🤖How Duolingo Became an AI Company

Plus: Deep in the Deepfake Economy

In partnership with

Get in Front of 50k Tech Leaders: Grow With Us

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.

  • How Duolingo Became an AI Company

  • 📰 News and Trends.

  • Senate Removes AI Regulation Ban from Megabill

  • 🧰 AI Tools - Architecture

  • Deep in the Deepfake Economy

  • 🧠 Learning Corner - DeepLearning Wizard

Get access to the most exclusive offers for private market investors

Looking to invest in real estate, private credit, pre-IPO venture or crypto? AIR Insiders get exclusive offers and perks from leading private market investing tools and platforms, like:

  • Up to $250 free from Percent

  • 50% off tax and retirement planning from Carry

  • $50 of free stock from Public

  • A free subscription to Worth Magazine

  • $1000 off an annual subscription to DealSheet

  • and offers from CapitalPad, Groundfloor, Fundrise, Mogul, and more.

Just sign up for our 2-week free trial to experience all the benefits of being an AIR Insider.

🌐 Other Tech news

Senate Removes AI Regulation Ban from Megabill

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

🧩 The Duolingo AI Flywheel

User InteractionsAI Learns Mistakes & PatternsGenerates Smarter LessonsBoosts Engagement & CompletionFeeds 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

  1. 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.”

  2. Train AI on Your Own Data
    → Their models are fine-tuned on billions of user interactions, making feedback hyper-relevant.

  3. Mix AI with Gamification
    → AI adapts what is shown, but game mechanics make you want to show up.

  4. 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.

🧠 Learning Corner.

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.

  • 25% of small businesses faced at least one AI scam in the past year.

  • Microsoft blocks 1.6M+ bot signups per hour.

  • 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.

What Can We Do:

  • Verify IDs before interviews.

  • Use tools like Spokeo and Beyond Identity to authenticate users.

  • Educate customers about scam detection.

  • Watch out for too-good-to-be-true offers with AI-polished visuals.

AI isn’t just building businesses, it’s also arming scammers. As one cybersecurity expert put it: “This is the industrial revolution for scams.”

🧰 AI Tools

Architecture

  • Spacemaker (by Autodesk) - Uses AI to optimize building layouts based on sunlight, wind, noise, and zoning. Site planning and early-stage development

  • Hypar - Automates architectural workflows using generative design functions. Parametric design and rapid prototyping for commercial buildings

  • ArkDesign.ai - AI-powered floorplan and room layout generator from natural language prompts. Quick residential layouts and conceptual design

  • LookX AI - AI-generated architectural visualizations from sketches or references. High-end renders, facade studies, and aesthetic inspiration

  • TestFit - AI-driven feasibility studies for multifamily, industrial, and urban developments

Download our list of 1000+ Tools for free.

Subscribe to keep reading

This content is free, but you must be subscribed to Yaro on AI and Tech Trends to continue reading.

Already a subscriber?Sign in.Not now

Reply

or to participate.