RepoJournal
OpenAI OpenAI
110 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 2026-09-05
stories 152
RSS Archive Topics

© 2026 RepoJournal Home Showcase Explore How it works Privacy

$ the-wire · showcase

OpenAI ships retry fix, prompt cache diagnostics, sandbox security

By RepoJournal · Filed · About OpenAI

copyberry Hayden markstuart-oai apcha-oai
4 people shipped this
copyberry @copyberry 5 cited Hayden @HAYDEN-OAI 5 cited markstuart-oai @markstuart-oai 2 cited apcha-oai @apcha-oai 1 cited

Python SDK users get safer retry handling and prompt cache visibility, while Codex tightens Windows sandboxing and Guardian review behavior.

The Python SDK now refuses overflowing server retry delays, returning the original API error instead of retrying early with a short fallback backoff [1]. This affects both synchronous and asynchronous requests; the existing 120-second limit and header precedence are preserved. Separately, the beta Responses API gains prompt cache diagnostics [2], exposing typed outcomes and cache miss reasons. Codex also preserves Windows managed deny reads [3], ensuring permission profiles aren't discarded when starting a sandbox session; this is a breaking change for users relying on over-permissive sandboxes.

Codex Guardian reviews now preserve root authorization context even after parent context compaction [4], making approvals stale when the root authorization changes. Additionally, Guardian approvals route independently of the async scorer, remaining available even if scoring fails [5]. Managed WebMCP policy is exposed through the app server [6], and `codex doctor` now reports the managed filesystem policy source [7]. The openai-node SDK fixes a runner cancellation bug where aborting during a final `afterCompletion` callback could falsely report success [8], and corrects Edge runtime detection to recognize `Edg`, `EdgA`, and `EdgiOS` tokens [9], fixing a breaking change for Edge users who were previously misreported as a different browser.

Documentation fixes clarify error handling fields and example output draining [10][11], and a test CLI now rejects unknown ecosystem project names instead of silently ignoring them [12]. The prompt cache comparison also requires a migration if you've been reading the removed event field, which was matched to the API spec [13].

Action items

References

  1. [1] fix: refuse overflowing server retry delays — SDK-235 ↗ openai/openai-python
  2. [2] feat(api): Add prompt cache diagnostics ↗ openai/openai-python
  3. [3] Preserve Windows managed deny reads in the sandbox CLI ↗ openai/codex
  4. [4] Preserve root authorization context in Guardian reviews ↗ openai/codex
  5. [5] Route Guardian approvals independently of async scoring ↗ openai/codex
  6. [6] Expose managed WebMCP policy through the app server ↗ openai/codex
  7. [7] Report managed filesystem policy in `codex doctor` ↗ openai/codex
  8. [8] fix(runners): observe cancellation after completion callbacks (#2591) ↗ openai/openai-node
  9. [9] fix: identify modern Edge browser runtime headers (#2628) ↗ openai/openai-node
  10. [10] docs: correct error handling diagnostics (#2630) ↗ openai/openai-node
  11. [11] fix(examples): drain manual conversation error output (#2629) ↗ openai/openai-node
  12. [12] fix(test): reject unknown ecosystem project names (#2626) ↗ openai/openai-node
  13. [13] feat(api): correct function argument completion event fields (openapi-545) (#3801) ↗ openai/openai-python

$ ls openai/reviews/ # the step-back reads

the week in review Codex hardens sandbox and auth across platforms the month in review August: Codex hardens, SDKs consolidate security, GPT-5.5 lands
Topics: AI / ML Full archive →

Quick answers

What shipped in OpenAI on September 5, 2026?
Python SDK users get safer retry handling and prompt cache visibility, while Codex tightens Windows sandboxing and Guardian review behavior. In total, 75 commits, 75 pull requests, and 2 releases landed.
Who contributed to OpenAI on September 5, 2026?
4 developers shipped this update, including copyberry, Hayden, markstuart-oai, and apcha-oai.
What were the notable OpenAI updates?
fix: refuse overflowing server retry delays — SDK-235, feat(api): Add prompt cache diagnostics, and Preserve Windows managed deny reads in the sandbox CLI.
♥

$ log OpenAI --recent

Codex TUI and openai-node fix stale state and timers 66 shipped Codex voice-host ships real-time microphone audio; openai-node fixes streaming hangs 108 shipped macOS Codex sandbox hardened against terminal input injection 174 shipped Windows sandbox provisioning goes live, SDKs gain canonical security models 107 shipped

$ log --all # elsewhere on the wire

Meta React Native adds Stable API guards to renderer attributed string FastAPI & Pydantic Pydantic fixes Mypy plugin crash, tightens datetime constraints

$ ls showcases/

Rails Django Laravel Node.js Go Vercel Vue.js Kubernetes HashiCorp Arch Linux Linux Anthropic

$ init your-desk

RepoJournal can write one of these for any repo you care about. Sign in with GitHub, pick your repos and sources, and your desk gets a wire like this one every morning.