The one idea underneath all three
Human computation: take a task computers can't do, split it into ten-second pieces, and get millions of people to do those pieces as a side effect of something they already wanted to do.
CAPTCHA (2000, built at Carnegie Mellon with Manuel Blum) verified ~200 million humans per day. Von Ahn did the math: at 10 seconds each, humanity was burning ~500,000 hours daily typing squiggles. That waste bothered him.
So reCAPTCHA (2007) made the squiggle a scanned word that OCR software had failed to read. You typed two words: one known (the test), one unknown (the work). Agreement across users transcribed the unknown word. Google bought it in 2009 and used it to digitize the New York Times archive and millions of Google Books pages — roughly 100 million words deciphered per day at its peak.
Duolingo (2011) applied the same trick to motivation instead of OCR: gamified streaks, five-minute lessons, and A/B-tested notifications made learning free at global scale — over 500 million registered learners, with courses built for dozens of languages. Von Ahn, a MacArthur "genius grant" winner at 27, now runs it from Pittsburgh while funding education projects in Guatemala.
What to steal from this playbook
1) Find wasted human effort and attach it to a real problem. 2) Make the byproduct the product. 3) Price at zero for the people who need it most and monetize the ones who can pay — Duolingo's subscription funds the free tier. It's a repeatable pattern, not a lucky break.