OpenAI plays catch-up as Anthropic's Claude Code steals the AI‑programming lead

What changed

Anthropic's Claude Code has rapidly turned AI programming into a commercial battleground, reportedly driving more than $2.5 billion in annualized revenue and forcing OpenAI to scramble to regain ground. OpenAI's Codex — an early system that translates natural language into code — has seen usage surge recently too, but remains behind: Codex's January run rate was reportedly a little over $1 billion. It has been reported that OpenAI even pursued a roughly $3 billion acquisition of AI‑coding startup Windsurf before Microsoft (微软) intervened, a move that sources say accelerated OpenAI's decision to rebuild Codex in‑house.

How we got here

OpenAI first showed Codex in 2021, then handed the related Copilot product to GitHub and its largest backer, Microsoft. That transfer, and OpenAI's subsequent redirection of engineering talent toward DALL·E 2 and GPT‑4 as the company chased broad, AGI‑style capabilities, left the coding niche under‑resourced. ChatGPT's meteoric 2022 rise further concentrated effort on general chat models. Meanwhile, startups such as Cursor and teams at Anthropic focused on the messy, real‑world codebases that make for robust developer tools — not just neat academic benchmarks — and found product‑market fit faster.

Product and strategic differences

Claude Code shifts programming from “autocomplete” to an agentic model that can enter terminals, run code and manage project workflows — a qualitative leap from the Copilot era. Internal demos like “Jam” and the so‑called vibe‑coding movement reframed expectations: the AI is not merely a suggestor but an executor. Microsoft’s strategic interest in protecting Copilot, and its reported demand for Windsurf IP access, highlights how investor ties and platform control can shape which models flourish. It has been reported that these tensions are a key reason OpenAI moved to rebuild Codex internally.

Why it matters

This is more than a developer tools story. The shift illustrates how early leadership in one AI frontier (chat) can come at the expense of another (programming), and how investor and platform dynamics — think Microsoft’s influence over GitHub — can determine market outcomes. For Western readers less familiar with China's tech landscape: Chinese cloud and AI firms such as Baidu (百度) are watching and will adapt features competitively, but this episode mostly reflects dynamics inside the U.S. AI ecosystem. Can OpenAI reclaim the programming crown? That question will determine which firms control developer workflows — and the commercial engine of next‑generation software.