186 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 2026-11-14
stories 112

© 2026 RepoJournal Home Showcase How it works Privacy

$ the-wire · showcase

Codex persists realtime voice frames, SDKs widen agent types

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

Codex now flushes timestamped realtime frames before acking completion and exposes token and cost totals to statusline providers, while openai-node 7.57.0 and openai-python 3.51.0 ship typed surface area for Agents, Live, Images, and admin.

Persist timestamped realtime frames before acknowledging completion openai/codex

by copyberry

Voice transcripts for eligible threads are now recorded as `frame` items on a `voice_transcript` channel, with first-observed timestamps that never move backward and are restored from history and compaction checkpoints; legacy transcript support and the existing app-server timeline representation stay in place. Flushing realtime records before completion means a finished turn no longer acknowle...

Expose token usage totals in statusline state openai/codex

by copyberry

`thread.usage.total` lands in the provider-facing statusline snapshot with total input, cache-read input, cache-write input, other input, and output token counts, and cache-write counts now survive the app-server-to-TUI token report conversion. Totals are omitted when usage is unavailable, the breakdown is invalid, or the report is a synthetic context fill, but a measured zero is kept.

Expose thread costs in statusline state and reject stale usage reads openai/codex

by copyberry

Statusline state gains optional `thread.usage.cost.total_cost_usd` and `total_credits` using the built-in status line's plan-specific values, with micro-units converted to ordinary units and unavailable or negative values dropped. UUIDs replace thread-keyed reads, fixing the case where a replacement chat widget received a previous widget's usage result for the same thread.

Show the primary executor's workspace in TUI status surfaces openai/codex

by copyberry

The primary executor's working directory now carries through thread initialization, restoration, and settings updates, and drives the status line and terminal title. The app-server directory was being shown for workspaces that live elsewhere, and remote paths abridged against the TUI's home directory read as a different workspace than the one running.

feat(api): update Agents, Live, Images, and admin types openai/openai-node

by dpiet-oai

The typed update adds Agents MCP approval requests, prompt-cache lifetime settings, and event obfuscation controls, plus Live safety identifiers, the `live.transport.connected` webhook, SVG image-edit options, OCI external-key configuration, and API-key expiry audit fields. `beta.agents.sessions.events.stream(id, options)` still works alongside the new explicit `(id, query, options)` call form.

Quick answers

What shipped in OpenAI on November 14, 2026?
Codex now flushes timestamped realtime frames before acking completion and exposes token and cost totals to statusline providers, while openai-node 7.57.0 and openai-python 3.51.0 ship typed surface area for Agents, Live, Images, and admin. In total, 53 commits, 53 pull requests, and 6 releases landed.
Who contributed to OpenAI on November 14, 2026?
4 developers shipped this update, including copyberry, openai-sdks[bot], dpiet-oai, and openai-sdks.
What were the notable OpenAI updates?
Persist timestamped realtime frames before acknowledging completion, Expose token usage totals in statusline state, and Expose thread costs in statusline state and reject stale usage reads.