OpenAI merges ChatGPT and Codex under Greg Brockman

OpenAI is bringing ChatGPT, Codex, and its developer API under a single team. Greg Brockman, the company's co-founder and president, is taking full control of product strategy in an effort to end the fragmentation that had been slowing development down.

One agentic platform instead of a patchwork quilt

In an interview with Wired, Brockman explained the reasoning behind the merger: OpenAI is investing in a single agentic platform, with ChatGPT (consumer-facing) and Codex (developer-facing) serving as interfaces to the same foundation. Until now, the two products evolved in parallel. Each had its own product team, its own roadmap, and often a diverging vision of the future.

Now everything sits under one roof, and a single team owns the full path from idea to production-ready solution. That lets the company keep development in sync:

• Ship base model updates to the consumer product and the developer API at the same time
• Keep one product from falling behind the other on features
• Move insights and learnings between development tracks faster
• Simplify the architecture and cut the tech debt that builds up from duplicated logic

Brockman uses a telling phrase: "side quests are over." OpenAI used to run too many parallel projects and experiments (Whisper for speech, DALL-E for images, GPTs for customization, Canvas for integration), which pulled attention away from the main goal and diluted the company's focus.

From experiments to a single direction

Over the past three years, OpenAI has shipped a wide range of products, from speech input (Whisper) and image generation (DALL-E) to custom GPTs and the Canvas integration. Each project demanded its own resources, team, and constant attention. Through all of it, though, the company's main direction stayed clear: agents.

Agents are systems that make decisions and act on their own without human intervention: they click through websites, fill out forms, write and run code, and look up information. What sets them apart from ordinary chat interfaces is that they can carry out multistep tasks autonomously, double-checking their results and adapting to unexpected obstacles.

Putting ChatGPT and Codex under one team signals that the agent-centric approach is a company-wide architectural decision, not merely a new feature. An agent-first approach demands deep consistency: an agent running in consumer ChatGPT needs access to the same set of tools and logic as an agent in the developer API.

Codex was the ideal entry point for developers who want to build automated code generation into their apps and services. But it evolved separately, with its own owner and its own roadmap. That slowed innovation and created confusion across the ecosystem.

What this means for the ecosystem

For developers, the merger could significantly simplify integration: instead of different APIs and interfaces for ChatGPT and Codex, there will be one platform, one set of documentation, and one set of abstractions for working with agents.

For competitors (Anthropic, Google, Meta), it's another clear sign that OpenAI has picked a strategic direction instead of spreading its resources thin. Anthropic has Claude and the Claude API, but no explicit unification around agents. Google has the Gemini API and MakerSuite, but its structure remains fragmented. Meta is investing in LLaMA, but doesn't yet have a unified agent frontend.

OpenAI has chosen focus: one clear priority instead of a pile of parallel experiments. And that priority is agents that work everywhere.

*Meta is designated an extremist organization and banned in Russia.