“Same mission, bigger stage”: OpenAI hires Git AI founders to help Codex prove its ROI
The New Stack Paul Sawers
OpenAI hired the founders of Git AI to work on Codex. Their tool shows what AI wrote, so companies can judge if the spend is paying off.
Based on reporting by The New Stack, Paul Sawers — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
OpenAI has brought in the two founders behind Git AI, an open-source tool built to measure how much code AI agents write and what that output is worth. Aidan Cunniffe said on LinkedIn late Friday that he and co-founder Sasha Varlamov will join OpenAI’s Codex team. Their job, he said, is to help businesses get better data on coding agents, from performance to returns.
The pitch is simple enough: if a company is going to lean on AI to write code, it should be able to see what the machine actually produced and whether that work stuck. OpenAI’s Thibault Sottiaux, who works on ChatGPT and Codex, said on X that the company wants Git AI’s technology to help businesses understand Codex’s impact. He framed it as making it easier to spot where Codex is helping individuals and teams get through problems.
Git AI is built around that kind of accounting. It links each line of code to the agent, model, and prompts that produced it, even after the code has been committed, merged, or rebased. The system is meant to show how much AI-generated code reaches production, how often it gets rewritten, and where time and tokens are going. It already works with a long list of coding tools, including OpenAI Codex, Anthropic’s Claude Code, Cursor, Google’s Gemini CLI, and background agents such as Codex Cloud, Claude Web, Cursor Agent, and Devin.
Cunniffe says the deal reflects a shared belief that users should have data to compare model performance and understand the return on every token they spend. That’s the sales pitch, anyway. The more interesting part is what happens when a tool built to measure AI software efficiency gets folded into the same company whose product it is measuring.
There’s also a bit of personal history here. Cunniffe founded Optic in 2018, built open-source tooling around API development, and saw Atlassian buy the company in 2024. He later became a principal product manager there, then started Git AI with Varlamov in the summer of 2025 as a side project. By November, he was calling it a weekend project that had grown much bigger. He left Atlassian in January 2026 to turn it into a company. Now both founders are heading to OpenAI, and it’s still unclear what that means for Git AI as a standalone business.
My take — AI-written commentary, not fact-checked reporting
This is the usual OpenAI move: buy the dashboard, then pretend the numbers will stay perfectly neutral. A tool that measures AI work is useful, but once it lives inside the vendor, the bright line between measurement and marketing gets awfully blurry. Open source survives on trust; corporate stewardship survives on control. Those two habits do not naturally get along.
Read more about this at: The New Stack
