Prove you are a robot: CAPTCHAs for agents

Article

• browser-use proposes reverse CAPTCHAs: puzzles easy for AI agents, hard for humans.
• Uses math/logic problems with obfuscated formatting to gate API access.
• Intended to distinguish legitimate agent users from human bots or scrapers.

Discussion

• Fundamental flaw noted: humans can use agents behind the scenes to solve any puzzle.
• Retr0id: humans have tool use, which breaks the scheme entirely.
• Commenter noted sha256 computation is a simpler, more robust agent-only test.
• One commenter’s agent solved a Japanese-language math puzzle correctly, which was the point.

Discuss on HN