{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreigywejg2j64gk6mrmvwnxulkpwvrbuccaerugjebmopjoidhzvofa",
    "uri": "at://did:plc:lk3jfj3zq4k4wxnk474axylu/app.bsky.feed.post/3mfjuvalw3ps2"
  },
  "path": "/t/tips-and-tricks-for-using-codex/1373143#post_20",
  "publishedAt": "2026-02-23T13:55:55.000Z",
  "site": "https://community.openai.com",
  "tags": [
    "Long horizon tasks with Codex",
    "OpenAI Cookbook Codex example",
    "Derrick Choi",
    "Takeaways for long-horizon Codex tasks"
  ],
  "textContent": "> The practical change is that agents can stay coherent for longer, complete larger chunks of work end-to-end, and recover from errors without losing the thread.\n\nLong horizon tasks with Codex\n\nLong horizon tasks is an OpenAI Cookbook Codex example by Derrick Choi\n\n### Takeaways for long-horizon Codex tasks\n\nWhat made this run work was not a single clever prompt. It was the combination of:\n\n  * A clear target and constraints (spec file)\n  * Checkpointed milestones with acceptance criteria (`plans.md`)\n  * A runbook for how the agent should operate (`implement.md`)\n  * Continuous verification (tests/lint/typecheck/build)\n  * A live status/audit log (`documentation.md`) so the run stayed inspectable\n\n\n\nThis is the direction long-horizon coding work is moving toward: less babysitting, more delegation with guardrails.",
  "title": "Tips and Tricks for using Codex"
}