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 2026-10-29
stories 133

© 2026 RepoJournal Home Showcase How it works Privacy

$ the-wire · showcase

Codex fixes token accounting and lease isolation, SDKs document model shutdowns

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

Codex closed two thread-lifecycle leaks that distorted context budgeting, while openai-python put shutdown dates in the type signatures and openai-node exposed hosted tool usage separately from model tokens.

chore(api): document upcoming model shutdowns (#4199) openai/openai-python

by Justin Beckwith

The python SDK's model enums now carry announced shutdown dates with links to replacement guidance, and audio translations are deprecated with migration to Realtime translations before February 26, 2027; calling them emits the standard `DeprecationWarning`. Existing model values stay available, so nothing breaks today, but the migration deadline is now written into the types your editor shows you.

Preserve reasoning token accounting across resume and fork openai/codex

by copyberry

Resuming or forking a thread dropped the flag recording that reported token usage already includes retained reasoning, so the reasoning was counted twice and automatic compaction fired too early. `last_usage_includes_reasoning` is now persisted with `TokenUsageInfo` and recorded alongside completed sampling and compaction usage.

Validate thread providers before resume and input openai/codex

by copyberry

A saved chat keeps its provider configuration when the provider selection changes, and resume and input are now rejected until the two match, with history remaining readable and recovery possible once configuration is restored. Provider validation runs before resume overrides are applied, and model or reasoning-only changes preserve the saved provider.

Add per-chat leases for shared MCP connections (#57978) openai/codex

by Charlie Marsh

Pooled stdio MCP connections now negotiate a `codex/connection-lease` extension: each chat opens a unique lease whose host-owned ID is attached to tool calls and resource reads via `codex/connectionLease`, overriding caller-supplied lease metadata. Releasing a lease writes cancellations for pending requests before `codex/connection/release`.

feat(api): add SIP DTMF commands and delivery events openai/openai-node

by jbeckwith-oai

SIP sideband keypresses ship as typed `transport.dtmf.send` commands with optional event correlation and `transport.dtmf.sent` acknowledgements, and the legacy notification types are retained. Also in openai-node, hosted tool usage is available in stable and beta Responses separately from model token usage, and organization external keys can be registered, validated, listed and filtered across ...

Quick answers

What shipped in OpenAI on October 29, 2026?
Codex closed two thread-lifecycle leaks that distorted context budgeting, while openai-python put shutdown dates in the type signatures and openai-node exposed hosted tool usage separately from model tokens. In total, 65 commits, 65 pull requests, and 3 releases landed.
Who contributed to OpenAI on October 29, 2026?
5 developers shipped this update, including copyberry, Charlie Marsh, jbeckwith-oai, markstuart-oai, and Marcus Wood.
What were the notable OpenAI updates?
chore(api): document upcoming model shutdowns (#4199), Preserve reasoning token accounting across resume and fork, and Validate thread providers before resume and input.