On April 16, 2026 — just hours after Anthropic released Claude Opus 4.7 — OpenAI announced a major overhaul of Codex, its AI coding app. The updated Codex is being marketed as 'Codex for (almost) everything,' with Mac-level computer-use permissions, a built-in browser, persistent memory that carries across sessions, automations that can wake up and continue work across days, and more than 90 new plugins for third-party tools. It is a direct push to turn a coding assistant into a general desktop agent.
The timing is not an accident. The major AI labs are increasingly releasing products on the same day to steal attention from one another, and this release cycle focused squarely on agentic coding — AI that does not just suggest lines of code but actually operates a computer, runs tests, browses documentation, and hands back finished work. OpenAI's pitch is that Codex can now sit in the same workflow as a human engineer for much longer stretches without supervision.
Giving an AI Mac-level computer use means letting it click, type, and open apps as if it were a person sitting at the keyboard. That unlocks enormous productivity, but it also raises real safety and security questions: what happens when an agent runs for days with access to files, browsers, and developer tools? OpenAI says persistent memory and automations are gated behind user controls, but the industry is still writing the rulebook for this kind of access.
For students, agentic coding is one of the most important trends to watch. The job of a software engineer is shifting from 'write this function' to 'supervise an agent that writes many functions and make sure the system works.' Learning to read AI-generated code critically, to design clear tasks, and to verify results will matter more than memorizing syntax. The careers that survive this shift will be the ones built on systems thinking, not typing speed.