$ cat openai/week/2026-10-26.log
the week in review · Oct 26 – Nov 1, 2026
Codex 0.165.0 adds Vertex AI provider and GPT-6.1 Terra
By RepoJournal · composed from the cited sources · human-reviewed weekly · methodology
Both SDKs widened Live Decision item types, openai-node shipped 7.44.0 automations, and openai-python fixed binary frame loss on WebSocket reconnect.
0.165.0 openai/codex
Set `model_provider = "gcp"` to route Codex through a built-in Vertex AI provider, and GPT-6.1 Terra is now in the model catalogs. The same release adds a repository filter to the session picker and directory grouping in the agent command center.
feat(websocket): recover queued binary frames after reconnect (#4191) openai/openai-python
On reconnect, `send_raw(bytes)` used to decode queued bytes as UTF-8: valid UTF-8 payloads went out as text frames and other bytes raised `UnicodeDecodeError` before anything was sent. Queued binary frames now survive the reconnect as binary.
release: 7.44.0 (#3025) openai/openai-node
Adds `client.beta.agents.automations` for scheduled automations, plus delegation controls and prompt cache lifetimes. If you rely on scheduled agent runs, the entry point changed.
feat(api): expand Live Decision item and question types openai/openai-node
Live Decision items now accept user, developer, system, and assistant messages, along with function and custom tool call and output types, plus strings or structured instruction parts. Teams sending tool calls through decisions can stop flattening them into plain messages.
Preserve reasoning token accounting across resume and fork openai/codex
Resuming or forking a thread dropped the flag recording that reported token usage already includes retained reasoning, so the reasoning was counted twice and triggered automatic compaction early. The flag is now persisted across both paths.
Add per-chat leases for shared MCP connections (#57978) openai/codex
For pooled stdio MCP servers that advertise the `codex/connection-lease` extension, Codex opens a unique lease per chat and attaches its host-owned ID to tool calls and resource reads via `codex/connectionLeases`.
chore: enforce generated-file ownership openai/openai-node
Generated files no longer get customization budgets: fix the inputs or compiler and regenerate, or put custom behavior in separately owned handwritten extensions. Hand edits to generated code are expected to fail review.
$ 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.