ToolsThe story, in brief

Addendum to o3 and o4-mini system card: Codex

OpenAI just shipped Codex: a cloud coding agent powered by o3, trained on real-world PRs. It writes code that passes tests.

Paper-cut illustration of a coral software window opening into a three-dimensional drafting space.
New tools for building and creating with AI.AI illustration by KeyNews
The KeyNews take

Why it matters

OpenAI is moving o3 from research artifact to production product layer. Codex represents the shift from model capability claims to deployed developer tooling—and signals how o3's reasoning is being monetized in app-layer products.

The key facts

5 to know
  1. Codex is a cloud-based coding agent powered by codex-1 (o3 variant optimized for software engineering)

  2. codex-1 trained using reinforcement learning on real-world coding tasks

  3. Agent trained to generate code matching human style and PR preferences

  4. Iterative test-driven approach built into agent behavior

  5. Product announcement published May 16, 2025 as system card addendum

Go to the source

OpenAI Blogopenai.com

Publisher excerpt: Codex is a cloud-based coding agent. Codex is powered by codex-1, a version of OpenAI o3 optimized for software engineering. codex-1 was trained using reinforcement learning on real-world coding tasks in a variety of environments to generate code that closely mirrors human style and PR preferences,…
Read original report
Back to today's editionMore tools news

The wider picture

View all
Paper-cut illustration of a coral software window opening into a three-dimensional drafting space.
AI illustration by KeyNews
Tools01

The Genie One MCP is now Generally Available

Genie One MCP is a production-ready tool layer for integrating coding agents into enterprise workflows. Practitioners building agentic systems now have a standardized, vendor-backed protocol for connecting agents to IDEs and development environments — lowering friction from pilot to deployment.

Databricks
Illustration of independent geometric mechanisms passing paper tasks along branching amber tracks.
AI illustration by KeyNews
Tools02

Rabbit Is Back, This Time With an AI Agent App

A failed AI hardware play pivots to cross-platform agent software — a test case for whether agent UX can drive mainstream adoption outside dedicated devices.

Wired AI
Illustration of a transparent lens revealing connected networks across layers of paper.
AI illustration by KeyNews
Tools03

llm-typesafe 0.1a0

A new Python library for enforcing type-safe outputs from LLMs—useful for practitioners building production applications where unpredictable output shapes break downstream code. Early alpha, but addresses a real friction point in AI app development.

Simon Willison