I have a 120-day French streak on Duolingo. I still can’t speak French that well. And I’m not alone: I hear the same thing from a lot of people who use it.

I haven’t skipped a day, and I really like Duolingo, but what it asks you to do barely resembles using a language. Real life is not filling in the blank of a sentence you have already seen fifty times, choosing from a row of words where only one could possibly fit.

Multiple choice without the choice

Take a typical exercise: fill in the blank in Je ___ au marché (“I ___ to the market”). On the left is the kind of question Duolingo mostly asks. On the right is one that would actually test grammar.

In the left one, only one option is a verb, so all you need to know is what a verb looks like. In the right one, every option is a form of aller, and the only way to pick correctly is to know the conjugation. Duolingo mostly gives you the left kind: the correct option is obvious, you get it right, and the game moves on.

That’s the problem in one sentence: learning requires failing, and the app is built so you rarely fail. Real learning means getting things wrong, spending time on why, and memorizing what you missed. Duolingo is optimized for something else: finishing lessons.

To be fair, Duolingo also has speaking and listening exercises, and those help. But most of the time you’re solving tests like the one on the left.

The streak is the product

At 120 days, the streak isn’t about French anymore. It’s a chore. It’s something I don’t want to lose, so I keep going. It has become a self-propagating habit rather than a learning activity.

I’ll probably keep the streak anyway, but I’ve stopped pretending it’s how I’ll learn French. The other standard route is the textbook, and it has its place: you need the grammar, and there is no skipping it. But I believe there are alternatives that are more fun, just as educational and more effective. I know because I’ve used one before, without meaning to.

How I actually learned English

I took the usual English courses as a kid, and I can’t claim they didn’t help. But the biggest gains came by accident, from YouTube.

I wanted to understand how certain mechanics worked in the games I played. There wasn’t content about them in Turkish, or I’d already consumed all of it and wanted more. So I searched in English. I played the games in English too: even when a Turkish translation existed, every guide and video I watched used the English terms, so it was simply easier to find things that way. The same went for channels I followed for years, like The Mighty Jingles and Lindy Beige.

At first there was a lot in those videos I didn’t understand. I translated the words I didn’t know, one by one, and learned them over time. And because I kept watching the same people, I kept hearing the same words. That’s the repetition Duolingo is going for when it asks you for chapeau for the fiftieth time, except every repeat came inside something I actually wanted to watch.

I wasn’t studying English in any of this. The English was a byproduct, the toll I paid to reach content I wanted.

Babies don’t learn languages with textbooks

Nobody hands a toddler a conjugation table. Children pick up their first language from a huge amount of input they only partly understand, surrounded by context, with a strong incentive to figure it out.

While writing this, I found out the idea has a name. The linguist Stephen Krashen calls it comprehensible input: you acquire a language by understanding messages that are slightly beyond your current level, focused on the meaning rather than the form (Krashen, Principles and Practice in Second Language Acquisition, 1982). From what I’ve read so far, the idea is influential and also debated, with many researchers arguing that output, feedback and deliberate practice matter too.

And adults aren’t babies. We have one big advantage over them: we already speak a language, and that language may be full of the one we’re trying to learn.

A goldmine hiding inside Turkish

Turkish is my native language, and it has thousands of words borrowed from French. Look at older Ottoman vocabulary and you’ll find even more, like the Ottoman word for railway, şimendifer, which is just French chemin de fer: literally, “road of iron”.

If you’re into etymology like I am, the French words Turks use every day are a free starter vocabulary:

Turkish	French	What it literally means
şimendifer	chemin de fer	road of iron (railway)
abajur	abat-jour	knocks down the daylight (lampshade)
randevu	rendez-vous	present yourselves (appointment)
şoför	chauffeur	heater, originally the stoker on early steam cars
pardösü	pardessus	over the top (overcoat)
vites	vitesse	speed (a car’s gear)
asansör	ascenseur	the thing that ascends (elevator)
kuaför	coiffeur	from coiffe, a headdress (hairdresser)
kuzen	cousin	cousin

Once you know the French root, the Turkish word stops being arbitrary, and the French word stops being foreign.

The same is true for English, which absorbed a huge share of its vocabulary from French after 1066. Many English words have a French relative you’d hear every day:

English	French relative	What it means in French
déjà vu	déjà vu	already seen
essay	essai (essayer)	an attempt (to try)
jettison	jeter	to throw
agent	agir	to act
dormitory	dormir (dormi)	to sleep (slept)

There’s an obvious limit. This works best when the languages you know share a lot of vocabulary with the one you’re learning. An Arabic speaker learning Norwegian won’t find much of a goldmine. Across French, English and Turkish, it works very well.

Learning French from videos

To learn French the way I learned English, I need French videos I actually want to watch, and a quick way to look up the words I miss. I’m building a tool for that, for myself first. I believe it will work because it’s the same loop that taught me English. If it does, I might turn it into something other people can use someday.

You give it the name of a song or the URL of any YouTube video:

• Songs: it finds the lyrics in an open-source lyrics database and translates them with AI into your language, in my case French into English.
• Any other video: it pulls the transcript from YouTube and translates it the same way. A French vlog works exactly like a French song.

The video plays on the left. On the right, the French line being spoken shows up in sync, with its English translation underneath, and the next one takes its place as the video moves on. Hover over any French word and you see what it means on its own. Think Google Translate meets Duolingo, for videos. There’s a quiz at the end, which is the least finished part and the one that makes it feel a bit like Duolingo, minus the obvious answers.

The plan is simple: find a French video I’d genuinely enjoy, drop it in, and learn French the way I learned English. Except this time, on purpose.

Where this breaks: stages

I should be honest about a gap. When I see a Turkish or English word with a French root, I can usually guess what’s going on, but that’s because I already have a French baseline. I’m not sure the video-plus-etymology approach works for an absolute beginner. Understanding half of a video is useful. Understanding nothing isn’t.

So I think the approach needs stages:

1. Foundation. Pronunciation, the most common words, core grammar. Yes, this is where a course, or even Duolingo, is actually useful.
2. Bridge. Mine the loanwords you already know. It turns “French is a foreign language” into “I know hundreds of French words already”, which is a much better mood to learn in.
3. Immersion. Videos and songs you’d watch anyway, with the app as scaffolding. This is where vocabulary grows fast.
4. Output. Actually speaking. None of the above fully solves this, and it’s the thing my 120-day streak was supposed to fix.

The streak continues

I’ll keep the streak. Probably. The little flame has emotional leverage over me at this point. But I no longer expect it to make me speak French. That’s going to come from a French vlog I actually want to watch, a hover over a word I didn’t know, and the quiet satisfaction of noticing it was in Turkish all along.