ChatGPT Just Became a Work Agent

The AI industry is undergoing a significant shift as labs prioritize cost-efficient, agentic infrastructure over raw model performance. OpenAI’s release of the GPT 5.6 model family and the "ChatGPT Work" interface exemplifies this transition, moving beyond simple chat to autonomous, multi-step knowledge work. Simultaneously, Meta’s surprise launch of MuseSpark 1.1 challenges the status quo by delivering frontier-level capabilities at a fraction of the cost, forcing competitors to re-evaluate their pricing and efficiency strategies. As enterprises integrate these tools into core workflows, the focus has moved from static benchmarks to "harnesses"—the systems that allow models to interact with tools, manage context, and execute complex tasks. This evolution marks a pivotal moment where AI utility is measured by real-world operational impact rather than traditional model-only metrics.