Productivity platform Notion is using OpenAI's Codex, an AI coding model, to speed up its engineering work, according to a case study published on OpenAI's blog. The company says its engineers are using a workflow it calls "one-shot specs," where a plain-language specification is fed to Codex to generate a working code prototype, cutting down time spent on manual coding and boilerplate.
One example cited is Notion's AI Voice Input feature for its web app, which the company says it built faster than would have been possible without Codex handling much of the underlying code generation, including translating spoken input into structured commands within the app.
OpenAI frames the case study as evidence that small engineering teams can ship more ambitious features by offloading routine coding work to AI, allowing developers to spend more time on architecture and user experience rather than repetitive implementation. Beyond Notion's own account and the specific voice-input example, the post does not include performance benchmarks, timelines, or team-size figures, so the extent of the speedup is not independently verified.