173 wires and counting

$ follow OpenAI

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

or

fair warning: these emails are deeply technical. diffs, version numbers, CVEs, benchmark deltas. if that's not your idea of a good read, this isn't your newsletter.

Folds into your digest — weekly by default, monthly if you prefer. Unsubscribe in one click.

$ status

wire 2026-10-31
stories 109

© 2026 RepoJournal Home Showcase How it works Privacy

$ the-wire · showcase

Codex hardens auth refresh and context accounting

By RepoJournal · Filed · About OpenAI · Composed from the cited sources · methodology

Codex stopped treating a corrupt auth.json as a permanent account mismatch and fixed context token estimates that were delaying automatic compaction, while the Node and Python SDKs moved agent session review, tool replacement, and typed output formats forward.

Treat auth reload failures as transient during token refresh openai/codex

by copyberry

Codex used to let an unreadable or malformed auth.json surface as an account mismatch during token refresh or unauthorized recovery, which produced a permanent failure; storage errors from guarded reloads now propagate as transient, keeping cached credentials so recovery can retry, and credential-safe error formatting covers file deserialization and agent identity JWT decoding.

Fix context token estimates for responses without usable usage openai/codex

by copyberry

Output retained from failed responses or responses without usage data could be left out of context token estimates, which delayed automatic compaction while reasoning kept accumulating. Codex now tracks the history boundary covered by the last usable usage report and estimates every item added after it, with spending-only updates preserving that boundary.

feat(api): add agent session review and tool settings openai/openai-python

by markstuart-oai

Sessions now accept review settings on create and update, expose optional session review values, and let you replace tools wholesale through agent.tools on a session update. The same change landed in openai-node, so both SDKs describe one API surface.

feat(api): parse Responses using the latest text format update openai/openai-node

by markstuart-oai

The Node parse and stream helpers infer their output type from an inline format update, covering Zod and Standard Schema formats, and a later plain-text update stops structured parsing; when the server knows the format from a previous response it passes parseAs in the second options argument, local to the SDK.

feat(api): support WebSocket warmup and preserve cleanup cancellation openai/openai-python

by markstuart-oai

Python gained an optional generate field on stable and beta Responses WebSocket create events, unset by default and sent on the wire only when explicitly set to true or false. Cancellation raised while a WebSocket connection closes now survives even when the context body already raised, and the Realtime, translation, Live, and Responses cleanup paths still surface the original body error when o...

Add scoped host metadata hooks for MCP tool requests openai/codex

by copyberry

The long tail: codex-message-board-mcp exposes message-board tools through rmcp::ServerHandler with an McpToolRequestMetaHook for per-invocation _meta, openai-python 3.42.0 adds agent context compaction settings, and openai-node 7.47.0 supports typed Responses WebSocket warmup.

Quick answers

What shipped in OpenAI on October 31, 2026?
Codex stopped treating a corrupt auth.json as a permanent account mismatch and fixed context token estimates that were delaying automatic compaction, while the Node and Python SDKs moved agent session review, tool replacement, and typed output formats forward. In total, 53 commits, 53 pull requests, and 3 releases landed.
Who contributed to OpenAI on October 31, 2026?
4 developers shipped this update, including Hung, copyberry, openai-sdks[bot], and markstuart-oai.
What were the notable OpenAI updates?
Treat auth reload failures as transient during token refresh, Fix context token estimates for responses without usable usage, and feat(api): add agent session review and tool settings.