Christian Darnton — $DUOL Is Like Netflix 15 Years Ago

0:00 The thesis in one chart — FCF/share up, price down

• Duolingo's free cash flow per share is "going up and up" while the stock is "going down and down" — and over the long run stocks follow FCF/share. The stock is down 78% from its peak even as FCF/share rises at an increasing pace, which is "very, very rare."
• The investor's job is to anticipate future FCF/share; when it inflects and increases, the price later follows — so he expects "a violent re-rating."

1:03 AI disruption fear is "trivial" — the innovation-stack moat

• He put the odds of Duolingo being disrupted by AI at ~1% ("greater chance an asteroid hits your head"). The "no moat" claim for digital apps (Spotify, Netflix, Instagram, Duolingo) misunderstands the business.
• These apps can't be vibe-coded in a basement; they are thousands-to-millions of tiny iterations that "stack up over time" into a far more delightful, friction-free user experience — the innovation stack, cited in Duolingo's latest earnings.

2:36 Spotify proves the moat

• If digital apps have no moat, "what is the moat of Spotify?" Spotify took 60–70% audio-streaming share against Apple, Amazon and YouTube despite less cash — purely via the innovation stack.
• The innovation stack only arises from an organizational culture obsessively focused on the end user. He calls Duolingo the most qualitatively strong company he's "ever come across."

4:14 Not a language app — a superhuman AI tutor

• Duolingo is becoming a personalized, superhuman AI tutor democratized "to the world at the lowest cost possible," with the potential to disrupt a static, slow, generic education system — as good as (or better than) a one-to-one tutor, fun, and globally available.

5:25 The funnel and the only real bear case

• The funnel: monthly active users (top) → daily active users → paying subscribers. Duolingo is at ~12% paid penetration vs Spotify's ~50%; management aspires to that but isn't focused on it this year.
• The only serious bear case is execution: can they keep bringing more users into the funnel (re-accelerate flat MAU growth).

6:29 The Bezos algorithm — scale precedes monetization

• 2026 is an "investment year": Duolingo is deliberately sacrificing short-term monetization for long-term dominance — the Bezos algorithm — "more aggressively than I've ever seen." For digital apps, scale precedes monetization (get billions on the app, monetize later, like Spotify).

7:20 The TAM and the balance sheet

• ~3 billion people are learning math and English alone; Duolingo has just over 50M DAUs — under 10% penetration of even that figure. $1B cash, no debt, ~$400M FCF for full-year 2026.

8:14 The "increasing speed and efficacy" diagnostic

• His mental model: "Can this organization overcome problem X with increasing speed and efficacy?" Every business (Google, Costco, Amazon) hits problems it must overcome — Google faced the AI-kills-search fear months ago.
• For Duolingo's problem (top-of-funnel slowdown) there's "outstanding reason" to believe they can — management has a track record of focusing on one vertical and solving it fast, which the CEO cited.

9:48 DAU growth healthy; MAU is the metric to watch

• DAU growth still healthy (~20% expected in 2026); the DAU/MAU ratio keeps rising (engagement/retention) — "on par or just under TikTok" stickiness. The problem is MAU top-of-funnel, which was flat this quarter — "the metric to watch."

11:37 50M DAUs grown entirely by word of mouth

• All MAU growth so far has come from word of mouth on "close to zero" performance marketing — the CEO admits it's "totally crazy" and they should start investing in marketing (which they're now doing).
• Per the Bezos quote, word-of-mouth spread is indicative of a product's true utility and stickiness — exactly what he saw at Palantir (~3 salespeople vs sales-heavy Snowflake/Databricks).

13:01 Word of mouth ← teaching efficacy ← AI

• Word of mouth occurs when the product produces learning outcomes; AI now lets Duolingo improve teaching efficacy faster (e.g. voice mode instead of clicking buttons boosts retention). There's a lag between improving efficacy and word-of-mouth spread — you can't flip word of mouth like a switch.

14:55 Adding and monetizing network nodes; Asia next

• The whole question reduces to: can Duolingo keep adding nodes to the network and monetize them over time? He thinks yes — including new performance marketing in Asia/China, where uptake is just beginning and "they love learning."

15:47 Saturation: management says "nowhere close"

• The bear case in one sentence is saturation; management explicitly denied it in Q4'25. Luis von Ahn: ~2% of US internet users use Duolingo on a given day (3% UK, 4% Germany); if every country reached just the US 2% rate, DAUs would more than double.

17:04 AI makes Duolingo stronger; proprietary-data moat

• In Q1'26 Duolingo shipped 20,000+ course units (10x its output of two years ago) thanks to AI. The CEO says generic LLMs "are not good enough" — they must be fine-tuned, which requires Duolingo's proprietary data: "the most important moat ever." A grandma using Claude can't replicate it.

18:05 A "monopoly on sale" and the AMD track record

• ~$400M FCF, ~£500M buyback, ~85% of global language-app DAUs — "a monopoly on sale right now." Echoes his $500-by-2026 AMD call that was mocked but proved right; the key question for investors is whether management can re-accelerate top-of-funnel growth.

19:45 Conclusion — building a superhuman AI tutor

• "Duolingo is not a language app, it is building a personalized global superhuman AI tutor at the lowest cost." Moats: proprietary data, innovation stack, network effects, brand, and organizational culture. He happily publicizes his thought process and holdings — some may fall 20–40% in a year, but rising FCF/share wins long term. "If you're watching this in 2029, you'll know if I'm terrible or good."