Codex desk
OpenAI · Codex updates
A focused desk for OpenAI Codex updates, workflows, repositories, and practical engineering notes.
Track the public data: Codex vs Claude Code Indicator →
Hot news
Featured guides
11 min readFrom Thousands to Millions of Tokens: Do Context Windows Still Need Optimization?A source-backed history of LLM context windows from GPT-3 to million-token models—and a practical guide to deciding when to send everything, retrieve, or summarize.
10 min readHow to Use REA with Claude Code and Codex: Inspect, Rebuild, and VerifyConnect REA to Claude Code or Codex, trace an Electron export, verify recovered behavior, and turn the evidence into a tested implementation.
9 min readCan AI Agents Test Your App Like a Human? OpenAI’s New Computer Use ExplainedOpenAI added a hosted browser to the Agents API. Learn how agents can exercise web-app flows, what evidence to collect, and where automated tests and human approval still matter.
10 min readGraph Engineering for AI Coding Agents: Graphify, Claude Code, and CodexUnderstand code knowledge graphs and agent workflow graphs, then use Graphify to explore a repository in Claude Code or Codex with evidence-backed examples.
12 min readHow to Evaluate a Coding Agent: Traces, Tests, Reverts, and Cost per Merged PRA source-backed framework for evaluating coding agents on real repository work, from traces and tests to post-merge rework and fully loaded cost.
15 min readBuild a PR Risk Gate with Jev: Typed AI Decisions for Safer Code ReviewsA practical TypeScript guide to classifying pull-request risk with Jev while deterministic policy and human reviewers retain control.