Our AI Agent Could Use a Computer, But Nobody Cared

Half a year on an autonomous computer-use agent. A breakdown of the failure.

There were four of us: a backend dev, a frontend dev, me as the ML/AI engineer, and a mobile dev. I owned the ML part. Half a year in (way too long), I realized it was going nowhere and stopped putting in time and effort. A month later everyone else stopped too.

The idea

A year ago, when Manus had just appeared and a big corporation hadn't bought it yet, we got excited about where AI could go next. A person works at a computer with a set of tools: eyes and hands. So to make an agent as capable as a human, you give it the same set of tools.

The idea was simple: give the agent the ability to look at the screen, click, scroll, and type, and we'd finally get what everyone had wanted...

Copyright of this story solely belongs to hackernoon.com. To see the full text click HERE