OpenAI's first post in a series explaining the internals of the Codex CLI agent loop — the core harness that orchestrates interactions between the user, model, and tools across Codex CLI, Cloud, and VS Code extension. The post covers the agent loop architecture with lessons learned since the April launch, referencing their open-source repo where design decisions are documented in issues and PRs. Directly actionable for engineers building or reasoning about agentic coding systems.
Infrastructure
Unrolling the Codex agent loop
OpenAI's first post in a series explaining the internals of the Codex CLI agent loop — the core harness that orchestrates interactions between the user, model, and tools across Codex CLI, Cloud, and VS Code extension. The post covers the agent loop architecture with lessons learned since the April launch, referencing their open-source repo where design decisions are documented in issues and PRs. Directly actionable for engineers building or reasoning about agentic coding systems.
Saturday, March 21, 2026 12:00 PM UTC2 MIN READSOURCE: OpenAI BlogBY sys://pipeline
Tags
infrastructure
/// RELATED
Infrastructure3d ago
Show HN: Pollen – distributed WASM runtime, no control plane, single binary
Distributed WASM runtime using CRDT consensus hits 4000 req/s across 10 nodes with no control plane, proving decentralized infrastructure architecture can scale.
Safety3d ago
Meta’s historic loss in court could cost a lot more than $375 million
Meta faces a New Mexico public nuisance trial that could impose far costlier operational mandates than its $375M settlement—including age verification, encryption restrictions for minors, aggressive CSAM detection, and usage caps—setting industry precedent.