Inside OpenAI's Race to Catch Up to Claude Code
Summary
This in-depth article explores how OpenAI, the biggest name in AI, found itself racing to catch up to Anthropic's Claude Code in the AI coding agent market. The piece is based on interviews with over 30 people inside OpenAI.
🔑 Key Findings
- Revenue Gap: Claude Code generates $2.5B annualized revenue vs. OpenAI Codex's $1B
- Historical Context: OpenAI had Codex in 2021 but paused it after ChatGPT launched in late 2022
- Anthropic's Advantage: Focused on coding earlier, trained models on real-world messy code repositories
- Windsurf Acquisition Failed: OpenAI tried to acquire Windsurf for $3B but deal fell apart due to Microsoft complications
- Strategic Importance: Altman calls coding agents "one of these rare mult trillion-dollar markets"
Detailed Analysis
The Race Timeline
- 2021: OpenAI demonstrates Codex, shows it can turn English commands into code
- Late 2022: ChatGPT launches with 100M users in 2 months; Codex team moves to other projects
- June 2024: Anthropic releases Claude Sonnet 3.5 with impressive coding abilities
- Early 2025: Multiple internal groups at OpenAI start working on coding agents
- February 2025: Claude Code launches as limited research preview
- March 2025: OpenAI forms sprint team to compete
- July 2025: Windsurf acquisition falls apart
- August 2025: OpenAI hits accelerator on Codex
Why Anthropic Won Early
Greg Brockman acknowledged that Anthropic was "focused very hard on coding" from an early stage. They trained their AI models not only on difficult coding problems from academic competitions but also on real-world problems from messy code repositories - a lesson OpenAI was "delayed on."
Codex vs Claude Code Today
By September 2025, Codex had been getting just 5% as much use as Claude Code. By January 2026, Codex's user base shot up to closer to 40% of Claude Code's. OpenAI spent heavily on customer acquisition, with developers reporting $200/month plans giving them over $1,000 of usage.
Enterprise Adoption
Cisco's president Jeetu Patel tells employees: "I guarantee you'll lose your job if you don't use them, because you won't be relevant."
Safety Concerns
A nonprofit called the Midas Project accused OpenAI of falling back on safety commitments with GPT-5.3-Codex. Amelia Glaese, OpenAI's head of alignment, rejects this idea.
Why This Matters
This article provides rare insight into how even the biggest AI companies can miss market shifts. OpenAI's focus on ChatGPT and multimodal models caused them to overlook the coding agent opportunity - a market they actually invented.
The competition between Codex and Claude Code represents a broader shift in how software is built, with AI agents becoming capable of increasingly autonomous software development.