and it's not a smooth distribution at all.
Looks like some kind of fixed reasoning budget or truncation maybe? Complex runs sometimes finalize right at these low buckets. Public repros show wrong answers clustering at 516 and 1034, while correct runs use way more reasoning tokens.
Someone reported it here: https://github.com/openai/codex/issues/30364
I checked my own sessions and I'm seeing the same thing. A lot of my xhigh sessions had a big cluster exactly at 516 reasoning tokens. worth a look
@Alex_m @mark_k @OpenAI No. Tell me more
@Alex_m @mark_k @OpenAI Team will look!
@thsottiaux @mark_k @OpenAI Can't wait dude. No tabs please
@thsottiaux @mark_k @OpenAI Did you notice the ~500 thinking tokens truncation issue on codex?
@thsottiaux @mark_k @OpenAI i just yelled that out loud during our 4th of july party and got tackled
@thsottiaux @mark_k @OpenAI What’s ChatGPT? Just create a no folder chat on codex. I never open ChatGPT since moving to codex
@thsottiaux @mark_k @OpenAI
@petergyang @thsottiaux @mark_k @OpenAI what about toggles?
please add deep research in codex but in whatever format we want. It is disgusting in chatgpt.
I want outputs as an html file but file stays in sandbox and i cant access it in chatgpt and research is gone now. chatgpt memory does not hold research detail too.
It is a do or die task.
ı asked html code instead now. Lets see what it will gonna do. Codex cant search long enough unless i build a freaking deep research tool
@thsottiaux @mark_k @OpenAI Codex is already in the @ChatGPTapp, look forward to having ChatGPT in Codex. Love where this is going.
@thsottiaux @mark_k @OpenAI Before you do that, please please please fix the GitHub sync issue in Codex, it’s actually becoming unbearable dealing with multiple repos 😭 GitHub issue below
@thsottiaux @mark_k @OpenAI wut lol
@mark_k @OpenAI It's pretty simple. We are bringing ChatGPT to Codex so that Codex can be in ChatGPT in Codex in ChatGPT.
@jxnlco Goblins coming?
@thsottiaux @mark_k @OpenAI 帅气 我还是挺期待的
I've been using Codex since the very beginning, and it has made significant progress and seen tremendous improvements for us programmers. We hope it won't be turned into a general-purpose app and that the aspects that matter to developers and programmers won't be neglected. Please don't forget that.
@thsottiaux @thesherlocker @mark_k @OpenAI Yes
@thsottiaux @mark_k @OpenAI If that means my chat starts hitting my codex usage, this will not be a positive change for me.
@jxnlco yeh but is that combining limits as well if so i don't want it take it back ty