CAPTCHAs are a terrible experience for users and a nightmare for accessibility, and they’re not likely to get better any time soon.
This Verge piece by Josh Dzieza highlights how the underlying technology—machine learning—is in an arms race with itself.
The more humans click through CAPTCHAs the more training the algorithms involved receive. That makes the machines better at solving the very same CAPTCHAs they present to users, and hands bad actors the perfect tool for overcoming CAPTCHAs. The process eats its own tail.