What Codex unlocks for Notion

According to OpenAI, Notion is utilizing Codex to fundamentally change how its engineering teams build software. Ryan Nystrom, who runs AI Product Engineering at Notion, notes that Codex has unlocked new possibilities for what his team can accomplish. A prominent example of Codex's autonomous execution involved bringing Notion's AI voice input to the web. Although a functional mobile version already existed, Nystrom was unfamiliar with its exact mechanics. By pointing Codex at the mobile codebase, providing a description for the web implementation, and supplying a verification method, Codex returned a complete first cut in one shot that shipped the next day. Nystrom stated that a project of this scale previously required two engineers and two weeks of work, but he completed it alone in three to four hours using Codex. Furthermore, engineers can now hand tasks and verification methods to Codex while stepping away, allowing them to run multiple tasks in parallel without being bottlenecked by meetings or peer support. The technology has also enabled engineering managers who have not written production code in years to return to the codebase and ship alongside their teams. Nystrom highlights that he can now queue up tasks, attend meetings, and return to finished features, or even pose research questions before bed and wake up to completed reports. Overall, Codex has reset the baseline for what small teams can ship while giving engineers more room to focus on priority work.