By asking visitors to correctly say which direction of say, four (up,
down, left or right) a part of a captcha appears to be "moving", it
filters out the bots while allowing humans with optical illusion
fooled eyeballs to pass the test. Given 16 of those captcha requests
in a row and given that there are 2^2 choices per captcha, there is a
2^16*2 - 1 possible ways to enter 16 captchas wrong, but only 1 way to
enter them right, leading to a filtration failure of around 1 in 4
billion.<p>Also, multiple captchas can be done in a row for each failure, but once say, 16 captchas have been solved in a row, access is granted for say, a month, until the captcha has to be done again.<p>It just feels like Hacker News is full of bots and I have a solution to this.
It seems like this would bring all of the problems of other CAPTCHAs, but with the additional problems of excluding the visually impaired as well as those who can't see optical illusions (which is as high as 20% of the population, depending on the specific illusion).<p>And would it really be effective against bots? It isn't obvious to me why it would be any more effective than other schemes.