Some public Internet services have a policy to the effect: "This service is designed for human beings, not for robots." For example, GeoCities and Tripod doesn't want scripts signing up for accounts and uploading warez to them; it just creates more work for GeoCities' copyright enforcement staff. Likewise, Freenet and AltaVista want to keep out spammers who would otherwise insert advertisements and overload the database.

There are two solutions I can see: mass moderation (a la Slashdot, Kuro5hin, and Everything), and "Think Cash." To generate Think Cash, a server generates a test that satisfies these properties:

  • It should be easy for Server to generate the questions (which will necessarily involve a one-way transformation) and check for correct answers.
  • It should be easy for User to generate correct answers given only the questions and feed the answers through Client back to Server.
  • It should be hard for Client to generate a correct answer to the current question given only the previous questions and their correct responses.
In other words, a function that's one-way for computers and two-way for humans. A correct answer would give User a token redeemable for (e.g.) five URL submissions or one free web site. (Source: )

The problem is that the current think cash systems implemented on the Web are based on an OCR problem; they require the user to be fully sighted and using a GIF-capable web browser. This leaves the blind and otherwise visually impaired users out in the cold, along with users of character-cell browsers such as lynx, links, and w3m. (To see what AltaVista's think cash form looks like in Lynx, visit