$ cat openai/week/2026-11-09.log
the week in review · Nov 9 – Nov 15, 2026
Codex 0.168.0 adds model provider picker and stable untrusted context
By RepoJournal · composed from the cited sources · human-reviewed weekly · methodology
gpt-6-rosalind appears in openai-node's beta Responses model types, and both SDKs now type async function tools.
0.168.0 openai/codex
by github-actions[bot]
Sessions now lists model providers with a per-chat default, and `features.tui_workspace` puts a side conversation beside the main chat, resizable by dragging the divider or zooming with `leader z`.
Enable untrusted context tool outputs by default openai/codex
`untrusted_as_tool_output` is stable and on by default. Browser information, goal objectives, and other untrusted context reach the model as `function_call_output` items in the `agent` namespace, so anything parsing the request stream should expect that shape.
Treat goal objectives as untrusted context openai/codex
Goal objectives now travel as untrusted `goal_objective` context rather than being embedded in continuation instructions, and the model is told to call `get_goal` when the objective is missing or truncated.
feat(api): recognize gpt-6-rosalind in beta Responses (#3126) openai/openai-node
gpt-6-rosalind is now among the known beta Responses model values in openai-node, which means typed callers can name it without a cast.
feat(api): add typed async function tools openai/openai-node
Function-tool request and response types for sessions and persisted agents gain optional `async` booleans, giving callers typed access to a setting they previously reached for untyped.
Support gateway device-code sign-in in headless TUI sessions openai/codex
Gateway sign-in now prefers device authentication in SSH sessions and headless Linux environments, and app-server gateway OAuth advertises `supportsDeviceAuth` and returns a `verificationUrl`, `userCode`, and `loginId` before approval.
Limit MCP product attribution to configured REPL-backed calls (#60697) openai/codex
Turn metadata no longer forwards a client-supplied `product_sku`, which previously could reach unrelated servers and connectors; attribution is limited to configured REPL-backed calls.
Enable model and reasoning effort selection for Cloud conversations openai/codex
Cloud conversations fetch the account-scoped catalog on demand, so model and reasoning effort can be chosen from the TUI instead of defaulting to whatever the activation returned.
$ ls openai/week/ # the briefings behind this review
Keep up with OpenAI in about 3 minutes: what actually shipped — the commits, pull requests, releases, and security advisories that matter.
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.