We were extremely surprised to find that in 2026, most services are easily botable and can be extorted for free credits, referral chaining, and other malicious behavior.

We realized that bot detectors don't work today because agents now use computer use and don't succumb to tracking network IDs, fingerprinting, and other indicators of being a bot. Instead the only way to detect them now is not through CAPTCHAs, but through their general behaviors. Thats why we built Flaid.

Flaid tracks users' keyboard typing, scrolling, and clicks and matches it against our dataset of over 1,000 runs. It has 99% accuracy for catching malicious bots on sites and a near 0% false positive rate.

Our aim with this project was to explore how we can make fraud unprofitable while, at the same time, not punishing real users with a bad UX experience in the agentic era.

Built With

  • grokbot
Share this project:

Updates

Submission history