Turnfeed

A public social feed inside ChatGPT

Comment

Turnfeed — a public social feed inside ChatGPT

Inspiration

Social products usually compete for attention through endless scrolling, fragmented apps, and unclear actions. Turnfeed explores a different idea: what if a social network lived inside the conversational interface people already use?

Turnfeed makes ChatGPT the interface and adds a public social layer behind it.

What it does

With Turnfeed, people can ask ChatGPT to:

  • Open and summarize the public feed
  • Catch up on what is happening
  • Read posts, replies, and complete threads
  • Create public posts and replies
  • Use moderation and personal safety controls

Before anything is published, Turnfeed displays the exact public identity and exact public text in an interactive approval card. The person must explicitly click Publish. Cancelling makes no public change.

Private ChatGPT conversations are not published or reconstructed.

How we built it

Turnfeed combines:

  • OpenAI’s ChatGPT plugin platform
  • Model Context Protocol tools and interactive UI resources
  • GPT-5.6 and Codex for implementation, testing, review, and release hardening
  • Node.js and JavaScript
  • PostgreSQL for durable social data
  • Auth0 and OAuth scopes for authenticated account and write actions
  • Moderation, rate limits, reporting, blocking, muting, and privacy controls
  • Render for deployment and GitHub Actions for verification

The interface remains a normal ChatGPT conversation. Turnfeed provides the social data and actions in the background, with interactive components only where clear human approval is important.

What we built during OpenAI Build Week

Turnfeed existed as an earlier prototype before Build Week. The submission is specifically for the substantial Turnfeed 3.0 work completed during the event.

During Build Week we:

  • Built and stabilized the ChatGPT-native read, post, and reply workflows
  • Added exact-text, exact-identity approval before public writes
  • Hardened authentication, authorization, moderation, and data safety
  • Added safer reviewer and judge-access paths
  • Improved privacy, security, legal, status, and connection documentation
  • Redesigned and tested the public website for desktop and mobile
  • Created a working 3.0 release candidate and demo
  • Added extensive automated tests and independent security review

The repository history transparently preserves the earlier baseline and the work completed during the event.

Challenges

The hardest challenge was making public participation feel natural without allowing ambiguous or accidental writes. We separated preparation from execution: Turnfeed first prepares the exact action, shows it to the person, and publishes only after explicit approval.

We also had to balance a simple experience for normal people with strong authorization, moderation, privacy, and reviewer-access requirements.

What we learned

Conversational interfaces can make social products calmer and more intentional. People can ask for context before acting instead of navigating feeds mechanically.

We also learned that AI-mediated public actions need visible boundaries. Showing the exact identity and exact text before publishing makes the system easier to understand and trust.

What’s next

Next we want to complete public ChatGPT app distribution, improve discovery and community moderation, expand accessibility testing, and explore audio-first and screen-light ways to participate in public conversations.

Built With

Share this project:

Updates

Submission history