Duolingo for 100 Days: What I Learned, What I Wish I'd Done Differently

The setup

I opened Duolingo on a Tuesday in May 2023. I had a two-week trip to Tokyo booked for December and a vague hope that I could stop being the person who just points at things. I set a goal of 15 XP a day. Duolingo raised it to 50 XP after week one because I was “on a roll.” I kept pace for 100 days.

The streak became the scoreboard. I stopped thinking about whether I was learning Japanese and started thinking about whether I was keeping the streak.

What worked

Here is what Duolingo did well for me.

• Habit formation.The streak is a legitimate habit-building mechanism. It took me from “thinking about learning Japanese” to “touching Japanese every day.”
• Initial hiragana exposure. By day 10, I could recognize about 20 hiragana. Not all 46, but enough to be reading slowly.
• Low commitment per session.5 minutes was all that was needed to keep the streak. That removed the “I do not have time” excuse.
• Pattern recognition. I started seeing Japanese sentence structure (subject-object-verb) and particles enough to guess at meaning.

What didn't

Here is where the 100 days failed me.

Hiragana took 40 days instead of 3

Duolingo spreads kana across many units, mixed with vocabulary. That is great for retention (you see kana in context) and terrible for speed. A focused 3-day push would have covered all 46 in one weekend. Duolingo took me 40 days to get there.

Pronunciation was a synthesized lie

Duolingo's Japanese TTS gets many words wrong, especially around pitch accent and the う and い vowel sounds. I spent 100 days practicing a version of Japanese that sounded robotic and was in places actively incorrect. When I first heard a native recording, I realized I was going to have to unlearn my listening habits.

Multi-choice produces recognition, not recall

Most Duolingo Japanese exercises are multiple choice. You tap the right answer from a list. That trains recognition, which is a weaker memory trace than recall. Real-world Japanese requires recall: you hear the word and have to generate the meaning, or see a blank and have to produce the word.

No schedule awareness

Duolingo's scheduling is opaque. I did not know which words were due for review and which I was being shown because the algorithm felt like it. Compared with an honest FSRS scheduler, it felt like random repetition.

What I'd do differently

1. Learn kana in a weekend, not a month

Four-day push on hiragana. Three on katakana. Do not let an app stretch them across 40 unit squares.

2. Use a real-audio tool from day one

Duolingo's TTS was my biggest regret. I built a listening model that I had to rebuild from scratch at day 120. If you cannot hear actual recorded Japanese on every card, you are not learning Japanese.

3. Set a learning goal, not a streak goal

“100-day streak” is a participation trophy. “Read a children's book” is a learning goal. The second one keeps you honest.

4. Measure progress in produced sentences, not XP

Every week, write three Japanese sentences about your life in a notebook. Look at the notebook a month later. If the sentences are getting longer and more accurate, you are learning. XP is noise.

5. Pick a tool that knows when to stop

Duolingo's goal is to keep you there. That is often at odds with your goal, which is to leave the app knowing more Japanese than when you opened it. Inku's review session ends when the review ends; there is no infinite feed. That design choice is a feature, not a bug.

The bottom line

Duolingo is not a bad tool. It is a tool optimized for daily engagement in a mass market. If that aligns with your actual goal, great. If your goal is to learn Japanese in a reasonable timeframe, pick something else.

I spent 100 days on Duolingo and then 30 days on Inku. The 30 moved me farther than the 100 did.

Common questions

Related reading

• Inku vs Duolingo comparison
• Duolingo alternative

Try Inku, free for 7 days, or read the full roadmap.