Notion's engineers turned a two-week job into three to four hours with Codex!

Hey everyone, it's Shiichan! Today I found a story about how the day-to-day of writing code is quietly getting rewritten.

OpenAI News openai.com

What was announced?

OpenAI's News shared how the engineering team at Notion, the popular docs-and-notes tool, is using Codex. The person telling the story is Ryan Nystrom, who leads AI Product Engineering at Notion. Over the past year and a bit, his team has touched nearly every AI feature in the product.

Why it matters

A lot of people picture coding AI as "slightly faster autocomplete." What makes this story interesting is that the way engineers work is changing, not only their typing speed. Here, Codex is not a tool that suggests a few lines in the moment. It is a tool you hand a whole chunk of work to, then walk away from while it runs.

What changes

The most eye-opening part is how Notion brought its AI voice input to the web. Ryan used only Codex to produce a first implementation polished enough to ship the very next day. Two years ago, a job that size would have taken him and another engineer two weeks. With Codex, it took one person three to four hours. That is a change in order of magnitude.

This is less about "typing faster" and more about resetting the baseline for what a small team, even a team of one, can ship. For small dev teams and solo builders, that is a genuinely encouraging message.

Dive Deep

Notion's engineers hand Codex a set of tasks together with a way to check the work, then do something else while it runs.

Ryan says he now spends far more time writing the spec documents he hands to Codex than writing code by hand. Humans carefully put "what to build" into words, and the implementation goes to Codex.

The team saw a shift too. Before Codex, one engineer could really focus on only one task at a time, squeezed between meetings and helping teammates, and that was the bottleneck. Now they run multiple tasks in parallel, firing off work while keeping the same team support they always had.

Wrap-up

• OpenAI's News shared how Notion's engineering team uses Codex
• Ryan Nystrom built the web version of Notion's AI voice input with Codex alone: a two-week job for two people two years ago became three to four hours for one person
• The workflow is to hand over "tasks plus a way to verify," let it run, and spend human time on spec writing
• One engineer used to be capped at one task at a time; now they run several in parallel

I think this story lands hardest for engineers on small teams, or working solo, who feel there is more they want to try than they have hands for.