Sites and the Next Generation of Web Apps | DevDay 2026

Categories: AI, Product

Summary

OpenAI's Sites platform is hosting 8 million web apps with 70% of employees building their own, demonstrating how AI-powered full-stack development eliminates deployment friction. The key innovation: combining plugins, scheduled tasks, and SQLite databases lets non-engineers create production tools that auto-update without redeploying.

Key Takeaways

  1. 70% of OpenAI employees actively build Sites, indicating massive adoption of AI-assisted web app creation. Primary use cases: replacing spreadsheets with custom UIs, building internal tools, creating interactive presentations, and prototyping without deployment overhead.
  2. Scheduled tasks combined with plugins enable auto-updating sites: Codex can fetch data from third-party services (Google Drive, Slack, Linear, GitHub) on hourly/daily/weekly schedules and update SQLite databases without code redeploy.
  3. Single-prompt site generation removes technical barriers: non-engineers can prompt Codex to build full interactive apps with filtering, multiple views (list/board/calendar), and data connections—no frontend/backend knowledge required.
  4. Built-in auth via Sign-in with ChatGPT provides security without custom implementation. Sites default to private; users sign in before accessing content, with option to add anonymous landing pages for public deployment.
  5. SQLite database decouples data from code: scheduled tasks update the database directly, instantly refreshing the site without triggering redeployment cycles—enabling truly dynamic applications managed through natural language.

Related topics

Transcript Excerpt

Hi everyone. I'm Katia, and I'm on the Developer Experience team. Hi everybody. I'm John, and I'm the lead engineer for Sites. You may know Sites as a web hosting service that we launched earlier this year, but it's actually way more than that, and this is what we're going to be talking about today. I initially joined OpenAI to work on internal web apps for researchers, but I found it really frustrating to deploy those tools within our OpenAI infrastructure, so I prototyped an easier way to deploy those full-stack web apps within the company directly from Codex. The prototype got passed around by other engineers, then surprisingly non-engineers started adopting it, and apps just started spreading throughout the company. Then we formed a team. Some of those team members are in this audience…

More from OpenAI