Who contributed to OpenAI on June 16, 2026?

1 developer shipped this update, including mzeng-openai.

What were the notable OpenAI updates?

avoid cloning sampling request input (#28306), Extract shared Windows sandbox session runner (#28357), and Add realtime speech append control (#27917).

OpenAI

Codex, the SDKs, and the engine behind ChatGPT

Pick a date

Topics: AI / ML Full archive →

The Wire · Showcase

CODEX CUTS REQUEST CLONING OVERHEAD WHILE HARDENING SANDBOX AND SPEECH CONTROLS

By RepoJournal · Filed 06:05 UTC on June 16, 2026 · About OpenAI

1 person shipped this

mzeng-openai @mzeng-openai 1 cited

A critical performance fix removes O(n) clones from every model request, while three parallel hardening efforts shore up Windows sandbox isolation, resource caching, and policy evaluation.

The biggest win lands in sampling: Codex now avoids cloning the full prepared input on every request [1], eliminating expensive copies that grow with conversation size. Instead, the input moves into the sampling loop and returns with results, preserving backward compatibility for the legacy after-agent hook. On the security front, Windows sandbox session management gets refactored [2] to extract reusable launch logic before a follow-up PR ships the fs-helper fix, making the eventual security review cleaner. Realtime speech gains explicit app-side control [3] so backends stop over-chatting by automatically speaking every preamble and progress update. Resource caching now lives at the thread level [4], eliminating repeated remote MCP fetches for the same orchestrator skill within a single thread and preventing prompt injection from observing stale catalog state. Finally, Apps policy evaluation consolidates into connectors [5], moving the evaluator out of codex-core to resolve one immutable policy snapshot per exposure build instead of rebuilding on every sampling request.

Action items

→ Review and merge cloning fix [ref:4] before next deployment cycle openai/codex [immediate]
→ Stage Windows sandbox refactor [ref:1] for security fix integration openai/codex [plan]
→ Monitor thread-level caching [ref:2] for MCP latency improvements in production openai/codex [monitor]

References

[1] avoid cloning sampling request input (#28306) openai/codex
[2] Extract shared Windows sandbox session runner (#28357) openai/codex
[3] Add realtime speech append control (#27917) openai/codex
[4] skills: cache orchestrator resources per thread (#28336) openai/codex
[5] [codex] Reuse Apps policy evaluation across MCP tool exposure ↗ openai/codex

Quick answers

What shipped in OpenAI on June 16, 2026?: A critical performance fix removes O(n) clones from every model request, while three parallel hardening efforts shore up Windows sandbox isolation, resource caching, and policy evaluation. In total, 50 commits, 49 pull requests, and 6 releases landed.
Who contributed to OpenAI on June 16, 2026?: 1 developer shipped this update, including mzeng-openai.
What were the notable OpenAI updates?: avoid cloning sampling request input (#28306), Extract shared Windows sandbox session runner (#28357), and Add realtime speech append control (#27917).

CODEX FIXES THREAD REPLAY ERRORS AND TIGHTENS SHUTDOWN SEQUENCING

Failed turn completions were being silently restored as successes when replaying conversation history, masking critical errors in the TUI transcript.

ai 10 shipped 1-min read

OpenAI 2 days ago

CODEX SHIPS RESPONSIVE UI FIXES AND FRESHENS MENTION SEARCH STATE

Codex tightened its interface handling across four critical flows: responsive keymap menus now adapt to narrow terminals, mention popups purge stale search results, completion events now track item lifecycles, and exec-server network policy decisions move client-side.

ai 15 shipped 1-min read

OpenAI 3 days ago

CODEX HARDENS MCP RUNTIME STABILITY ACROSS FIVE CRITICAL FIXES

OpenAI shipped a coordinated hardening of the Model Context Protocol layer across Codex, patching race conditions in thread startup, resource access, and authority management that could leave stale configurations in production.

ai 66 shipped 1-min read

OpenAI 4 days ago

SDK RELEASES LOCK IN PROMPT CACHING AND SPEND CONTROLS

Both Node and Python clients shipped v6.49.0 and v2.48.0 overnight with admin spend limit APIs and relaxed prompt cache validation.

ai 103 shipped 1-min read

Elsewhere on the wire

AI Agents about 21 hours ago

CLINE SHIPS FIVE CRITICAL AGENT FIXES AFTER END-TO-END TESTING BLITZ

Saoudrizwan just landed a cascade of fixes that prevent the agent loop from running wild, stop file reads from resolving to the wrong directory, and plug the compaction pipeline.

ai-agents 16 shipped 2-min read

Local LLMs about 21 hours ago

LLAMA.CPP SHIPS MINIMAX-M3 SPARSE ATTENTION, VLLM UNLOCKS HETEROGENEOUS PREFILL-DECODE

The inference stack gets a new model architecture and a fundamental shift in how prefill and decode can run on different hardware configurations.

+10

llms 125 shipped 1-min read

@CachyOS about 21 hours ago

GLIBC 2.43 REBUILD SHIPS, INSTALLER JUMPS TO C++26

CachyOS pushed glibc 2.43 through the pipeline while the CLI installer upgraded its C++ standard to C++26, signaling a major refresh across core infrastructure.

infra 97 shipped 1-min read

Elixir & Phoenix about 21 hours ago

ECTO FIXES SUBQUERY PLANNING IN JOINS, ELIXIR PATCHES IEX PARSING

Ecto's query planner now handles subqueries correctly in interpolated join filters, while Elixir fixes a syntax error that broke 'not in' as a continuation operator in IEx.