OpenAI Codex release notes: app 26.602 becomes a rollout check

When someone searches for Codex release notes, the real question is rarely “what did OpenAI ship?”. It is usually closer to: “can we update Codex without losing control of files, logs, plugins or rollback?”. That is why this June 5 post has been refreshed as a rollout check, not as another news recap.

Codex app 26.602 from June 4 was a smaller app update: activity insights and share cards in the Profile section, better Computer Use startup, clearer appshot error reporting and several UI fixes. Read together with later CLI notes, especially 0.142.5 where full Responses WebSocket payloads stopped being written to trace logs, the pattern is clear: release notes become operations work once the AI agent touches real repositories, tools and data.

Source: OpenAI Codex changelog, 2026-06-04 and Codex CLI 0.142.5 release.

OpenAI Codex release notes: what app 26.602 changed

The Codex app is OpenAI’s desktop environment for coding agents in local projects, threads and review flows. A coding agent can read code, propose changes and sometimes use tools under human control. App 26.602 did not change the base model, but it made usage easier to see and review.

The important parts:

• Activity insights and share cards: users can review usage highlights in the Profile section and save a profile card. Sharing is listed for consumer ChatGPT plans.
• Computer Use and appshots: OpenAI improved Computer Use startup readiness and appshot error reporting. Computer Use means Codex can work in desktop apps; appshots capture context from app windows.
• Review and browser surfaces: OpenAI mentions fixes for fullscreen browser composer controls, hex color swatches, terminal scrollbar alignment, animated diff stat alignment and onboarding with more role choices.

Source: OpenAI Codex changelog, 2026-06-04, “Codex app updates 26.602”. Appshots are also described in the 2026-05-21 changelog.

The answer the searcher needs: should we update?

Yes, but treat Codex as a tool that can change the workflow, not as a normal chat app. If one person uses Codex in a test repository, it may be enough to update, open the Profile section and check that the app behaves as expected. If Codex is used for customer code, internal tools or remote sessions, run a short check first.

Use this order:

• Version: which app or CLI version are we actually running?
• Doctor/status: does codex doctor, the app state or the vendor status page show anything that affects the work?
• Login and permissions: do login, SSO, plugins and MCP servers still work after the update?
• Sandbox and files: does the agent only write where it should, and is the diff visible before merge or publication?
• Logs: do trace/debug logs store sensitive prompts, repository data or WebSocket payloads?
• Rollback: does the team know which version or routine to return to if the workflow becomes unstable?

This check matters even more after Codex 0.142.5. That release note was narrow, but relevant: a trace statement was still logging full request text and was not covered by existing request-log filtering. The fix did not change request behavior or the app-server API, but it shows how easily logging becomes a data risk.

Source: GitHub PR #30771 and OpenAI Codex 0.142.5 release.

Turn usage insights into a control habit, not a trophy

Activity insights only help if someone turns them into a decision. A simple pattern is to have a human open the Profile section, read the usage highlights and write one small operating rule for the week.

Example prompt for internal Codex work:

Here are our Codex usage highlights from the Profile section: [short summary]. Compare them with our three most important repository or automation goals this week. Suggest two changes to how we use Codex, one thing to measure next week and where human review should be required. Do not change files.

Good output should not just say “use Codex more”. It should identify which tasks fit, which ones need an approval gate and where the team should write a clearer rule in AGENTS.md or in plugin/MCP configuration.

Sibling links: read the right Codex release note first

If you arrived from a general Codex query, start with the version or feature that matches your situation:

• Codex 0.135.0: diagnostics, profiles and doctor fits when you need local environment and support data checks.
• Codex Sites preview fits when the question is hosted web apps, secrets and publication gates.
• Codex 0.142.5 trace logs fits when logs, traces and sensitive prompts are the risk.
• Codex 0.143.0-alpha.36 fits when installation, plugin versions and alpha tracks need review.

That is why Hammer keeps Codex posts in a release-note series. The series is not vendor watching for its own sake. It helps a reader find the right version decision faster.

The Hammer angle: Tool Forge starts with an update routine

For Hammer customers, a Codex update is rarely “just a new version”. It is a question of who may let the agent write files, which tools it may use, what gets logged and how quickly the team can roll back.

A small Tool Forge routine can be enough:

• owner for the Codex installation
• approved repositories and test repositories
• plugin/MCP list with purpose and owner
• log level and where traces are stored
• one test case for each important workflow
• one rollback row with version, command and responsible person

Start there. Once that routine works, Codex can take on more responsibility without every release note becoming a guess.