Streaks are a trap. Here's what to count instead.

A friend of mine has a 1,247-day Duolingo streak in French. She mentions it in passing the way you'd mention having lived in the same apartment for a long time. She is also, to her own enormous frustration, unable to introduce herself to her French landlord without going into a small panic spiral. Both things are true at the same time.

I think about this a lot. The streak is real. The skill she was hoping it would buy her never showed up.

She isn't unusual. Spend ten minutes in any language-learning forum and you'll find hundreds of people in this exact situation. They've been "studying" for years, by their own definition of studying. The number gets bigger. The fluency does not.

The trap isn't the app. The trap is what the streak is actually measuring.

What a streak counts

A streak counts whether you opened an app. That's the whole thing. It doesn't notice whether you learned anything, whether what you learned was useful, whether you can produce any of it, or whether the time you spent was the right kind of time.

In the early days of streak design (Duolingo wasn't even the first to use them, just the most successful), this metric was chosen because it was the only one that could be cheaply measured. The product can tell if you opened it. It cannot tell if you got better. So it tracks the proxy and calls it consistency.

The trouble is that for the first few weeks, the proxy is honest. You start from zero. Every day you open the app, you do learn something. The streak and the actual learning move together. You feel virtuous; you are also actually getting better. The experience is real.

Then somewhere around day 60 or 90, depending on your starting point, you hit a plateau. You're still opening the app. You're still tapping the right answers. But what's increasing now is your recognition of the language, not your ability to produce it. Recognition is the easy half of fluency. Production is what you'll actually need at the cash register, at the dinner party, in the job interview. Production gets practiced about 5% as much as recognition.

By day 200, the streak is still going and the plateau is still there. By day 500 the gap between your streak number and your fluency has become embarrassing. By day 1,247, you've made the streak a personality trait without making the language a skill.

Why streaks feel like progress anyway

Game designers borrowed streak mechanics from gambling, and they work for the same reasons. A streak is variable-cost (you have to do something every day) and high-cost-to-lose (one missed day breaks it, and the bigger the streak the bigger the loss). That asymmetry creates a feeling of accumulated investment that has no necessary relationship to whether the activity is achieving anything underneath.

Once a streak gets long enough, you keep going partly to protect the streak. That's a perfectly fine reason to do something. It just isn't the same reason as "I'm getting better at French." The two motivations start identical, drift apart silently, and end up unrelated. The streak becomes a separate goal, in quiet tension with the original goal of fluency.

Worse: when you finally do break the streak, the natural emotional response is rarely "great, I'm free of this, now I can practice in a way that actually works." It's "I lost the streak, the project is dead." So people quit entirely, having spent years failing to learn the language they wanted to learn, mourning a number that stopped meaning anything around month four.

What to count instead

The fix is to count something that actually correlates with the skill you want.

For language learning past the absolute-beginner stage, the things that predict fluency aren't days-on-the-app. They're more like:

Minutes spoken out loud. Out loud. Not tapped. Not read. Spoken, in your target language, in real or simulated conversation. This is the metric the apps don't show you, because it's the metric that exposes how little of it most learners actually do.

Conversations completed end-to-end. Not phrases. Conversations. A back-and-forth with a real partner or a patient simulated one, where you started something, navigated it through a misunderstanding or two, and finished it. Two of these a week beats two hundred vocabulary drills.

Words used in production for the first time. A word you recognize is not a word you can use. A word you've said out loud, in context, becomes one you can pull from later. Counting "first use" makes vocabulary feel like building something rather than like checking a list.

None of these are easy to gamify, which is one reason no big app has tried. They require either a real partner, who's slow and expensive to schedule, or a patient simulated one, which is much harder to build than a flashcard generator.

A reframe worth trying

Forget the streak for a week. Try this instead:

Pick one production goal. Not a process goal ("open the app"). A production goal: have three conversations in your target language; complete a café order out loud five times; use the past tense correctly in a sentence you've never said before. Track only that.

You won't have a number to defend on social media. You won't get a green notification badge. You'll just, slowly, start being able to do the things you originally said you wanted to do.

This kind of progress is slower and less visible than a streak. It is also the only kind that becomes fluency.

If you want a patient place to do the production half, Glauda was built for exactly that. No streak counter. No daily reminder shaming you about the badge you lost. Just a partner that waits while you find your words. Open it when you want to do the real work. Close it when you don't. Your fluency will not care about the days.