This week, we share our case study about how Duolingo became an AI powerhouse and why it’s eyeing all subjects next. The Senate just gave states the green light to regulate AI. Meanwhile, deepfakes are exploding across the internet, and it’s only getting worse. Let’s dive in, and as always, stay curious.
In a 99–1 vote, the U.S. Senate struck down a proposed 10-year ban on state-level AI regulation that was part of President Trump’s sweeping tax and spending bill. The amendment, led by Sen. Marsha Blackburn (R-TN), restores states’ power to regulate AI technologies.
Key Points:
The AI regulation ban would’ve prevented states from accessing a new $500M AI infrastructure fund.
Major tech firms like Google and OpenAI supported the ban, favoring federal-only oversight to avoid fragmented laws.
The final tax bill passed 51–50, but the AI restriction was removed.
Democrats and GOP governors united against the moratorium, citing child safety, robocalls, and deepfake threats.
Sen. Blackburn said states must act until Congress passes laws like the Kids Online Safety Act.
The Senate’s move signals growing bipartisan concern about the risks of unregulated AI, and a shift toward state-led protections amid stalled federal legislation.
How Duolingo Became an AI Company
From Gamified Language App to EdTech Leader
Duolingo was founded in 2009 by Luis von Ahn, a Guatemalan-American entrepreneur and software developer, after selling his previous company, reCAPTCHA, to Google. Duolingo started as a free app that gamified language learning. By 2017, it had over 200 million users, but was still perceived as a “fun app,” rather than a serious educational tool. That perception shifted rapidly with their AI-first pivot, which began in 2018.
🎯 Why Duolingo Invested in AI
Scale: Teaching 500M+ learners across 40+ languages required personalized instruction that human teachers could not match, and Luis von Ahn knew from first experience that learning a second language required a lot more than a regular class.
Engagement: Gamification helped, as it makes learning fun and engaging, but personalization drives long-term retention.
Cost Efficiency: AI tutors allow a freemium model to scale without increasing headcount.
Competition: Emerging AI tutors (like ChatGPT, Khanmigo, etc.) threatened user retention.
“We realized we weren’t just a language app — we were an AI education platform.” — Luis von Ahn, CEO, 2023
🧠 How Duolingo Uses AI Today
🚀 Product Milestone: Duolingo Max
Duolingo Max is a new subscription tier above Super Duolingo that gives learners access to two brand-new features and exercises, launched in March 2023 and powered by GPT-4 via OpenAI. Its features include:
Roleplay: Chat with fictional characters in real-life scenarios (ordering food, job interviews, etc.)
Explain My Answer: AI breaks down why your response was wrong in a conversational tone.
Result: 4x increase in daily active users of premium tier + 30% increase in time spent on app
📊 Business Impact
Share
🧩 The Duolingo AI Flywheel
User Interactions → AI Learns Mistakes & Patterns → Generates Smarter Lessons → Boosts Engagement & Completion → Feeds Back More Data → Repeat.
This feedback loop lets them improve faster than human content teams could manage.
🧠 In-House AI Research
Duolingo AI Research Team: Includes NLP PhDs and ML engineers.
Published papers on:
Language proficiency modeling
Speech scoring
AI feedback calibration
AI stack includes open-source tools (PyTorch), reinforcement learning frameworks, and OpenAI APIs.
📌 What Startups and SMBs Can Learn
Start with Real Problems → Duolingo didn’t bolt on AI — they solved pain points like “Why did I get this wrong?” or “This is too easy.”
Train AI on Your Own Data → Their models are fine-tuned on billions of user interactions, making feedback hyper-relevant.
Mix AI with Gamification → AI adapts what is shown, but game mechanics make you want to show up.
Keep Human Touchpoints → AI tutors didn’t replace everything — Duolingo still uses human-reviewed translations and guidance where accuracy is critical.
🧪 The Future of Duolingo AI
Math & Music Apps: AI tutors now extend to subjects beyond language.
Voice & Visual AI: Using Whisper and potentially multimodal tools for richer interaction.
Custom GPTs: May soon let educators create their own AI tutors using Duolingo’s engine.
Duolingo’s AI pivot is a masterclass in data-driven transformation. Instead of launching an “AI feature,” they rebuilt the engine of their product around intelligence, adaptivity, and personalization. As we become more device-oriented and our attention gets more limited, gamification can improve any app’s engagement numbers, especially when there are proven results. Now the company will implement the same strategy to teach many other subjects, potentially turning it into a complete learning platform.
DeepLearning Wizard — A hands-on, code-first resource that teaches deep learning, machine learning, and PyTorch from the ground up.
What makes it good:
Clear math + code explanations
Step-by-step guides for building neural networks
Great for beginners who want to go deeper technically
Deep in the Deepfake Economy
Small businesses are under siege from a new wave of AI-powered scams. From fake job listings to cloned websites and deepfake video calls, scammers are using generative AI to deceive at scale, and it’s working.
Stats:
Since ChatGPT’s 2022 launch, GenAI-enabled scams have quadrupled.
A finance clerk at Arup was tricked by deepfaked coworkers into approving a $25M transfer.
Oishya, a Japanese knife brand, had a scam clone site defraud 100+ customers.
On Amazon, AI-written books dominate bestseller lists using fake reviews.
Recruiters now face an epidemic of deepfake candidates faking video interviews.
AI tools can cheaply and convincingly mimic websites, voices, and identities. Scammers now operate like startups, scaling fast, lowering costs, and targeting more victims across industries.
🤖How Duolingo Became an AI Company was originally published in Coinmonks on Medium, where people are continuing the conversation by highlighting and responding to this story.