RepoJournal
OpenAI

OpenAI

Codex, the SDKs, and the engine behind ChatGPT

Keep up with OpenAI in about 3 minutes a day: what actually shipped — the commits, pull requests, releases, and security advisories that matter.

One email a day. Unsubscribe in one click.

Pick a date

Topics: AI / ML Full archive →

The Wire · Showcase

CODEX TIGHTENS MEMORY WITH MOVE-ONLY SUBMISSIONS AS NODE SDK COMPLETES LINTING OVERHAUL

By RepoJournal · Filed · About OpenAI

Codex eliminated cloning overhead from submission operations while the Node SDK wrapped up a 185-part linting cleanup that touches every rule in the compatibility matrix.

Codex's top change removes Clone trait requirements from Submission and Op types [1], consuming operations directly instead of copying them through the submission loop. This is a memory efficiency win for high-throughput submission paths. Two security fixes landed alongside: the Windows sandbox backend now respects configured sandbox levels and rejects managed networking when restricted tokens are in use [2], and Statsig exports now exclude internal metrics like conversation turn count and response-engine latency [3]. The SDK also extracted persisted history types into a dedicated crate [4], cleaning up the domain boundary between protocol and runtime state. Over on openai-node, Hayden pushed five consecutive linting PRs (parts 145-149 of a 185-part stack) enabling object-shorthand, arrow-body-style, unicorn/no-useless-undefined, unicorn/prefer-module, and no-shadow rules [5] [6] [7] [8] [9]. Each validates cumulatively at the stack head with full lint and test coverage. The Python SDK regenerated its API surface and updated generated-file headers to credit Castiron [10]. Across all three repos: 92 commits, 92 PRs, 2 releases in the last 24 hours.

One email a day. Unsubscribe in one click.

Action items

References

  1. [1] Make submission operations move-only (#37901) openai/codex
  2. [2] Honor the configured Windows sandbox level for managed networking ↗ openai/codex
  3. [3] Keep runtime summary metrics out of Statsig exports ↗ openai/codex
  4. [4] Extract persisted history types into a dedicated crate ↗ openai/codex
  5. [5] chore(lint): enable object-shorthand (#2237) openai/openai-node
  6. [6] chore(lint): enable arrow-body-style (#2238) openai/openai-node
  7. [7] chore(lint): enable unicorn/no-useless-undefined (#2239) openai/openai-node
  8. [8] chore(lint): enable unicorn/prefer-module (#2240) openai/openai-node
  9. [9] chore(lint): enable no-shadow (#2241) openai/openai-node
  10. [10] chore(api): Update generated-file header attribution to Castiron ↗ openai/openai-python

Quick answers

What shipped in OpenAI on August 11, 2026?
Codex eliminated cloning overhead from submission operations while the Node SDK wrapped up a 185-part linting cleanup that touches every rule in the compatibility matrix. In total, 92 commits, 92 pull requests, and 2 releases landed.
Who contributed to OpenAI on August 11, 2026?
5 developers shipped this update, including Owen Lin, copyberry, cooper-oai, Hayden, and apcha-oai.
What were the notable OpenAI updates?
Make submission operations move-only (#37901), Honor the configured Windows sandbox level for managed networking, and Keep runtime summary metrics out of Statsig exports.

More from OpenAI

Daily updates, in your inbox

Follow OpenAI

Keep up with OpenAI in about 3 minutes a day: what actually shipped — the commits, pull requests, releases, and security advisories that matter.

or

One email a day. Unsubscribe in one click. Read a past issue →

Elsewhere on the wire

Want every project, not just this one?