Why most language apps fail.

I. The streak lie

You know someone with a 700-day Duolingo streak who can't order a coffee in Spanish.

This isn't a glitch. It isn't user error. It's the product working exactly as designed.

The streak is not a measure of progress. It is a measure of compliance. It tells you, with great precision, how many consecutive days the user opened the app. It tells you nothing about whether they can use the language.

These are different things. The whole category has confused them on purpose.

II. What apps actually optimize for

Language apps are subscription businesses first, learning products second. The order matters.

A subscription business runs on retention. The metric that determines whether a language app is valuable to its investors is not how many users become fluent. It is how many users keep paying month after month. These goals are nearly opposite. A user who becomes fluent stops paying. A user who never becomes fluent — but keeps trying, keeps opening the app, keeps maintaining the streak — pays forever.

The dominant model in the category has solved this elegantly. It makes the practice of language learning feel rewarding in the moment, with little green checkmarks and shimmering animations and milestone celebrations, while quietly ensuring that the user rarely accomplishes anything substantial enough to leave.

You don't blame a business for following its incentives. You notice them.

III. What fluency actually is

Fluency is not knowing the most words. It is not having the longest streak. It is not finishing the most lessons.

Fluency is the ability to say something true about your life, in another language, without rehearsing it first.

It is the moment a waiter in Mexico City doesn't switch to English. It is the moment a stranger in a bar in Berlin doesn't slow down for you. It is the moment your spouse's grandmother understands the joke you just told.

These moments don't come from completing trees. They don't come from streaks. They come from spending hundreds of hours producing language about things you actually care about, and having someone or something correct you when you get it wrong.

IV. Why tutors work and apps don't

This is what a great tutor delivers, and what every great language student in history has had access to.

A tutor sits down with you. They ask what you do. They ask where you're going. They ask who you want to talk to and what's at stake. Then they teach you that — not the textbook, not the curriculum, you. They give you the words and structures you actually need, in the contexts you actually face, and they correct you when you produce them wrong.

This works. It has always worked. It is not pedagogy that distinguishes the great tutor from the great app. It is personalization. The tutor teaches your Spanish. The app teaches Spanish.

The result, over hundreds of hours, is that the tutor's student can talk about their life. The app's student can complete the app.

The tutor teaches your Spanish. The app teaches Spanish.

V. Your Word World

Every person moves through their own personal vocabulary every day — in their own language.

The finance professional's words are leverage, P&L, client, term sheet, allocation. The new parent's words are bedtime, tantrum, daycare, pediatrician, swaddle. The chef's words are reduce, sear, fold, deglaze, plate. The traveler's words are visa, transit, layover, exchange rate, reservation. Every one of us speaks a few thousand words in our own language every week — words about our work, our families, our hobbies, the things at the center of our actual life.

This is your Word World. Not the most common words in a language. The most common words in yours.

These are the words you need most in any new language. They are also the words no generic course can know. A standard curriculum teaches the same 2,000 high-frequency words to a finance professional and a chef, even though each would be better served by twelve hundred entirely different ones.

A great tutor learns what your Word World is, and teaches you that.

VI. The scale problem

So why hasn't every app done what the tutor does?

Because it couldn't. The economics didn't work. A great tutor costs forty dollars an hour. Personalized lessons require someone — a human someone — to know the student well enough to build them. At app scale, with millions of users, this was never possible. So the apps did the only thing they could do: they built one course and sold it to everyone.

This was a compromise born of the technology of the time. It was never the goal. It was the limit.

The limit has lifted. The work a tutor does — learning who someone is, building lessons around their life, correcting them in context — can now be done at scale. Not someday. Now.

This is the moment the category has been waiting for. Whether anyone uses it well is a different question.

VII. What real practice looks like

Real practice in a language is not a multiple-choice quiz. It is production.

You produce a sentence. The sentence is right or wrong. If it is wrong, someone tells you why. You produce another one. Over thousands of these small loops, your brain builds the structures that let you produce, eventually, sentences you have never produced before — about things you have never specifically practiced. This is what fluency is.

Most apps don't have you produce. They have you select. Tap the correct translation. Drag the words into the right order. Choose the matching image.

This is easier to grade. It is easier to gamify. It is easier to score. It is also, fundamentally, not what learning a language is. The student of an app like this gets very good at the app. That skill does not transfer to a conversation.

VIII. What we believe

This is the bedrock Arbor is built on — not as marketing claims, but as the arguments the entire product flows from.

Most language products are built to keep you. We're building the one that wants to set you free.

IX. The invitation

If any of this resonates, you are the person we built Arbor for.

Arbor is the app that learns your life before teaching you anything else. It teaches you the language you actually need, in the contexts you actually face — the way human tutors have always taught, except at the scale apps have always promised and never delivered.

We are not for everyone. We are not for someone who wants to feel productive. We are for someone who wants to actually become fluent — and is willing to spend a few months of consistent effort to get there.

If that is you, the waitlist is below.