OpenAI Launches ChatGPT Work, Agent for Cloud-Based Knowledge Tasks

ChatGPT Work represents OpenAI's most accessible deployment of cloud agents to a mainstream audience. By embedding it directly into ChatGPT—which has crossed 1 billion weekly active users this month—OpenAI is positioning agents not as a specialized tool for developers or power users, but as the default interface for how a billion people will soon work. The architecture reflects deliberate trade-offs: rather than giving the agent unrestricted access to a single computer holding every file, conversation, and memory (as OpenClaw did), OpenAI has separated task-level freedom from cross-task continuity. Within a task, the agent can create directories, install dependencies, and manipulate files freely; across tasks, continuity flows through ChatGPT's existing product layer—Conversations, Library, Personal Context, Memory—which the company already operates at scale. This split serves as both a technical necessity (refactoring proven infrastructure would be costly) and a guardrail (unrestricted access poses user safety and data governance risks). Work's early steps toward proactivity—suggesting personalized tasks based on calendar and email context—hint at where OpenAI believes the product is headed, though today the user remains the triggering agent. The absence of a meta-layer coordinator that could manage work across multiple tasks and projects suggests this capability is still being designed.