Codex v0.141.0 Ships Record & Replay and Noise Relay Remote Execution; v0.142.0 Alphas Roll Daily

Codex v0.141.0: Noise relay remote execution, MCP per thread, MCP marketplace, Record & Replay macOS skill (June 18)

• Released June 18, v0.141.0 is the most feature-dense Codex release in months: authenticated Noise relay channels enable authenticated cross-platform remote code execution; MCP context is now configurable per thread rather than globally; an in-app MCP marketplace lets you discover and add servers without manual config; realtime speech controls land for hands-free sessions.
• The companion app changelog (26.616) adds Record & Replay for macOS - demonstrate any workflow once and Codex converts it into a reusable skill - alongside bulk automation actions and thread handoff between hosts.

Editor’s read: Record & Replay is the standout: instead of prompting the same multi-step workflow repeatedly, you show it once and it becomes a persistent, shareable skill - test it with your most repetitive coding tasks.

Codex 0.142.0-alpha.1 through alpha.10: ten pre-release iterations ship in three days (June 18-21)

• Starting June 18, the day v0.141.0 landed, OpenAI began shipping daily pre-release alphas for v0.142.0, reaching alpha.10 by June 21; the rapid cadence signals active development of the next stable release, though no feature manifest has been published for the alpha series.

Codex v0.140.0 stable (June 15)

• v0.140.0 shipped on June 15 as the last stable before the v0.141.0 major drop, capping the previous development cycle.

levelsio on what to build as an indie developer in 2026 (June 15)

• On June 15, Pieter Levels shared his framework for indie developer positioning in 2026, arguing that niche specificity matters more than ever as tooling parity closes and generic AI-assisted products converge.

Editor’s read: Read this before you start your next side project - the window for generic AI wrappers is closing fast.

Rauchg: Vercel launches Fluid compute on microVMs, no function runtime ceiling (June 15)

• Guillermo Rauch announced Vercel Fluid compute on June 15.
• microVM-based infrastructure removing the traditional serverless runtime ceiling, which is directly relevant for agentic Codex workflows that run longer than typical serverless function limits allow.

SpaceX acquires Cursor for $60B in stock - largest VC-backed startup acquisition ever (June 16)

• On June 16, SpaceX announced the acquisition of Cursor, Codex's primary IDE competitor, for approximately $60 billion in SpaceX stock days after the company's blockbuster IPO; Cursor co-founder Michael Truell called it a chance to 'advance the frontier of useful AI' and the Cursor team committed to accelerating product development under SpaceX backing.
• The deal prompted community concern about Musk-era influence on product direction alongside cautious optimism about expanded compute and distribution.

Editor’s read: Cursor under SpaceX likely means more compute and distribution, not less product ambition - watch the roadmap closely over the next 60 days.

Cursor Compile 2026: Origin (agent-first git hosting), Frontier Model, and Cursor Mobile announced (June 16-17)

• At Cursor's Compile 2026 conference June 16-17, the team announced three products: Origin, an agent-first git hosting platform positioning as a GitHub alternative; a Cursor Frontier Model being trained in-house; and a Cursor Mobile iOS beta on TestFlight with remote Mac connection.

Editor’s read: Origin competing with GitHub is a serious bet - if Codex adds similar git-native collaboration, the winner will be whoever makes agent-driven PRs feel native.

Cursor 3.8: /automate skill, GitHub and Slack triggers, computer use for automations (June 18)

• Cursor 3.8 released June 18 adding the /automate skill that lets agents execute recurring workflows triggered by GitHub events or Slack messages, plus computer use capabilities for automation tasks.
• a direct capability overlap with Codex Record & Replay that launched the same day.

GLM-5.2: Z.ai's 753B open-weights model beats GPT-5.5 on coding benchmarks (June 17)

• Simon Willison covered Z.ai's GLM-5.2 on June 17.
• a 753-billion-parameter open-weights model that claimed the top spot on the Artificial Analysis leaderboard and outperformed GPT-5.5 on coding benchmarks, offering a self-hostable alternative at frontier quality.

Fable 5 still offline on Day 9 after US export control order (June 21)

• Anthropic's Fable 5, which launched June 9 to widespread developer acclaim, remains globally suspended on June 21 following a US government export control order issued June 12-13; an NSA breach claim has complicated restoration talks between Anthropic and Trump officials.

Microsoft AutoJack: a single malicious web page can hijack an AI coding agent and achieve RCE on the host (June 18)

• Microsoft disclosed AutoJack on June 18.
• a prompt-injection technique exploiting browser access in AI coding agents where a single malicious page can execute arbitrary code on the developer's machine; applicable to any Codex session with web access enabled.

Editor’s read: If you use Codex in a browser-integrated mode or have web fetch in your tools, read this and sandbox accordingly before your next session.