Happy OpenAI DevDay!
Today we’re introducing a new generation of apps you can talk to – built right into ChatGPT – and the tools for developers to build them. Whether it’s creating a playlist with @Spotify, finding your next home with @Zillow, or taking a course with @Coursera, these apps meet you in the conversation and adapt to your context.
It’s the start of a new ecosystem that brings more utility to users and new opportunities for developers.
ChatGPT can now do work for you using its own computer.
Introducing ChatGPT agent—a unified agentic system combining Operator’s action-taking remote browser, deep research’s web synthesis, and ChatGPT’s conversational strengths.
Connecting outside data to @ChatGPTapp has been a real game changer for my daily work. Excited to finally ship it externally! Give it a shot and let us know how it goes.
🔭🔌 We're rolling out more connectors to ChatGPT! Deep Research can now access your @googledrive, @HubSpot, @linear , @github and more. Pro and business users can also add custom MCP connectors.
This is just the beginning of ChatGPT being deeply embedded with all your applications. We're excited to see how you use it!
@k_kohlbrenner@OpenAIDevs@k_kohlbrenner There are a few reasons why a repo may not appear at first. Take a look at this help center article and let us know if you're still having trouble after trying the suggestions there: https://t.co/0ZBNBpp1VF
@testingcatalog@OpenAIDevs@testingcatalog - there are a few reasons this may happen. Take a look at the help center article here and let us know if you're still having trouble: https://t.co/0ZBNBpp1VF
You can now connect GitHub repos to deep research in ChatGPT. 🐙
Ask a question and the deep research agent will read and search the repo’s source code and PRs, returning a detailed report with citations. Hit deep research → GitHub to get started.
@DanielLurie can we do better than leaving recently repaved streets in a hand painted state for months? Several of these photos have been this way for 9 months+. It’s unsafe and also just embarrassing? I see this all over SF.
As we enter the world of test-time compute, we are seeing increasing returns by simply letting our agents do their thing for longer.
For the first time, we are running our agent for hundreds of steps on these benchmarks. Instead of accumulating errors, CUA introspects, updates, and tries over and over again until the final success.
We launched a research preview of Operator today! It's a model built on top of GPT-4o that can control a browser — it is very early and will make mistakes, but it's a taste of things to come https://t.co/eYQbIqI1Lw