· By the ToolNav Team · 6 min read OpenAI Codex AI Coding AI Agents Developer Tools

OpenAI Codex ships stable Goal Mode, Appshots, and post-lock Computer Use

TL;DR

OpenAI's May 22 Codex update graduates Goal Mode from experimental to stable across the app, IDE extensions, and CLI, adds Appshots screen injection on macOS, and lets Codex continue Computer Use after your Mac locks — three changes that push agentic coding further toward unattended, long-horizon runs.

Stable

Goal Mode status — exits experimental across Codex app, IDE extensions, and CLI v0.128.0+

⌘⌘

Appshots trigger — double Command key on macOS injects frontmost window screenshot and text into Codex thread

Post-lock

Computer Use now continues after Mac screen locks, triggerable remotely from Codex Mobile on iOS/Android

May 22, 2026

Date of Codex changelog update shipping all three features

OpenAI shipped three notable Codex updates around May 22, 2026, per the official Codex changelog. Taken together, they push Codex meaningfully further into long-horizon, low-supervision agentic territory: Goal Mode is now stable, Appshots brings one-keystroke screen context injection to macOS, and Codex can keep running Computer Use tasks after your Mac screen locks.

Goal Mode exits experimental and goes stable. Goal Mode — OpenAI's long-horizon objective-setting mode for Codex — has graduated from experimental to generally available across the Codex app, VS Code and JetBrains extensions, and CLI (version 0.128.0 or later required). In Goal Mode, instead of issuing one instruction and waiting for a single output, you assign Codex an objective: a defined outcome, success criteria, and a scope of work. Codex then drives toward that objective autonomously across session breaks, token budget resets, and task interruptions, without requiring you to prompt it forward at each step. The company says this enables hours- or days-long autonomous coding runs. The "stable" designation means the feature is no longer marked experimental and can be treated as production-ready in supported environments — a meaningful threshold for teams evaluating whether to rely on it for production-track work.

Appshots: inject your screen into Codex with a double-tap. Appshots is a new macOS-only shortcut that injects the frontmost window directly into your active Codex thread. Press both Command keys at once — Codex captures a screenshot of the frontmost app and pulls any accessible text the app exposes, including content outside the visible scroll area, and attaches it to the current thread as context. The practical use case: share what you are actually looking at — a browser tab showing an error, a design file, a build log, a document — without manually screenshotting, cropping, and pasting. Appshots is macOS-only at launch; Windows and Linux availability has not been announced.

Post-lock Computer Use. Codex can now continue executing Computer Use tasks after your Mac screen locks. If you set a long-running Computer Use task and need to step away, Codex no longer stops when the screen locks. Combined with Codex Mobile — the iOS and Android app that can trigger and monitor Codex sessions remotely — you can start a task at your desk, lock your Mac, and track progress or redirect the task from your phone.

What is not changing. These are feature additions, not model upgrades. The underlying Codex model and its capabilities are unchanged by this update. Goal Mode's ability to complete a complex multi-day task still depends on the scope of the objective you set, the quality of your task definition, and the specifics of your codebase. Appshots requires macOS and frontmost-window accessibility. Post-lock Computer Use requires the same Computer Use setup as before — it does not unlock Computer Use for accounts or environments where it was not already configured.

Who it matters for. The biggest immediate beneficiaries are developers who currently run long or complex Codex sessions and have to actively babysit them — checking in, unblocking stuck steps, re-prompting when context runs thin. Goal Mode stable and post-lock Computer Use together remove most of that overhead. Appshots is narrower in impact: it removes a copy-paste friction point on macOS, which matters most if you frequently need to share screen state with an AI coding assistant mid-session. See our best AI coding tools roundup for how Codex compares to Cursor and Claude Code on agentic and long-horizon features, and our Cursor vs Claude Code comparison for the current head-to-head at the interface layer.

Why It Matters

Long-horizon autonomous coding is moving from experimental to everyday. Goal Mode going stable — available in production-grade form in the app, extensions, and CLI — means developers can reliably assign Codex multi-hour objectives without babysitting each step. Combined with post-lock Computer Use and Appshots screen injection, OpenAI is closing the gap between 'AI coding assistant that waits for you' and 'AI coding agent that runs while you're not looking.' For teams evaluating their agentic coding stack in mid-2026, Goal Mode stable is the most significant signal: OpenAI now has a production-ready long-horizon mode alongside Claude Code's agent loop and Cursor's background agent.

Who's Affected

  • Codex users running complex or long-duration tasks. Goal Mode stable is the most directly relevant change. If you have been using Goal Mode in experimental and working around its limitations, the stable release means you can now rely on it for production-track work — check that your CLI is on v0.128.0 or later.
  • macOS developers who share screen context mid-session. Appshots removes a specific friction point: the need to manually screenshot, crop, and paste visual context into an AI thread. If you debug by showing the AI what you're looking at, the double-Command shortcut changes that specific workflow.
  • Developers running Codex for overnight or unattended sessions. Post-lock Computer Use means a screen timeout no longer kills a long Computer Use run. Combined with Codex Mobile, you can start a task and monitor it remotely without keeping the machine active and unlocked.
  • Teams comparing OpenAI vs Cursor vs Claude Code for agentic coding. Goal Mode going stable is a feature parity signal. All three platforms now have production-grade long-horizon agentic modes. The comparison has tightened and the differentiation is increasingly in reliability, context handling, and integration depth rather than capability presence.

What To Do Now

  1. 1. Upgrade the CLI before relying on Goal Mode in production. The stable release requires CLI version 0.128.0 or later. Check your installed version with `codex --version` and update before assigning Goal Mode to production-track tasks.
  2. 2. Test Goal Mode stable on a non-critical task before going all-in. 'Stable' means it is no longer experimental — it does not mean it handles every edge case perfectly. Run a representative but low-stakes task through Goal Mode on your codebase before committing complex workflows to it.
  3. 3. Appshots is a quality-of-life improvement, not a workflow revolution. If you frequently share screen state with Codex, it saves meaningful friction. If you rarely do, it does not change much. Enable it and use it naturally — do not reorganize your development workflow around it.
  4. 4. Post-lock Computer Use still requires proper setup. Codex can run after your screen locks now, but it still needs a proper task definition and occasionally needs human intervention on ambiguous states. Check in on long tasks periodically even if you are not locked out.

More on this topic — Best AI Coding Tools

Independent Review

ChatGPT

Pricing, pros and cons, real-world verdict — no affiliate spin.

Read the ChatGPT review

The AI Hustle Playbook Newsletter

Get one practical AI playbook each week.

Tools, workflows, and side-income ideas — curated for people who want to build, not browse forever.

No spam. Unsubscribe anytime. We respect your privacy.