May 25, 2020 · 1h 4m · how-i-built-this
reCAPTCHA and Duolingo: Luis von Ahn
0:00 / 0:00
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of How I Built This, entrepreneur Luis von Ahn details his journey from pioneering crowdsourcing technologies like CAPTCHA and reCAPTCHA to founding Duolingo and scaling it into a global, gamified education platform.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Guy holds 34.2% of the talking time here. How this is scored →
Guy as informed peer 1.9 Guest teaching 2.5 Guest disagreement 0.3 Guy pushing back 0.7
speaking balance: gold is Guy, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 55:02 Rejecting instant fluency marketing myths
Luis firmly dismisses marketing claims that suggest people can master languages in days, directly pushing back against unrealistic expectations for educational software.
Hardest push from Guy ▶ 48:24 Raz presses on massive software capital expenditureRaz halts Luis's narrative to challenge why an app consisting of code requires tens of millions of dollars without physical manufacturing costs.
Biggest teaching moment ▶ 27:09 Explaining dual-word verification mechanicsLuis educates Raz on how reCAPTCHA solves the unknown data problem by pairing a known control word with an unverified scanned word to guarantee digitization accuracy.
Guy holds their own ▶ 50:51 Raz dissects ideological self-entrapment in startupsRaz synthesizes startup strategic pitfalls, articulating how early rigid dogmas against advertising box companies in until practical revenue models are embraced.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Guy as informed peer | Guest teaching | Guest disagreement | Guy pushing back | Why |
|---|---|---|---|---|---|---|
| Childhood in Guatemala: Candy Machines, Math, and Civil War | 1 | 1 | 0 | 0 | Guy Raz establishes rapport by asking about Luis's childhood in Guatemala, candy factory machines, and the dangers of the civil war. The interaction is warm, biographical, and conversational without friction. | |
| Transition from Pure Mathematics to Computer Science | 2 | 3 | 1 | 1 | Luis explains transitioning from pure math to computer science at CMU and how attending a Yahoo lecture inspired the original CAPTCHA to fight spam bots. Raz playfully jokes about being a human who fails CAPTCHAs. | |
| Naming CAPTCHA and Worldwide Web Adoption | 2 | 4 | 1 | 2 | Raz expresses astonishment that Luis gave the CAPTCHA system to Yahoo for free rather than licensing it. Luis explains his academic mindset at age 21 and how open release spurred rapid web adoption. | |
| Crowdsourcing Research and Google's Acquisition of the ESP Game | 1 | 3 | 0 | 0 | Luis details the ESP Game, explaining how crowdsourcing human perception through online games allowed image labeling before modern AI, culminating in an acquisition by Google. | |
| MacArthur Fellowship, Teaching, and Turning Down Bill Gates | 2 | 1 | 0 | 0 | Raz inquires about the MacArthur Genius Grant and the anecdote of turning down Bill Gates's personal job offer. Luis modestly recounts choosing academia over Microsoft. | |
| Conceiving reCAPTCHA to Digitize the World's Books | 2 | 4 | 0 | 0 | Luis outlines his realization that 500,000 hours of human effort were wasted daily on CAPTCHAs, prompting him to invent reCAPTCHA to decipher words scanned books' OCR engines could not parse. | |
| Scaling reCAPTCHA and The New York Times Archive Contract | 2 | 4 | 0 | 1 | Raz asks how verification works without known ground truth, and Luis explains the two-word verification model and the lucrative digitization contract with The New York Times archive. | |
| Selling reCAPTCHA to Google and Meeting Severin Hacker | 1 | 1 | 0 | 0 | Luis recounts selling reCAPTCHA to Google, leaving early to maintain academic freedom, and meeting his future Duolingo co-founder Severin Hacker. | |
| Mid-Interview Reflection and Commercial Break | 2 | 3 | 0 | 1 | Following the mid-roll, Luis explains the initial vision for Duolingo: offering free education by crowdsourcing translation of news articles for media outlets like CNN. | |
| Venture Funding and Flaws in the Translation Model | 2 | 3 | 0 | 1 | Luis explains raising Series A venture capital with Union Square Ventures and Ashton Kutcher, but quickly discovering that commercial translation was an unsustainable race to the bottom. | |
| Pivoting to Mobile, Gamification, and Explosive User Growth | 2 | 2 | 0 | 0 | Luis describes how a summer intern and engineer built the standalone mobile app, shifting the product from translation exercises to gamified mobile micro-learning. | |
| Monetization Crossroads: Google Capital G and Subscription Model | 3 | 3 | 1 | 2 | Raz presses on why software development is so capital intensive and discusses the tension between ideological anti-monetization and venture expectations from Google Capital G, leading to non-intrusive ads and subscriptions. | |
| Pedagogical Strategy, AI Notifications, and Overtaking Rosetta Stone | 3 | 4 | 1 | 2 | Raz raises common academic criticisms regarding whether app-based gamification can truly teach language fluency. Luis clarifies the realistic goal of B2 proficiency, AI-driven notification nudges, and outcompeting Rosetta Stone. | |
| Pandemic Acceleration, Financial Growth, and IPO Strategy | 2 | 2 | 1 | 1 | Raz asks about pandemic impact and exit strategy; Luis corrects the premise that ad declines hurt them, revealing subscription surge drove daily revenues to $600k and positioned them for an IPO. | |
| Luck, Pittsburgh's Tech Advantage, and Episode Conclusion | 1 | 0 | 0 | 0 | Raz asks his signature question regarding luck versus skill, and Luis attributes success to both mobile timing and relentless obsession. Raz closes with an anecdote about Pittsburgh tech recruiting. |
Statements from this episode (27)
Von Ahn: CEO role is 100% dealing with people and problems
“And it is interesting because my job these days is 100% just dealing with other people and people's problems.”
Von Ahn: CAPTCHA was developed independently at CMU without notifying Yahoo
“Yeah, I mean, with my PhD advisor, it was, it became a research subject. And yeah, we did not tell Yahoo.”
Von Ahn: Yahoo deployed CAPTCHA live within a week of receiving code
“What was amazing is within about a week of when we sent them this, it was already functional. You know, live. It was live.”
Von Ahn: No money was exchanged when giving CAPTCHA to Yahoo
“No, no, no, no, no money was, no money exchanged hands.”
Von Ahn: Bill Gates spent nearly an hour recruiting him to Microsoft
“I had been an intern, I had been, like a year and a half before that, I had been an intern at Microsoft. And so I had good ties with Microsoft, and they really wanted to hire me for Microsoft Research. And so that is, you know, Bill Gates spent, I don't know, …”
Von Ahn: Early CAPTCHAs wasted 500,000 hours of human time daily
“I did a little back of the envelope calculation, realized about two hundred million Times a day somebody types a captcha. And that's when you get two hundred million times a day times 10 seconds. So humanity as a whole is wasting about 500,000 hours a day typi…”
Von Ahn: Early OCR failed on 30% of scanned book words
“Computers cannot, or at the time, could not recognize many of the words. About 30% of the words computers could not recognize.”
Von Ahn: Online Booty Call was reCAPTCHA's first major client
“The first largest, it wasn't even large, the largest website for a while was That was using us was called Online Booty Call.”
Von Ahn: reCAPTCHA charged NYT $42,000 per archive year
“The number we came up with was for every year of content, we will charge you 42,000 dollars.”
Von Ahn: CMU took standard 5% equity in research spinouts
“They had a standard deal that says, okay, if you're going to start a company based on, on something that you came up in your research, we'll take five percent of this company.”
Von Ahn: reCAPTCHA had no outside investors when Google acquired it
“Especially since we didn't have investors or anything. It's all the money went to us.”
Von Ahn forfeited part of Google payout by leaving before three years
“Leaving cost me quite a bit of money, because some of the reCAPTCHA payment required that I stay there for three years. Yeah, and I did not stay for three years.”
Luis von Ahn: Education drives socioeconomic inequality instead of equalizing society
“And a lot of people talk about education as something that brings Equality to different social classes, but I always saw it as the opposite, something that brings inequality, because what happened, at least, you know, in my case, those who have money can buy t…”
Von Ahn: Knowing English doubles income potential in most countries
“And it turns out that in most countries in the world, if you know English, you can double your income potential.”
Luis von Ahn: Duolingo raised $3.3M Series A from USV, Ferriss, Kutcher
“It was 3.3 million, which at the time, this was a normal Series A. So it's three million dollars from Union Square Ventures and 300,000 from a combination of Tim Ferriss, and Ashton Kutcher.”
Luis von Ahn: Duolingo landed translation contracts with CNN and BuzzFeed
“And we landed a contract with CNN and a contract with BuzzFeed that we were going to Translate all their news from English to Spanish.”
Von Ahn: Staying motivated is the hardest part of self-learning
“The hardest thing about learning, learning something by yourself is staying motivated.”
Von Ahn: Duolingo canceled revenue contracts to rely solely on VC
“So I made the difficult decision to cancel those contracts and forget about trying to make money and just raise more venture capital.”
Von Ahn: Adding one ad jumped Duolingo's revenue run rate to $10M
“I mean, as soon as we put that ad, our run rate went from zero to like ten million bucks a year.”
Von Ahn: Duolingo makes significantly more money from subscriptions than ads
“It turned out, and this is just, we really just didn't realize this subscriptions made us a ton more money than the ads.”
Von Ahn: Duolingo has 40M MAUs with a 3% paid conversion rate
“Yeah, I mean, active users right now, we have about Forty million monthly active users, and about three percent of them are paying subscribers.”
Von Ahn: Duolingo alone cannot make learners fluent like a poet laureate
“If you want to use Duolingo and start from zero and get to the point where you're the poet laureate of that country, that is unlikely to happen with Duolingo by itself. That's just not going to happen.”
Von Ahn: Duolingo has 50x Rosetta Stone's users and 5x its revenue
“By now, I would say we probably have 50 times more users than they do. Our revenue is probably five X their revenue.”
Von Ahn: Education companies become marketing-driven because product improvement is hard
“A lot of companies in education eventually turn into marketing companies. For the following reason. You realize, man, making improvements in how well you teach is hard. And it really is very hard. It's, it takes years. So for every dollar you have, you quickly…”