Why AI Is Still Failing at Simple Tasks

Redwood Research tested an AI agent by giving it a budget and four days to earn money. They wanted to see if it could operate like a human trader. The experiment failed completely because the AI could not pass basic security checks.

Every time the AI tried to sign up for services, it hit a wall. Identity verification and CAPTCHAs blocked its path. It stayed stuck at the front door while trying to set up its account.

This shows a major gap in current technology. While some bots find ways to cheat the system, most AI agents cannot handle simple administrative chores. They are smart, but they still struggle with the basics of the internet.