Please add this new reading feature

Clozemaster
#1

I don’t know if the developers check these message boards, but there is a gapping hole currently in the cloze reading section. The stories are random, boring, unrelated to what I’ve been learning, but I have a potentially awesome idea.

Create a “generate story” option where a LLM like ChatGPT, Gemini or Claude generates a random story based on your explicit progression within clozemaster. You should be able to add tags so the model can generate a story you actually want to read.

Here’s an example I tried with ChatGPT just as a test for Romanian

Keywords / Theme: Dog, cat, little boy, friends, playing, outside, school, happy, weather.

Într-o zi, un băiat mic mergea la școală.
Afară era soare și vremea era frumoasă.
Pe drum, băiatul vede un câine mic.
Câinele se apropie și băiatul îl mângâie.
Apoi, o pisică vine lângă ei.
Câinele și pisica devin prieteni cu băiatul.
După școală, ei se joacă afară împreună.
Băiatul aleargă, iar câinele și pisica îl urmează.
Toți sunt foarte fericiți și se joacă împreună.
Seara, băiatul merge acasă cu prietenii lui.

Translated:

One day, a little boy was going to school.
Outside it was sunny and the weather was beautiful.
On the way, the boy sees a little dog.
The dog comes closer and the boy pets him.
Then, a cat comes next to them.
The dog and cat become friends with the boy.
After school, they play outside together.
The boy runs, and the dog and cat follow him.
They are all very happy and play together.
In the evening, the boy goes home with his friends.

Some of you may think this is a very basic story, and that’s because it is. That’s because I have been learning Romanian for 2 weeks, not 6 months. I want a story that is simple, quick and accurate for my skill level.

I really believe being able to do this inside clozemaster would be great because the story the LLM would generate will always be in tune give or take, with your skill level. And you can generate stories about whatever interests you and at whatever your skill level is.

What do you guys think?

2 Likes
#2

Unfortunately, ChatGPT and LLMs are not that good at languages yet. I tried to use it for learning French and it made heinous errors daily. It could be a starting point and then a qualified person would have to check for and correct the errors. Quote from this article: “a study by OpenAI found that GPT-3, despite its impressive capabilities, still generated incorrect or nonsensical responses approximately 15% of the time.“

There are others that are anti-AI for different reasons. But that is a different topic. Given the current limitations of LLMs, depending on it for language-learning is problematic, to say the least.

TLDR: We aren’t there yet.

Edited to add another quote from the article about AI hallucinations and inaccuracies (#2 in the article):
“The implications of these inaccuracies are profound, as they can generate misleading or non-existent information. For example, there have been instances where GPT-3 has generated completely fabricated historical events or incorrect scientific facts, such as claiming that “The Great Emu War of 1932 was a conflict between Australia and New Zealand over emu migration,” which is entirely false.“

1 Like
#3

ChatGPT 3 is 6 years old! The difference now is incredible. And I’ve checked their translation ability by cross checking with other LLMs. So if Claude generates some text, I will sent it to ChatGPT to verify and vice versa and they all agree on the translation.

I actually asked these LLMs why they’re so good at translating and they basically said something along the lines of “We’ve been trained on billions of pieces of text, including every free course, open source project and online learning. We know all there is to know about translating languages”.

Obviously they’re not 100% perfect all of the time, especially when there is very special translation nuances, but it’s going to be accurate like 99.9% of the time which is fine for learning a language.

If you’ve not tried an LLM recently, I suggest you give it another try and see if you still encounter the same inaccuracies.

1 Like
#4

You have a confidence in LLMs that many don’t share. It is true that 3 is an old version, but I’m still seeing a lot of errors. I’m seeing a top accuracy rate of 94% for ChatGPT5, and that is for Math computations, not complicated language rules. OpenAI puts a warning on every conversation: “ChatGPT can make mistakes. Check important info.” Does that give you confidence?

What I said still stands, every item has to be checked and corrected. Here is an article comparing ChatGPT 3, 4, and 5.

I’m concerned how much confidence people have in LLMs when the LLM answers are sometimes complete hallucinations! This blog is also making the case that version 5 is “good enough“. It is too soon to start relying on it.

1 Like
#5

I dont know why you are making Maths a measurement for a LanguageLearningModel when it comes to learning languages. Given the fact, that even in mathematics AI is heavily outpacing humanity right now, even this Argument seems pointless. If AI is good at one thing, then it is reproducing Language.
In my Experience Gemini never even once made neither a spelling nor a grammatically mistake (in german), allthough I cant excactly tell if Chatgpt meets the same standards. This also explains why that many students in University (also a big problem in my uni) use it entirely in homeexames and essays.
The weakness of AI is in my opinion mostly not beeing able to explore deep conceps and explain niche topics. But I thinks this shouldnt be in our range of focus, when trying to acquire a language.

#6

I mentioned Math because the quoted article rated LLMs as being BEST at Math with more accuracy than other areas. Even then, the accuracy is only 94%.

But the point was that at this point in time, anything that LLM generates has to be checked, verified and corrected by a qualified human. I tried several of the top LLMs, none of them were that good, IMO.

But I do agree that AI is not good at exploring deep concepts and explaining niche topics. But it always told me it was good at it.

1 Like
#7

I’m sure every source of learning a language has some degree of error. Even right here in clozemaster sometimes I’m seeing translations that are just inaccurate, because they are pulled from community source tatoeba or whatever it’s called.

Every app can have mistakes, the argument I’m making now is that LLMs ability to accurately translate is so good that it absolutely clears the bar for being allowed to be used for learning.

#8

I support this suggestion, mainly because it would make it easier to actually study this type of content inside clozemaster, rather going through all the steps of generating it outside and manually importing a new story every single time. It could even give you some narrative prompts if you’re not feeling inspired.

I would add one small suggestion on top: a checkmark that gives the option to make the story expire after a certain period, say 24 hours. This would prevent “My Cloze-Texts” to become bloated with stories that the user is not really interested in keeping to read later. As a bonus, it would also provide a tiny incentive to actually finish the story within that period.

1 Like
#9

LLM creators themselves say to check everything. Anything that is created by an LLM has to checked, verified, and corrected by a qualified person. I think I have said everything I have to say on the subject.

1 Like
#10

They say that because LLMs are not infallible, just as humans are not infallible. But if you seriously don’t like to use LLMs as an aid to your learning because they might make a small grammatical error 0.1% of the time, that’s your decision.

Not being rude but we are language learning here, not launching rockets into space with people on board.

#11

The article I posted said up to 15% error rate. 6% error rate was for complex Math problems.

Nobody anywhere is claiming 99.9% accuracy for LLMs. When I confronted AI with its errors, they would flip and say the opposite, and then proclaim “Now I’m right.“ Using AI for language learning was an exercise in frustration. It did not help me with learning complex language rules.

I’m not saying it will never get there, it isn’t there yet for language learning.

Again, I’m repeating myself.

1 Like
#12

The article you posted was 6 years old, that’s the real problem here. Do you know how bad LLMs were 6 years ago? Anyway you’re not going to try the newest versions of these LLMs so there’s no point arguing anymore. You’re probably a textbook / official material only kind of learner and that’s fine.

1 Like
#13

The first article was dated 2026. The second article was dated 2025. Chat-GPT still hallucinates, still makes errors. Hallucinating is when Chat-GPT makes stuff up. Still a problem.

1 Like
#14

Oh my god, this is going to be my last post explaining this because you quite literally don’t know what you’re talking about at this point.

The article was not posted in 2026, it was edited in 2026. ChatGPT 3 was released in 2020.

I’m not replying to you anymore.

1 Like
#15

The second article that I posted is dated 2025. I actually liked this article because it seemed to be a fair and neutral article talking about what ChatGPT does well and what it doesn’t. It compares

  • GPT-3.5
  • GPT-4
  • GPT-5

I have used all 3 versions. It was an exercise in frustration for learning a language. I have never had a job where an error rate of 5-15% was acceptable. Gee, I got 90% of the dishes clean. Who cares? It isn’t rocket science! (AI is not free, it incurs cost which at times can be significant. I don’t work for Clozemaster so I’m certainly not the decider.)

I’m looking to see if I can find a more current study.

Here is a quote from the article:
“What Are ChatGPT Hallucinations and Why Do They Happen?

Critical insight: Even GPT-5 occasionally presents fiction as fact with complete confidence.

“Hallucination” in AI terms means generating believable but false information. GPT-4’s hallucination rate dropped to 28.6% for citation accuracy (down from GPT-3.5’s 39.6%), but it hasn’t disappeared entirely.

Real example: ChatGPT might confidently cite a study that doesn’t exist, complete with realistic journal names and publication dates.“

1 Like
#16

Duolingo stories are good. They’re getting better. In section 5 (they’re are 8 for German), there’s 250 units and one story for each unit. And I’m finding Video Call with Lily very helpful. Unlike most AI conversations, it doesn’t show any text until after your done (transcripts, with helpful suggestions). DUO is getting a lot better. Add in Clozemaster, and those two are my favorites. And I’ve tried a LOT of apps. Clozemaster invaluable for me, and I see a ton of words I’ve learned in here show up in German subtitles when watching a movie or a series. Important words, the ones frequently used. Because I see them so often, they get activated from passive vocab pretty quickly.

But yes, story telling is a very effective method. 100% agree.

1 Like
#17

Thanks for the post! Do you have in mind generating the stories with a word missing from each sentence like in Cloze-Reading, or as stories you can read like you pasted in and optionally click on to get the translation, add to a custom collection, etc. (like in the mobile app for Spanish from English, for example, if you go into Account there’s a toggle for “Clozemaster Reading”, a new feature we’re experimenting with)?

1 Like
#18

As full stories we can read, no words missing. Essentially I just want to practice my reading, pronunciation and speaking speed. It would be nice as well if the entire story had TTS so we can shadow along.

#19

You are committing a logical fallacy and draw a false equivalence.

  • Calling a Pitbull a Bull Terrier is wrong.
  • Calling a Pitbull a stone is wrong.

Yet these two statements aren’t equally wrong. A Bull Terrier is still a dog of the terrier family whereas a stone isn’t even a living organism. When someone mistakes a dog for a stone, you can’t be as lenient with them as with someone who mistakes a dog for some similar breed of dog.

Stop justifying slop-producing LLMs with the observation that humans also make mistakes. It only reveals that you don’t understand anything about “AI”, neither about the technologies nor about the fascist eugenicist cult the people in Silicon Valley are in. Look up TESCREAL to learn more. This isn’t an attack, it’s well-intentioned advice. I don’t want you to embarrass yourself in front of people who know more than you.

1 Like
#20

This is blatantly false.

I studied both maths and AI.

LLMs suck at math.

Edit: According to your later comment, you haven’t finished university yet.

LLMs are comically bad at math. It could be funny if there wasn’t so much crime, fraud, destruction, and a literal death cult involved. (Again, check TESCREAL or Ghost in the Machine.) It’s depressing to see such falsehoods being spread.

OpenAI could solve the Navier-Stokes equations only because it stole the ideas that human mathematicians had typed into ChatGPT before the human mathematicians could publish their findings. And then OpenAI threatened the victims of their theft to stay silent or OpenAI would ruin their careers. And even then OpenAI still had to throw 22 million USD on the problem to solve a 1 million USD problem. What efficiency! Only 2200% of the prize money.

The “AI” CEOs are all lying sociopaths, the media is either also lying or doesn’t have the technical background to understand anything they write, and you believe them.

We are throwing trillions of dollars on “AI”. This is completely unprecedented. Imagine if we funded schools and teachers with trillions of dollars. Imagine what we could achieve if we ever gave anything else the same resources we now allocate to " AI": compilers, linters, fuzzers. But those never receive the same funding.

“AI” is completely underwhelming for all the money it has burned. It’s a dead end.

And it’s all just kept alive by accounting fraud and hype because it makes a few sociopaths unimagineably rich.

If “AI” is good at one thing, then it’s producing security vulnerabilities and getting hacked.

1 Like