Pig.dev, a Y Combinator startup, initially aimed to revolutionise Windows desktop control using AI agents. The company sought to develop AI capable of automating complex workflows by interacting with the Windows graphical user interface. Users could connect their Windows machines and develop automation agents via a chat interface or SDK. Pig offered tools for AI to use, such as clicking, typing, and taking screenshots to control Windows apps.
However, Pig.dev faced challenges including maintaining accuracy over extended periods and the high computational costs associated with large language models. They discovered that customers wanted fully managed solutions rather than developer tools. In response, Pig.dev pivoted to Muscle Mem, a caching system designed for AI agents. This shift reflects a move towards addressing core AI agent efficiency issues rather than direct Windows automation.
Microsoft is also pursuing Windows automation, integrating computer use technology into Copilot Studio. Piglet, a computer-use driver maintained by Pig, exposes a high-level API for desktop automation tasks. It offers tools for getting screenshots, reading and writing the element tree, keyboard and mouse control, reading and writing files and running commands.
Related Articles
AI Browsers Emerge as Agents
Read more about AI Browsers Emerge as Agents →ChatGPT to Automate Tasks
Read more about ChatGPT to Automate Tasks →Bedrock Robotics launches with funding
Read more about Bedrock Robotics launches with funding →Thinking Machines Lab Secures Funding
Read more about Thinking Machines Lab Secures Funding →