OpenAI Atlas browser: early tests show promising concept but clunky execution

OpenAI released Atlas, a browser integrated with ChatGPT and featuring an AI agent that can take actions on a user’s computer to complete tasks. Experts interviewed by Sky News described the agent as primitive and occasionally unreliable, noting it can struggle with complex workflows and may require user intervention. While the concept is hailed as exciting, with potential to disrupt traditional browsing and push the shift from attention economy to the answer economy, critics caution that reliability, latency, and task success still need improvement. Some researchers highlighted that the agent can take control of navigation and automate steps, but in practice it often stalls or performs suboptimally on more complex sites. Despite these early shortcomings, there is consensus that the idea could push Google to accelerate AI integration in Chrome or pursue newer prototypes. OpenAI acknowledges the early-experience nature of the feature and emphasizes ongoing improvements. The broader takeaway is that Atlas represents a significant, though imperfect, step toward more AI-assisted web experiences that could alter how people seek information online.