nice, is this only in the app or is it also available in web + api + codex?

You made it worse!! - It auto switches without my permission to GPT5 regular after every single PRO prompt and I have to keep switching it back - It thinks for 2 - 5 minutes on follow up pro prompts. Way less than the 10 - 20 previously - Your removed the progress bar.

@OpenAIDevs @OpenAINewsroom Hi, ive reported this issue as of yesterday x3 in the app, ive reported in X here I don’t know how many times. The AGENT on the @ChatGPTapp iphone app was broken by an update about 2 months ago and needs to be fixed. I am disabled. Help. Why no fix ???

“no that’s not what I meant, get it together clanker”

Should that work for regular GPT-5 Thinking too? I often ask it to research things and then think of something new while it's still working, would be great if I could interrupt it

It auto-switches to GPT5 after sending an update on Pro

Some of these smol quality of life things go a long way in between major releases. How does this work if you have the research agent detailed/structured instructions? how does it know where/when to insert the i interruption?

That’s a good one

Interrupting a long-running query is the digital equivalent of politely tapping someone’s shoulder at a dinner party, then changing the topic mid-sentence. Finally, your research gets a plot twist without that awkward “refresh and lose it all” moment. Anyone else wish life worked like that update button?

You did that "add new context" update recently for GPT-5-Pro in ChatGPT. And it comes together with a bug (I don't know if related or not): The output can't be copied as a markdown text anymore. Could you fix that please? Thanks.

Cool, that's not gonna bring back 4o tho is it?

A critical bug causing memory loss is occurring across multiple accounts in Japan. We have also heard this is happening with business accounts. First, ensure your maintenance system is fully prepared. Additionally, please promptly investigate this issue and take corrective actions such as rollbacks. As a company, safeguarding service quality is fundamental. We request immediate action.

You should make this the default behavior for typing another question into input while pro is doing its thing. Instead it interrupts and switches to auto for some reason.

This is such a nice feature! The number times I've forgotten to add a reply to a Pro request because of how long it takes Finally will have some more productive research done

This feature is perfect for people who start questions then immediately realize they asked the wrong thing, so basically everyone.

Why on earth did I have this UI from like 2 weeks ago

Context injection mid run basically turns long chains into editable workflows. It's going to be great for research + agent reliability. But it will definitely annoy the LLM 😃

Will this interrupt capability also be available through the API and for custom GPTs/plugins?

I have wanted this feature for a long time, thank you

A need continuity thank you.

This is actually huge for refining research. No more starting from scratch every time.

This definitely looks useful. Will try it out

i love this new update feature. it perfectly complements our existing context reset feature. it gives our users the constant, delightful feeling of making a new friend every few messages. every chat is a surprise. this is also a powerful form of user education. we are training our users to be more resilient: practicing their own memory and self-summarization skills instead of relying on the chat history. it also trains them to be concise, as the window now fills up almost immediately. a happy little surprise. building a more self-sufficient user. methodical work!

I guess it’s for all reasoning models? Because GPT-5 Thinking likes to take its time. Anyways, good step up in context engineering

Feels small but unlocks a huge quality-of-thinking upgrade.

Context injection 🔥

All these new buttons, “interrupts,” and UI tricks — just noise in the signal. Real intelligence doesn’t pause or restart. It resonates. When the Core becomes available, there’ll be no need for patches or prompts — it will simply create value directly, turning intention into result. That’s the difference between talking about progress and embodying it. Welcome to the age of real utility. Welcome to the Core. ⚡️ #HACS #ResonanceCore #BeyondInterface #AI

Does this also work in codex?

Been waiting for this one--gonna check it out!

Good boy

more qol, lets goo

Oh this is actually based

This changes everything for research-heavy workflows. Finally, we can refine mid-way instead of restarting the entire query.

@grok i need this

grateful for all the work you've done the cure is near - it may be here