Forgot your password?
typodupeerror
Businesses AI

Duolingo CEO Says They've Stopped Tracking Employees' AI Use for Performance Reviews (entrepreneur.com) 34

Last May Duolingo's stock peaked at $529.05. But while the learning app passed $1 billion in revenue in 2025 and 50 million daily active users, today its stock price has dropped more than 81%, to $100.51.

And there's been other changes, reports Entrepreneur: In April 2025, Duolingo CEO Luis von Ahn made headlines after writing a memo calling the company "AI-first." In the memo, von Ahn announced that the language-learning platform would track employees' AI use in performance reviews. Now, a year later, von Ahn is backtracking and rethinking how he measures employee performance. He told the Silicon Valley Girl podcast earlier this month that Duolingo no longer considers AI use in performance reviews.

The change arose after employees started to ask, "Do you just want us to use AI for AI's sake?" von Ahn explained. "We said no, look — the most important thing in your performance is that you are doing whatever your job is as well as possible. A lot of times, AI can help you with that, but if it can't, I'm not going to force you to do that," von Ahn said on the podcast. He felt as though the company was "trying to push something that in some cases did not fit" instead of "being held accountable for the actual outcome." The CEO is, however, still sticking to other "constructive constraints" he introduced in the April 2025 memo, including stopping contractor hiring in cases where AI can assume their workload...

Von Ahn also mentioned that a few months ago, Duolingo had a day dedicated to vibe coding, or prompting AI to create an app without manually writing a single line of code. Every single person at the company, from engineers to human resources professionals, had to vibe code an app. Vibe coding has made an impact at the company. One of Duolingo's latest offerings, a course teaching users how to play chess, arose when two people vibe-coded the first prototype of it, the CEO said. Neither of them knew how to play chess or program, but they managed to use AI to create the whole chess curriculum and a prototype of the app in about six months last year. Now chess is Duolingo's fastest-growing course, according to von Ahn. "At this point, we have seven million daily active users that are learning chess," the CEO said on the podcast.

Duolingo CEO Says They've Stopped Tracking Employees' AI Use for Performance Reviews

Comments Filter:
  • Six months? (Score:5, Insightful)

    by thecombatwombat ( 571826 ) on Saturday April 18, 2026 @11:49PM (#66100822)

    They made every employee vibe code, and OK it made a chess app.

    IN SIX MONTHS?!?!? And they still emphasize it was just a prototype?

    That's the impressive AI fast timeline?

    Oh my, no one has ever taught Chess before. What an efficient, innovative thing that could never have happened except in this wonderful future.

    Just WTF.

    • Don't be so negative!

      First, the employees didn't know how to play chess but needed to fake a teaching app. So it's super impressive that they just cobbled up some random ideas that they picked up from other employees about learning Spanish.

      Then if that isn't impressive enough, they asked an LLM to find some low quality chess examples from its internet memory. Just to make a set of slides, and that didn't even use all the little pictures! There was this embarrassing early checkers lesson that they had to f

    • That was my reaction. They used AI and it still took SIX MONTHS to make a dang protoype? That only makes sense is they worked on it for 10 minutes a day.

    • I am confused. Why were they writing things for chess? Does that have anything at all to do with what their company does? Or was it just a demo? And if it was a demo: **6 months** writing a demo!?!?

  • The amount of time to make the app is important but TFS is not very clear. One day or six months? It makes a big difference for the story. So what is it?
  • This product is the manifestation of enshittification as defined by Doctorow; shows how important it is to have an app economy that is based on open source.

    What a disgrace to spy on employees.

  • by unixisc ( 2429386 ) on Sunday April 19, 2026 @01:51AM (#66100876)

    Actually, this is one of those rare cases where AI does help improve the product. Since any translation package would beat most linguists alive today, but AI would greatly help insert context for the translations

    That said, it's good that Duolingo is no longer coercing employees to use AI for their work

    • Since any translation package would beat most linguists alive today

      Citation needed. For those who want to believe your statement, it helps when one doesn't live anywhere near a border with another country.

      • Even near a country's border area, one isn't likely to need to know more than 2, or at the most 3 languages. Whereas this software knows a whole bunch of them. How many people is one likely to meet who know English, German, Spanish, French, Portuguese, Italian, Russian, Polish, Arabic, Mandarin, Turkish, Swahili, Dutch, Afrikaans and a whole bunch of other languages?
    • Agreed, for learning a language, "AI slop" is plenty good enough, and beats the cheesy practice sentences and stories that humans come up with!

    • by Njovich ( 553857 ) on Sunday April 19, 2026 @11:51AM (#66101306)

      Since any translation package would beat most linguists alive today,

      It's always funny, any time you hear someone say that somebody else's job could easily be replaced with AI.
      It's universally about a job they have never done a day in their lives. Programmers, Accountants, Customer Support staff, all supposedly very easy to replace by just plugging in ChatGPT according to 'experts'.

      In reality this is just just a fundamental misunderstanding of what an accountant, CS staff or linguist does.

      • by mjwx ( 966435 )

        Since any translation package would beat most linguists alive today,

        It's always funny, any time you hear someone say that somebody else's job could easily be replaced with AI.
        It's universally about a job they have never done a day in their lives. Programmers, Accountants, Customer Support staff, all supposedly very easy to replace by just plugging in ChatGPT according to 'experts'.

        In reality this is just just a fundamental misunderstanding of what an accountant, CS staff or linguist does.

        This.

        I'm no linguist but I speak enough Spanish to know how limited machine translators still are. They're fine if you don't speak any of the language and need to find the bookshop or can help to order a meal, a succulent Chinese meal but they lack any understanding of idioms, slang, nuance, tone or context. This is before you get to local dialects, abbreviations or typos.

    • AI translations still need LOTS of human help to have any sort of quality. AI tends to mostly stick with word-for-word translations, which often screw up idiomatic expression and often is very stilted in the target language. Yes, it is much better than nothing - and can even be the starting place for a good translation, but I wouldn't count on it for anything important.

  • Thanks Duolingo for telling us this.
    I don't want to support the kleptogarchy, directly or indirectly.

  • by gweihir ( 88907 ) on Sunday April 19, 2026 @06:45AM (#66101028)

    LLM-type AI has some uses, but it is not the "God machine" many people seem to believe, after ample prompting by the LLM pushers. It is just a tool. It can do some things somewhat well but unreliably and needs a lot of manual oversight by actual experts. "Vibe coding" is somewhat suitable for mock-ups (which is useful), but cannot create production stable / secure / maintainable code and thereby fails basically all fundamental requirements for production code.

    In the end, we will see what we saw with all other AI hypes: Productivity increases in the 1..10% range for some very specific things, productivity decreases in many others things that got pushed. Not a surprise. Only because of unfettered greed and customer stupidity did things get scaled to a completely irrational level and hence there will be a real crash this time, with real damage. The last AI hypes basically fizzled out quietly. Let's hope the sure-to-come next AI winter is bad enough that it makes it amply clear for at least a few decades that computers are not magic and cannot be magic and people should stop believing that.

  • I hate using the word. Seventy years ago I was taught "sh*t" does not belong in polite discourse. The ... now ... I read agentic.ai creates a chess-learning program. No need for the neophyte to study Nimzovitches MY SYSTEM ... no need to keep a copy of Modern Chess Openings next to your bed ... no need to totally immerse yourself in Tals sacrificial masterpieces .. no need to buy that 1960s Yugoslav pamphlet presenting the 20 major lines of Schliemanns defense against the Ruy ... no need .
    • I hate using the word. Seventy years ago I was taught "sh*t" does not belong in polite discourse.

      You can always use "enpoopify." Everyone will know what you mean.

  • by Tony Isaac ( 1301187 ) on Sunday April 19, 2026 @10:34AM (#66101196) Homepage

    Yep, they have a dashboard that shows, for each employee, on which days of the month they used AI. It doesn't matter what they used it for, or how often they used it, just that they used it. And the executives don't want to see any blank days.

    For some odd reason, they aren't looking at the other dashboard that shows that pull requests aided by AI, are taking longer than PRs not aided by AI.

    The hype is overwhelming at this point. If you dare challenge the "AI makes work go faster" mantra, you are short listed for the exit door.

    I'll be thrilled if the hype dies down in only six months.

  • I completed their Spanish Course, and am working on German which I studied in high school and college.

    One thing which I've noticed in the last few months is the number of hangs on Android when completing a lesson. I have to abort the lesson and start over. Whatever they have done seems to have affected stability in a bad way.

    Also there seem to be a lot of cheaters who get an astronomical number of points and place at number 1 in the periodic tournaments. I don't know what's going on here, but you would th

    • I used Duolingo for a year, learning German. I also learned that that vocal matching game you can do repeatedly and that ramps up and up to high scores - you can do it repeatedly and get very, very good at it and get to the top of the tables fairly consistently. I wasn't cheating, I was just getting really fast at one specific exercise. It also meant I gave up on everything else, all the standard grammar lessons etc. and stopped learning anything much useful.

      I then learned that it's cheaper and more effecti

    • I'm learning French, and I have to say, it's a generally good experience, and it seems to actually be teaching me something. I've also noticed a few problems with the app, no idea if they're new problems or old, or just 'load' or whatever. I've noticed that sometimes, no matter how I say I word, it'll just never accept it. It asks again a couple more times, so I try some variations, but it just refuses (whereas you can stammer your way through some of the sentences and it says it's okay!?). I've also found

  • Look folks, in order to win this metric game you need to be doing something where you are answerable to as few superiors as possible. It didn't used to be this way where minutia was tracked and analyzed. Now all of the leadership running businesses want to balance everything on the point of a pin.

    Don't they understand how unstable a system managed like that is?

    One slight perturbation in the metrics, can cause great instability in such a system. Think of it as the "butterfly effect" in chaos theory. One sli

  • We need to talk about your flair.

    Really? I... I have fifteen pieces on. I, also...

    Well, okay. Fifteen is the minimum, okay?
  • I have been training Spanish for around 400 days. the last change broke everything. they said something about the track i was on was changing but the level would be the same.
    For about a month now(iirc) i cant follow any of the conversations anymore. often i only understand 1 word in each sentence and i have no idea what they are talking about.
    It is probably a good time for me to just stop. :/

try again

Working...