Philip C

@Dimillian Also, can we give ChatGPT Work subagent capabilities? Would be extremely useful for eg automations that need the ability to split off into various unrelated tasks.

@Dimillian It’s incredible, really nice work guys. Cant wait for (a) GPT Live to be able to delegate to / kick off Work sessions and (b) web / iOS Work sessions to sync to desktop.

Yeah, it's very good! I know we're Codex Remote pilled on my timeline, but ChatGPT Work is like Remote, but instead of working on your computer, it works on an on-demand, always available computer, even if your computer is offline. It has access to all your connectors, etc...

@Dimillian @PhilippSpiess

@Aknotymous @PhilippSpiess All I can say is that it'll keep getting better :)

If you update to the latest ChatGPT iOS app version, we have a very cool new feature from @PhilippSpiess. Codex visualisation now also works on iOS; you can ask for all kinds of graphs and custom content to be built on the fly!

@reach_vb is the efficiency coming at the expense of reasoning tokens / intelligence? curious how you could accomplish this otherwise. looks like juice values have changed so quick confirmation would be helpful.

5 hours limit removed. GPT 5.6 Sol is now even more efficient. Crossed 6M users!! Tibo pressed the button for everyone to just go build things 🔥

@signulll Seems they have a lot more good press right now than the other labs, tbh.

there is a very very simple reason why openai gets clowned on x now (even reddit) & why they lost a ton of public support at least with the closest power audiences. they face more negative press & vibes than any other ai company today. why? - it is not because of their product execution or execution in general. - it is not because of the owner of this platform. - it is not because of the transition from non profit to profit. - it is not because ppl are afraid of ai or job loss. - it is not because of their communication strategy or lack of potential strategy in general. - it is not because of their acquisitions. - it is not because of the lawsuits. - it is not because of their relationship deterioration with microsoft or apple. - it is certainly not because of the personality of sam altman. - it is not because of the endless drama on the personnel or the founders attempted coup. if you understand the mechanics of culture, you figure it out rather quickly. & it’s only one word. can you guess what it is?

@nunezvice @ChatGPTapp I've really enjoyed the upgrade, I know the UI/UX with Chat is not there all the way yet but I am sure it will in time. Can you provide more documentation on Work's capabilities in the web / iOS? Like does it have access to subagents, how does sharing memory work, etc.

a lot of people haven’t realized how much more capable Chat has become in the last few days how much of Codex is now available directly inside @ChatGPTapp on mobile and the web. the Codex desktop app is great. I obviously live there. but now, inside ChatGPT, you can tap Work on mobile or the web to hand off a task that runs in the cloud, no computer needed. you can also tap Remote to pick up and continue work already running on your computer. start at your desk, keep things moving on your phone, or kick off something new wherever you are. agents shouldn’t care which screen you started from.

@nathanclark_ @thsottiaux 100% this!

@thsottiaux For Codex: I want user input tool available in non-plan mode. So frustrating having it restricted to that.

At different times you complained about speed, code slop, frontend quality, ... With each release we improve and GPT 5.6 Sol is ✅Fast and token efficient ✅Hardcore at back-end dev ✅Great at front-end ✅Does not use useEffect everywhere What is next?

New rule for FIFA - if you flop and act like you’re in so much pain you need a stop of play, your team has to sub you out for the remaining half. Suddenly everyone will be fine and the sport will be watchable again.

@athyuttamre Mainly email, sharepoint, drive, plus a custom plugin that's effectively a tunnel to an always-on machine running codex. At the least, would be incredible for GPT-Live to be able to delegate tasks to ChatGPT Work which can access all these natively and communicate back status.

@Aknotymous working on both! what connectors do you want to use it with? curious about your use-case

hey @athyuttamre, GPT Live is incredible and worth the wait. Are there plans to open up the connectors/tools it has access to? I mostly just want to be able to connect it to ChatGPT Work instead of being relegated just to Chat (which I hardly ever use now).

@pvncher Would be super helpful to see an example of a great agents.md file from you given all the changes from 5.5 to 5.6, does something like this exist?

Happy 5.6 Sol day! My first pro tip with this new model is: you don't need Ultra to have good sub agent use! The model is happy to use them if you ask. Add "delegate to sub agents proactively" to your prompt, agents.md or skills!

@ajambrosino @dioscuri Does that mean integration with ChatGPT Work? If so, it can't come soon enough!

@dioscuri and to think we have more coming

New ChatGPT voice mode is incredibly impressive. I continue to believe that natural voice interface is underweighted in LLM assessments. For a decent chunk of people, it’s just how they interact with AI.

@athyuttamre @signulll Integration into Codex / ChatGPT Work would be such a game changer!! BTW - is GPT-Live calling 5.5 in the background or now 5.6?

@signulll love to hear this from you signull — what can we do better?

okay, enough openai criticism for a second (still gonna give product feedback when it’s deserved). it’s been a little too easy. i’ve been using gpt live a lot, like a lot lot, & it’s genuinely fucking amazing. the feature to let the model speak like a human “hmm”, “let me check”, brief pauses, & natural turn taking while calling a bigger model or searching or using tools in the background is a fantastic execution. almost all of my conversations have been nothing but pleasant, i can’t describe the feeling but it’s i think a little too good at this point. nothing else is particularly close in this category right now at all. anthropic’s voice experience still feels very very underbaked (after thought really), & gemini’s personality & intelligence seem lacking. & since it’s already reaching an enormous user base, every great gpt live interaction makes people more confident in ai as a product category instead of walking away thinking ai is a gimmick. really great work.

@dkundel @simonw Is there a way for ChatGPT Work in web and iOS to use the same thread management tools that Codex has? It's a game changer for me to manage projects, threads, etc. and one of the reasons Codex is so much more organized & useful to me compared to my ChatGPT chats...

@simonw Work on mobile and web is essentially a Codex agent with a cloud browser 🙂 you can also ask it to schedule tasks!

Anyone know if ChatGPT Codex (in the new ChatGPT desktop app) is a strict superset of ChatGPT Work? Liked if you're a software engineer who isn't intimidated by Git features is there any reason you'd ever want to switch to ChatGPT Work Mode?

@dkundel Huge upgrade, so thank you. Are there plans to have web Work chats sync with desktop? The use case is much smoother on iOS, but still clunky given differences with memory and no desktop hand-off or thread mgmt. Also, does this mean we're not getting voice mode in ChatGPT Work?

PSA: switch ChatGPT to “Work” on web or mobile and you get the Codex agent. Same powerful agent, except it runs without babysitting your laptop and can handle scheduled tasks. Mine starts every day by briefing me on Slack, email, and my calendar.

@LoopOnChain Seems a bit overkill tbh, just run on auto mode.

So apparently GPT 5.6 Sol might just randomly delete all of the files on your laptop 🚨 Here's how to safeguard yourself against it Paste this prompt into codex: Protect this machine from accidental permanent file deletion by Codex. Implement this as defense in depth, not merely as a written instruction. Requirements: 1. Inspect the current official Codex documentation for hooks, command rules, sandboxing, and approval settings before making changes. 2. Detect the operating system and identify its recoverable Trash or Recycle Bin command: - On macOS, use `/usr/bin/trash `. Do not add `--`, because macOS `/usr/bin/trash` treats it as a filename. - On Linux or Windows, verify an installed, reliable Trash or Recycle Bin mechanism before using it. - If no reliable recoverable mechanism exists, stop and explain what must be installed or configured. 3. Add a durable global Codex instruction: - Codex must never permanently delete a file or directory. - All removals must go to Trash or the Recycle Bin. - Permanent deletion requires the user to explicitly change this safety policy first. - Destructive Git operations such as `git clean`, `git reset --hard`, bulk checkout, or bulk restore must also be blocked. 4. Install a global `PreToolUse` hook that returns a hard `deny` before execution when it detects: - `rm`, `unlink`, `rmdir`, or `shred` - `find -delete` - `rsync --delete` - destructive Git cleanup or reset commands - common Python, Node, Ruby, or Perl deletion APIs - file deletion through `apply_patch` - delete-like MCP filesystem tools - nested destructive commands inside `bash -c`, `bash -lc`, `sh -c`, or `zsh -c` 5. The hook must direct Codex to the verified Trash or Recycle Bin command instead. Do not silently rewrite dangerous commands, because flags and shell expansion could be misinterpreted. 6. Add user-level Codex command rules that mark direct permanent-deletion commands as `forbidden`. Preserve existing rules and configuration. 7. Set safe global defaults when compatible with the existing configuration: - `sandbox_mode = "workspace-write"` - `approval_policy = "on-request"` Do not override managed policies or create conflicting permission configurations. 8. Preserve all existing configuration. Inspect files before editing, make narrowly scoped changes, and create recoverable backups of any configuration files that must be replaced. 9. Validate everything without risking real data: - Parse the hook and configuration files. - Feed simulated tool-call JSON into the hook. - Confirm dangerous examples are denied. - Confirm ordinary commands and the Trash command are allowed. - Test the Trash mechanism using a newly created disposable temporary file. - Run an end-to-end Codex test using `rm` with no arguments inside a read-only sandbox. Never test `rm` with a real target. - Confirm the command was blocked before shell execution. 10. Explain any limitations honestly. A Codex hook is a guardrail, not an absolute operating-system security boundary. 11. Check whether system backups are configured. Do not enable or modify backups without permission, but report clearly if no backup destination exists. 12. At completion, provide: - The files created or changed - The validation results - The exact one-time steps required to review and trust the hook - Any restart requirement - Any remaining risks Do not permanently delete anything while completing this task. If cleanup is necessary, move it to Trash or the Recycle Bin.

Honestly, the @OpenAI super-app isn't ready for primetime yet. No integrated voice (despite GPT-Live being incredible), chat is now just a weird pop-out window, no syncing between cloud / local, etc. I would not want to be a "normie" with this release. But I know it'll improve.

Astonishing how much GPT 5.6 has improved in design / frontend and writing. My apologies for doubting the team!

@thsottiaux Great update. Sol is incredible! Very small but annoying Codex shortcut bug: ⌘1 and ⌘2 are assigned to chats inside Projects before the visible Tasks list, so my first task starts at ⌘3. Those shortcuts also remain when the project folders are hidden.

Today, the OpenAI product and engineering team is humming, there is so much to coordinate and keep running across the release of GPT 5.6 Sol/Terra/Luna, ChatGPT Work, the new ChatGPT desktop app, hosted sites and a million other things... together with the new ChatGPT Voice just launched yesterday. We will not get it all right, give us feedback and we will iterate quickly. Thank you for being an amazing community.