Duolingo: From Bot-Fighting Test to EdTech Icon

Luis von Ahn: From CAPTCHA Creator to EdTech Innovator

Luis von Ahn, a dedicated entrepreneur focused on expanding access to free education, has likely caused frustration for many internet users. His creations have, in all probability, been a source of annoyance on numerous occasions.

Prior to establishing the popular language-learning application Duolingo, he developed the technology behind CAPTCHA. This system, designed to differentiate between human users and automated bots, appears during online registration and login processes for widely used services like email.

The Connection Between Security and Education

Although seemingly disparate, the principles employed in creating scalable security tests for widespread consumer use ultimately formed the foundation for Duolingo’s success. The core DNA of this highly successful edtech company was born from this earlier work.

The entrepreneur, himself an immigrant, discovered that leveraging crowdsourcing, focusing on language learning, and maintaining adaptability – even in the face of criticism – could fundamentally reshape an entire industry.

Duolingo's Impact

Duolingo has become one of the most downloaded education apps globally, boasting over 500 million downloads and a substantial user base of 40 million active users.

This demonstrates the power of applying innovative security concepts to the field of education, creating a platform that is both accessible and effective.

Addressing Market Challenges with Innovative Solutions

Luis von Ahn’s early life in Guatemala City exposed him to the disparities in educational opportunities. His mother made significant financial sacrifices, reportedly exceeding $1 million, to provide him with a quality education. This experience instilled in him a desire to make education more accessible to others.

While pursuing his Ph.D. in computer science at Carnegie Mellon University, von Ahn attended a presentation by Yahoo’s chief scientist outlining the company’s most pressing issues. A key concern was the proliferation of bots creating fake email accounts for spam distribution.

Driven by a spirit of innovation, von Ahn and his team, under the guidance of Manuel Blum, developed a system to differentiate between automated bots and human users. This system, known as a CAPTCHA, presented distorted words that proved easily decipherable by humans but resistant to computer vision technology at the time.

The CAPTCHA test quickly gained widespread adoption through Yahoo, becoming a ubiquitous security measure online. At its height, users collectively spent an estimated 500,000 hours daily completing around 200 million CAPTCHAs, representing recognition by roughly 10% of the global population.

Despite its success, von Ahn recognized an inefficiency in the process. He observed that the time spent solving CAPTCHAs – time during which human brains performed tasks computers couldn’t – was being utilized for a relatively unproductive purpose. He questioned whether this collective effort could be redirected towards a more beneficial goal.

In 2005, this led to the launch of reCAPTCHA, an evolution of the original CAPTCHA. This new iteration maintained the security function but incorporated a secondary purpose: digitizing books for the Internet Archive by presenting scanned book pages as prompts.

Von Ahn understood the potential value of this innovation. In 2009, he sold reCAPTCHA to Google, following a previous licensing agreement with the company for another of his research projects focused on image labeling.

The acquisition provided both financial reward and increased industry recognition for von Ahn, shortly after completing his doctoral studies. However, he chose to remain in Pittsburgh and accept a position as a computer science professor at Carnegie Mellon University.

This role in academia aligned with his long-held ambition to improve access to education. Unbeknownst to him, his previous work was laying the groundwork for his next venture. Incubation, setbacks, and investment would all contribute to a new project, building upon the idea that language could serve as a bridge rather than a barrier.

The Birth of Duolingo

The success of both CAPTCHA and reCAPTCHA highlighted the power of leveraging human cognitive abilities.
Carnegie Mellon, crowdsourced translation, and Google all played crucial roles in the development of his next venture.
This ultimately led to the creation of Duolingo, a language learning platform valued at $2.4 billion, designed to make language acquisition enjoyable.

Von Ahn’s journey demonstrates how a solution to a technical problem – combating bots – could evolve into a transformative force in the field of education.

The Genesis of Duolingo

The year 2011 witnessed a surge in edtech startups, including companies like Coursera and Codecademy, which have since grown into multibillion-dollar enterprises. The increasing adoption of iPads and tablets within educational settings empowered founders who envisioned a future where education was primarily delivered online. A palpable sense of excitement surrounded virtual instruction, perceived as a promising area for investment.

Meanwhile, Luis von Ahn began to feel constrained by the traditional computer science classroom, finding it less stimulating than the potential of computers to broaden access to education. Consequently, he sought a motivated, first-generation immigrant student to collaborate with, mirroring the faith Blum had once shown in him.

Von Ahn connected with Severin Hacker, a Ph.D. student from Switzerland, and together they started developing the concept for a web-based learning platform.

In contrast to von Ahn’s vibrant and outgoing personality, Hacker is a reserved and intensely focused individual. However, he possesses a dry wit. He’s known to quietly declare, “This meeting has peaked,” before excusing himself when a discussion loses its focus. His self-introductions are equally unique, involving a playful pantomime of his name – pretending to sever an arm, followed by rapid typing.

Upon forming their partnership, the pair needed to determine a subject matter. Hacker leaned towards coding, while von Ahn considered mathematics or programming. Ultimately, they chose language learning, a field outside their primary areas of expertise.

Von Ahn reflected on the advantages he experienced growing up in Guatemala, recognizing the significant impact his English proficiency had on his opportunities. He stated, “I enjoy math, but mastering math alone doesn’t directly translate to income. You study math to then pursue physics and become an engineer, whereas English skills demonstrably enhance earning potential in many nations.”

This realization wasn’t novel. A substantial market already existed for language-learning resources, valued at $82.6 billion at the time, encompassing all languages. Furthermore, Rosetta Stone dominated the software landscape, offering products priced at $600 to $700 per CD, without a corresponding app.

Hacker and von Ahn identified a clear opportunity: Duolingo could attract users by providing a free and effective language-learning application.

The challenge before these two technologists, lacking formal training in linguistics, was significant. Fortunately, von Ahn had a proven track record of leveraging collective intelligence to overcome obstacles.

Crowdsourced Translation: The Duolingo Origin

At TechCrunch Disrupt in 2012, Luis von Ahn unveiled a preview of a novel product before an audience exceeding one thousand people: “Duolingo: Learn a language through website translation.” The company was officially releasing its initial offering, having already cultivated an exclusive beta group of 125,000 individuals.

The product’s introductory trailer highlighted a fundamental human need: “Throughout history, information has been exchanged between people using a shared language; however, effective communication requires either knowing or learning the language of others.” This challenge also existed online, where content appeared in numerous languages, each requiring reader proficiency.

Drawing parallels to his success with reCAPTCHA, von Ahn identified a potential two-sided market facilitated by technology. This market connected individuals seeking language acquisition with entities – websites, news organizations, and marketing companies – needing cost-effective, rapid content translation.

Initially, Duolingo functioned as follows: A native Spanish speaker aiming to learn English would receive English sentences from websites for translation into Spanish. These tasks were adjusted to the user’s skill level, preventing discouragement for beginners and boredom for advanced learners. Assistance was available; users could access suggested translations for unfamiliar words.

Following the translation exercise, Duolingo provided practice with newly encountered vocabulary through examples and flashcards. This system offered an elegant solution, addressing two distinct needs with a single platform.

With one million active users, Duolingo estimated the complete translation of English Wikipedia into Spanish could be achieved in just 80 hours. The company’s revenue model involved securing paying clients who would purchase these user-generated translations, while maintaining free access for learners.

“The service will be entirely free, with no advertisements, hidden charges, or subscription fees,” the trailer emphasized.

Von Ahn’s established reputation from reCAPTCHA proved advantageous for Duolingo from the outset. This recognition quickly positioned Duolingo as a media favorite, enabled von Ahn to present at international education conferences, and bolstered the company’s credibility, attracting initial investment offers.

Monolingo VCs: A Focus on Location Over Education

Prior to its demonstration and subsequent popularity, Duolingo initially sought $3.5 million in Series A funding in 2011. Even with a founder, Luis von Ahn, possessing a successful prior exit to Google, securing investment proved challenging due to VCs’ skepticism towards educational ventures and companies not situated in Silicon Valley.

Dedicated edtech venture capital firms like Reach Capital, GSV, and Owl Ventures were still in their formative stages. Mainstream Silicon Valley largely disregarded the potential of the education technology sector at that time.

Bing Gordon, previously a partner at Kleiner Perkins (known then as Kleiner Perkins Caufield & Byers or KPCB), demonstrated a keen interest in the gamification of learning materials and the modernization of traditional lectures as early as 2009.

He recounted that his investment thesis regarding education was largely met with dismissal. However, Kleiner Perkins later led Duolingo’s $20 million Series C round and invested in other edtech companies, including the unicorn Newsela and the publicly traded Coursera.

Duolingo benefited significantly from having von Ahn as its founder. While Silicon Valley wasn't prepared to invest in a language-learning company based in Pittsburgh, led by founders without direct language-learning experience, it was willing to back a seasoned founder with a history of successful company sales.

“We received offers from numerous investors,” von Ahn stated. “Later, I learned that their reasoning was, ‘We didn’t understand your vision, but you’ve sold two companies to Google, so we’ll take the risk.’” Rumors of a potential acquisition by Google Translate circulated even before Duolingo reached its first anniversary.

Von Ahn ultimately accepted a term sheet from Brad Burnham, who was spearheading the investment for Union Square Ventures. Burnham offered two key advantages beyond the capital itself. Firstly, USV was among the few firms that didn’t mandate a relocation to Silicon Valley following the investment.

Currently, the majority of Duolingo’s 384 employees remain based in Pittsburgh, with additional offices in New York City, Seattle, and China. Secondly, Burnham was a rare investor who recognized the value of education as a viable investment strategy.

He articulated in 2009 that the primary goal should be to “discover methods to leverage technology in order to lower the expenses and broaden the reach of education.” Burnham’s emphasis on accessibility resonated strongly with the core mission of von Ahn and Hacker.

With USV’s backing – the same firm that initially invested in Twitter, Etsy, and Zynga – Duolingo gained significant prestige within Silicon Valley. Ashton Kutcher and author Tim Ferriss also participated in the funding round, and within a year, Duolingo secured five times the capital in its Series B.

Venture capital provided Duolingo with a financial buffer – and ultimately a crucial lifeline – enabling the company to prioritize product development over immediate monetization for several years.

The Startup Transformed by its Interns

Duolingo’s application has received numerous accolades and enjoys greater recognition than any of the company’s other products, including its proficiency assessments. However, the success of its iOS app was largely accidental, driven by the initiative of two interns at the time.

In 2011, websites were still the primary platform for tech startups, with mobile applications only beginning to achieve the widespread popularity they have today. Luis von Ahn, believing users would be unlikely to engage in typing translations or sentences on mobile devices, prioritized development for desktop platforms. Shortly after its public release, Duolingo’s website attracted over 300,000 users, with an average daily engagement time of 30 minutes.

While other consumer startups followed a similar strategy, some explored the possibility of creating companion apps. These apps were often intended simply to demonstrate a startup’s mobile-friendliness, offering only a limited subset of the functionality available on the desktop site. Despite Duolingo’s focus on translation and the growing success of its website, the founders were keen to avoid falling behind emerging trends.

They envisioned a flashcard application as a suitable companion app and assigned the task of developing a prototype to two newly hired interns.

Tyler Murphy and David Klionsky, the assigned interns, had previously never owned a smartphone. They purchased an iPhone while traveling to a Duolingo website launch event in New York City to facilitate their research and development efforts.

“We examined numerous applications that are no longer available,” Murphy stated. Path was noted for its design quality, Evernote was gaining traction, but Instagram served as a “North Star” for optimal mobile user experiences.

Murphy aimed for Duolingo to deliver a similarly exceptional experience, providing a fully functional application free of charge, rather than a simple flashcard program. He also sought to emulate the sleek aesthetic of a particular iconic app, advocating against the creation of language-specific companion applications.

“At the time, this didn’t appear to be a particularly groundbreaking concept,” von Ahn commented. “However, in hindsight, I believe it was one of the most beneficial decisions we ever made.”

By concentrating on a single, unified experience for all users, Duolingo was compelled to carefully consider its core value proposition. “This necessitated a clear identification of Duolingo’s fundamental experience – the skill tree that users unlock to progress, tapping to initiate lessons, and receiving celebratory feedback upon completion,” Murphy explained. “It highlighted that as our primary strength.”

He proceeded to eliminate extraneous features, shaping Duolingo into the platform utilized by millions of users today.

Within one week of the iPhone app’s launch, Duolingo ascended to the position of the most downloaded application in the education category. A year later, in 2013, it was honored with Apple’s iPhone App of the Year award. The following year, it received TechCrunch’s Crunchie award for education startup of the year.

The recognition from Apple’s App of the Year award was the moment von Ahn and Hacker realized Duolingo’s potential exceeded their initial expectations. “My previous ventures were more along the lines of ‘This is interesting, I’ll pursue it,’” von Ahn said. “But Duolingo? I am prepared to dedicate years to this. It’s edtech, it’s a gradual process [and] it wasn’t a company I was inclined to sell.”

Re-evaluating the Monetization Strategy

Despite initial successes, von Ahn soon recognized a potential issue with Duolingo’s early business model. He anticipated that users would be reluctant to engage in extensive translation or sentence composition tasks directly within the mobile application. Consequently, the website translation monetization approach was discontinued for the app.

However, a significant challenge arose as approximately 80% of Duolingo’s user base was accessing the platform via mobile devices. With the most promising aspect of the product being detached from revenue generation, Duolingo faced a critical dilemma.

Von Ahn stated, “Our primary revenue stream relied on the website, yet the app represented the future of the platform.” This presented a considerable obstacle. Previous translation-based strategies, which had proven effective for von Ahn, began to appear less viable. Securing contracts with companies like Buzzfeed and CNN proved difficult, and emerging translation services consistently offered lower pricing.

Simultaneously, Duolingo encountered intense competition from numerous language-learning platforms, including Livemocha, PlaySay, Voxy, italki, MindSnacks, and Busuu. This competitive landscape necessitated a strategy to maintain free access while differentiating itself sufficiently to retain its user base.

Karin Tsai, an early engineering employee at Duolingo and currently a director of engineering, participated in the meeting where the decision to abandon translation services as a monetization method was made.

“A conflict existed between utilizing learner translation for revenue and prioritizing effective language instruction,” she explained. “Our core objective was to excel at the latter, and there was concern that prioritizing the former – relying on translation for business sustainability – would compromise our commitment to optimal learning and teaching practices.” Furthermore, she noted that Duolingo was “far from achieving profitability” with translation as its primary income source.

Von Ahn doesn’t consider the crowdsourced translation initiative a failure, and maintains its potential for broader application – excluding language education. “In a CAPTCHA scenario, the interaction with a user is brief, lasting only about ten seconds, making it inconsequential,” he clarified. “However, Duolingo fosters a sustained relationship with its users, which introduces a different set of considerations.”

The unexpected success of the app prompted Duolingo to discontinue its translation services business by the close of 2013. Rather than pursuing alternative monetization avenues, von Ahn communicated to investors that the company would concentrate on product enhancement and user base expansion, anticipating future revenue generation.

Murphy, the intern who later became Duolingo’s chief designer, recalls a shift in the company’s fundamental philosophy at this time. Without the translation component, the company was compelled to deliver a truly exceptional product capable of achieving high engagement and fostering customer loyalty. “Duolingo’s initial vision wasn’t simply to replicate Rosetta Stone at no cost,” he stated. “It was to understand why many individuals don’t develop a passion for learning and to create an accessible entry point.”

Addressing this question placed product growth at the forefront of the company’s engineering culture. This facilitated experimentation with gamification, educational techniques, and, as will be explored in the second part of this EC-1 series, the development of a memorable, albeit somewhat assertive, green owl mascot.