Why Dwarkesh is Wrong about Computer Use + How OpenAI shipped its Jev competitor in 1 Week

OpenAI’s latest developments in computer use agents and API infrastructure mark a shift toward highly efficient, autonomous software interaction. By providing agents with dedicated Linux virtual environments, these tools now execute complex, multi-step tasks—such as meal planning or YouTube management—faster than human operators. The newly introduced Decisions API complements this by enabling rapid, parallel inference for classification, distinct from the reasoning-heavy models required for long-horizon agentic tasks. Infrastructure enhancements, including async tool calling, WebSocket-based bi-directional communication, and extended caching guarantees, significantly lower latency and operational costs for developers. These advancements, combined with multimodal data representation like AppShots, allow agents to effectively navigate and manipulate software interfaces, moving the industry closer to a paradigm where agents reliably handle intricate, real-time digital workflows with superhuman speed and precision.