The Ideal Workspace for an AI Agent Is a File Folder
Computers copied the metal filing cabinet 50 years ago, and AI learned to work by watching humans use the copy.
Lightspeed Venture Partners · From Prompt to Loop Engineering | LightworkThe ideal workspace for an AI agent, software that can do multi-step work on a computer without constant human help, turns out to be something familiar: files and folders.
That is because the computer folder is itself a copy. When designers were figuring out how to represent information on early machines, they looked to the physical office, with its metal file cabinets, paper folders, and stacked documents. It was intuitive to import that system to the screen, and office workers have now lived with that file and folder setup for more than 50 years.
Modern AI models were trained on material about humans interacting with computers through that file system. That material covers people creating folders, naming files, moving documents around, and using simple commands to open, edit, and organize them.
So an agent arrives already familiar with the setup: how folders are organized, what a file system looks like, and which utilities it can call on to manage them. Give it a workspace of files and folders, plus commands for working with them, and it does not have to learn a strange new world from scratch.
That view of files as the natural setting for an agent emerged especially during the past year, when agents began to take off. Startup founders are now trying to shape that workspace to fit how agents operate. Starting with folders helps because the behavior is already familiar to the model.
Questions people ask
Is an agent-oriented file system different from the folders on my laptop?
A regular folder on a laptop works, because the model is already familiar with how folders behave. An agent-oriented file system would be the same basic idea, a file workspace arranged with agents in mind.
Why can't an agent just use apps and websites like a person does?
Agents do use browsers and apps for some tasks. Files are useful alongside them because they offer one organized place where an agent can do its work and keep track of a job over time.
What does it mean to say the AI understands files and folders?
It means the model can navigate folders, recognize the structure, and use fitting commands without learning the system from scratch. That familiarity comes from training on material about people using file systems.