OpenAI launches new MacOS app for agentic coding

OpenAI has made a significant leap in software development with the introduction of its new MacOS application for Codex, a tool designed to enhance coding efficiency through AI collaboration. This launch is a response to the rising trend of agentic software development, where AI agents autonomously tackle programming tasks, a shift exemplified by innovations like Claude Code and Cowork applications. The newly launched MacOS app integrates various agentic practices that have gained popularity over the past year. It allows multiple AI agents to operate simultaneously, utilizing their unique skills along with cutting-edge workflows. This release follows closely on the heels of GPT-5.2-Codex, OpenAI’s latest and most powerful coding model, which CEO Sam Altman claims is the best option for complex programming challenges. He emphasized, "If you really want to do sophisticated work on something complex, 5.2 is the strongest model by far." Despite Altman’s optimism, coding benchmarks present a more nuanced view. While GPT-5.2 currently leads on TerminalBench, which assesses command-line programming efficiency, its performance is closely matched by competitors like Gemini 3 and Claude Opus. Additionally, results from SWE-bench, a test for real-world bug fixing, show no definitive advantage for GPT-5.2 over its rivals. The Codex app comes packed with new features aimed at enhancing user experience and redefining efficiency. Users can automate tasks to run in the background according to a schedule, with outcomes queued for review upon their return. Furthermore, the app offers customizable agent personalities, ranging from pragmatic to empathetic, catering to individual working styles. Altman highlighted the transformative potential of AI in coding, stating, "You can go from a clean sheet of paper to building a sophisticated piece of software in just hours. As fast as I can type in new ideas, that is the limit of what can get built." This reflects OpenAI’s commitment to pushing the boundaries of what AI can achieve in the realm of software development.