Codex on mobile is not interesting because I suddenly want to write a tricky refactor from a phone keyboard. I do not. The interesting bit is supervision. Coding agents are becoming background work, and mobile changes where that work can be checked, steered, or stopped.
OpenAI’s Codex for almost everything announcement puts Codex across more of the product surface, including mobile. That matters because the agent workflow is not the same as autocomplete. You give it a task, it works for a while, then you review the result and decide what happens next.
That review loop is much more mobile-friendly than the actual coding loop. Reading a plan, checking a diff summary, nudging a task, cancelling a bad direction, or asking for tests is plausible from a phone. Editing a complex TypeScript file on a train sounds like a punishment. The value is being able to keep long-running work moving without being chained to the desk.
The tradeoff is that mobile makes discipline more important. If agents are easier to kick off casually, teams need better scopes, repository instructions, tests, and review habits. Codex on mobile is useful, but only if it becomes a way to supervise serious work, not a way to create more half-reviewed code.