Stop Adding AI Features. Start Building Something ChatGPT Can't Kill.

I’ve been watching what’s happening with Duolingo and honestly, it’s a microcosm of a much bigger problem across the entire software market.

Investors and users look at these new language models and ask a pretty reasonable question: why do I need a separate app when ChatGPT or Claude will soon do the same thing, only better and cheaper?

The market’s already pricing this in. Even if revenue hasn’t taken a hit yet. It’s fear of the future that’s killing valuations today.

Short term, I think Duolingo can recover. But only if they stop trying to adapt and instead double down hard on the one thing the big models can’t copy: user experience.

The winner isn’t whoever has the smartest model. It’s whoever built an ecosystem around their users that people don’t want to leave.

Not just bolting on a chatbot or checking the “AI features” box. I’m talking about building a real experience moat. The kind of moat that actually protects your castle.

Gamification that hits dopamine receptors just right. Social mechanics where you’re competing with friends. That little victory sound when you finish a lesson.

The feeling of a journey, of progress. Something a big faceless LLM just can’t give you.

For Duolingo, the hypothesis shouldn’t be “compete with models on intelligence.” It should be “create a unique environment where these models are useless without our context, our behavioral data, our experience design.”

They need to stop being just a language learning app. They need to become an environment where the learning process itself is the adventure.

Just adding AI features is signing up for mediocrity. Which I think is one of the biggest mistakes in business.

That’s defensive tactics. What you need is offensive. Build user experience so strong and emotionally charged that the thought “maybe I’ll just ask ChatGPT” doesn’t even cross people’s minds.

In a world where intelligence becomes as commoditized as electricity, the only sustainable advantage is the experience you give users.

That’s what you can’t copy by just training a model on more data.

So the choice for Duolingo (and a lot of other software companies) is pretty straightforward.

Either they build this experience moat and create unique, irreplaceable value for their users.

Or the market that wrote them off so early, looking far ahead, turns out to be right.