OpenAI has hired the founders of Git AI, an open-source tool that measures how much code AI writes and tracks coding agent performance, to help Codex prove its business value to enterprises. The move reflects a growing industry challenge: as AI coding tools proliferate, companies struggle to answer a fundamental question: are they actually worth the money?

Why Can't Companies Measure AI Coding Tool Value?

For months, enterprises have deployed AI coding agents like Codex, Claude Code, and Cursor without clear visibility into what those tools actually contribute. A developer might use an AI agent to write code, but tracking whether that code reaches production, how often it gets reworked, or whether the tokens spent on it generated real value has remained largely opaque. That's the problem Git AI was built to solve.

Aidan Cunniffe, Git AI's co-founder, and his colleague Sasha Varlamov announced on LinkedIn that they are joining OpenAI's Codex team. In a blog post, Cunniffe explained the shared mission: "OpenAI shares our belief that users should have the data to compare model performance and understand the ROI of every token they spend."

Aidan Cunniffe, Git AI's co-founder, and his colleague Sasha Varlamov

Thibault Sottiaux, a member of technical staff at OpenAI working on ChatGPT and Codex, confirmed the integration on X, stating: "We'll make it easier for businesses to see where Codex is making a difference when working through problems for individuals and teams."

How Does Git AI Actually Track AI-Generated Code?

Git AI works as a Git extension that automatically tracks every line of code written by an AI agent. The tool links each line to the specific agent, model, and prompts that generated it, preserving that context even after code is committed, merged, or rebased. This creates an audit trail that teams can use to measure impact.

The platform already supports a broad range of coding agents and tools, including:

• Coding Agents: OpenAI Codex, Anthropic's Claude Code, Cursor, and Google's Gemini CLI
• Background Agents: Codex Cloud, Claude Web, Cursor Agent, and Devin
• Measurement Capabilities: Tracks how much AI-generated code reaches production, how often it's reworked, and where time and tokens are spent

Once installed, supported agents automatically report their edits to Git AI. Developers can continue using their tools and Git workflows normally, with the tracking happening in the background. The system then shows the split between human-written and AI-written code, allowing teams to trace individual lines back to the agent that produced them.

What Does This Mean for Codex's Future?

Cunniffe's career trajectory suggests OpenAI is serious about this integration. He previously founded Optic, a developer tooling startup focused on API development, which Atlassian acquired in 2024. After joining Atlassian as a principal product manager, he started Git AI as a side project in summer 2025. By November 2025, the project had grown significantly enough that Cunniffe left Atlassian to turn Git AI into a standalone company. Less than a year later, both founders are now at OpenAI.

The exact fate of Git AI as a commercial business remains unclear, but Cunniffe has stated that OpenAI will continue backing the open-source project. "We'll keep investing in open source, while also giving enterprises the data they need to build effective software factories," he wrote.

One key question will be whether Git AI maintains its cross-model independence under OpenAI stewardship. The tool currently supports competitors like Claude Code and Cursor, which could create tension if OpenAI prioritizes Codex visibility over neutral measurement.

How to Evaluate AI Coding Tool ROI

For teams considering AI coding agents, understanding the measurement framework that Git AI provides offers a practical approach to evaluation:

• Track Production Impact: Measure what percentage of AI-generated code actually reaches production versus what gets discarded or heavily reworked
• Monitor Token Efficiency: Compare the cost of tokens spent on code generation against the value of code that survives to production
• Compare Agent Performance: Use consistent data across multiple tools to see which agents deliver the best results for your specific workflows
• Preserve Context: Maintain links between code changes and the agents that created them, even after commits and merges, for long-term analysis

This data-driven approach addresses what has been a blind spot in AI coding adoption: enterprises have been spending on these tools without clear metrics for success. OpenAI's acquisition of Git AI's founders signals that the company recognizes this gap and believes measurement is essential to Codex's enterprise value proposition.