Autonomous agent

An autonomous agent is an AI that pursues a goal across many steps on its own, deciding what to do, taking actions, observing results, and continuing until it's done or stuck, without a human approving each move. It sits at the far end of a spectrum from a simple assistant that waits for your next instruction.

For a founder this is the promise and the peril in one. A fully autonomous agent can run a whole workflow overnight, research, draft, file, follow up, which is enormous leverage when it works. But autonomy multiplies both the upside and the blast radius: a wrong turn compounds across steps before anyone catches it. That's why most useful deployments aren't fully autonomous; they're autonomous on the cheap, reversible parts and human-in-the-loop on the rest. Decide how much autonomy a task can safely take by asking one question: if it goes wrong unattended, can I undo it cheaply?