Table of Contents
How This Started
Feature #1: Articles (or the Lack of Them)
The Upside
The Downside
Feature #2: Gender
Englis...
For further actions, you may consider blocking this person and/or reporting abuse
In Polish, we have three grammatical genders, just like in German. Except... they don't match! 😂 And that's a nightmare, because in Polish, "kot" (cat) is masculine, but in German it's "die Katze", so suddenly the cat is a woman! 😂 And there are SO MANY differences like this!
We also don't have articles in Polish at all, which is probably why I keep messing them up in English. They just feel like such an artificial feature to me! 🤣
In Bengali, we also have three grammatical genders: male, female and neutral.
Which means, a cat is just a cat - neither male, nor female. But of course there can be a male cat and a female cat. Just like, human - neither male, nor female. Then there's man and woman.
This is understandable. But if cat is female by default, then what is a female cat? 🤣
Natural languages are ambiguous. That's why I keep telling people, AI will not replace programmers, because an ambiguous non-deterministic language can't replace an unambiguous deterministic language.
They think they'll not need us, until they'll actually need us, then they'll need us 😂😇🥰
I see your point but AI is getting better at covering ambiguity 😄but regardless, it’s still far from fully replacing programmers for sure.
shorturl.at/afo5o
Yeah, genders don't really match across languages, so you can't rely on what you already know, which makes things infinitely harder 😄
As for articles, yes, they can be confusing at first, but once the language becomes second nature, you start to understand their benefits.
My native language is Hindi But wait, I’m from India, where we have 22 officially scheduled languages and hundreds of other languages and dialects! Sometimes it feels like learning one language isn’t enough to survive here.
Hindi has grammatical gender, so even objects are either masculine or feminine. We don’t have articles like “a” and “the” in the same way English does, and we use postpositions instead of prepositions.
Basically, in India, you can travel a few hundred kilometers and suddenly feel like you’ve entered a whole new language universe!
Actually, we also have quite a lot of dialects here and we're a tiny country. So I can only imagine how many dialects are there in India.
I’m from Rajasthan, India, but because of my job, I live in South India, where people speak different languages. Sometimes, I can’t understand what they’re saying. Btw, I think we probably all think in our first language. At least, when I write in English, I still think in Hindi. Haha! 🤣
Actually, huge exposure to English, through films, games, music, also living in the US for a few months for my PhD research and working at an international company for years, actually changed the language I think in. It sounds strange, but I think in English now, probably also because some things are simpler to describe and express in it.
I understand! I’ve done all my studies and interviews in English too. I just meant that I really, really love Hindi! ❤️ I’ve watched many of your videos, and I know your English is good too. Now, I simply want to learn your language! ✨ I absolutely love learning new languages. Wish me luck!
One thing I find fascinating about India is how language can shape the way we see and experience the world. 🇮🇳 The same joke, emotion, or idea can feel completely different depending on the language we use. It makes me wonder whether we think differently because of the languages we speak, or whether language simply changes how we express our thoughts.
Here’s a question I’d love to explore: If you became fluent in a completely different language tomorrow, do you think it would change your personality and the way you see the world, or would you still feel like exactly the same person? 🤔
Thanks, Giorgi, I read your two posts. I'm Italian and now live in Georgia. Beyond the technical quality of your experiment, which I greatly appreciated, your explanation helps me better understand why my knowledge of the Georgian language is so poor after four years. My wife is Georgian, but she speaks Italian fluently, which is perhaps why I was also lazy and unmotivated. I've learned the entire alphabet, I can read, and I can write using Georgian characters, but I don't understand what I'm reading. And then there's phonetics... some sounds are truly difficult to reproduce (or even just perceive) for someone who hasn't experienced them since childhood. Anyway, thank you, it was truly educational. Didi Madloba.
Arafris!😄🙌 I'm glad my article helped a little. Georgian really is an S-tier language when it comes to complexity and I totally understand why you're struggling. It's a completely different world. Phonetically it resembles Italian a bit, since both love ending words with vowels, but beyond that they're nothing alike.
Don't stress too much about learning Georgian, especially since your spouse speaks fluent Italian. Just enjoy the process, and things will click much more naturally that way.
🇭🇺 In Hungarian, much like in Georgian, there are no grammatical genders; however, our system of inflection is highly complex, allowing for the expression of a vast array of concepts with great precision.
For instance, if I use the term párom (my partner), it refers to the person I am with—not specifically as a "wife," but rather as my significant other. This term doesn't inherently specify my gender or that of my partner, Melinda. The root word is pár (pair/partner); something can have a pár even if it isn't a person—for example, in the question "Where is the párja (pair/other half) of my shoe?", I am looking for the other half of the pair. Furthermore, the base word cipő (shoe) becomes cipőm (my shoe) with the addition of the suffix -m; if I were referring to your shoe, it would be cipőd.
This led to a funny moment when I was chatting with Claude and used the word párom (my partner). He replied by saying, "Give my regards to párom [my partner]"—but since párom implies my partner, he should have actually said, "Give my regards to párod [your partner]."
My familiar programming language is the JS that why I am so interested in the pipeline operator, because |> could make the JS/TS/TSX much more elegant language.
is much more natural to read compare to
I know we are so familiar with this revers direction of declaration.
So this is the main reason to I was make a move:
TS + |> to hack JS
This is really insightful, especially for someone like me who doesn't even know how to say "hello" in Hungarian 😄
Simply: "helló" with long o in write. Because we make a distinque between the short standard o or longer ó.
Just like a u -> ú, ö -> ő, ü -> ű,
but a is different than á, also e is different than é.
I almost forget for this week I make a program with gemini to hacktober challenges, in the post I inserted the full prompts in hungarian because I did it that way, so I can tell the used language under slop coding don't cause LLM is less usable. This is a good part of LLM liberation!
Slop Coding: The Vibe Archeologist Strikes Again!
The Tesler’s Law framing is what’ll stick with me — it reframes the whole thing from “which language is simpler” to “where did each one decide to put the complexity.” Georgian drops articles and gender, then quietly hides an entire sentence inside a single verb; the complexity didn’t leave, it just moved where you weren’t looking. And that pattern is everywhere: it’s why “no-code” tools aren’t simpler (complexity moves to configuration), why dynamic typing doesn’t remove type errors (just relocates them to runtime). Yours is the first time I’ve seen it argued through grammar. The AI point is the sharp sting at the end, too — the places where languages disagree on what to encode are exactly where the model has to guess, and every inserted “the” or forced “he/she” is it quietly inventing information the source never had. Brilliant, genuinely original piece.
Thanks for the detailed feedback!
☺️
A Chinese user of pictographs popping in, hahaha🤣
Speaking of Chinese, my favorite animal is Panda by the way.😄
I have zero idea how that language works and it looks like an impossible language to learn😄 it could easily take a lifetime to learn I guess.
Haha it’s tricky, but there’s no grammatical gender or articles — one less headache! 🤣
So it’s kinda like Georgian in that regard. The complexity is moved to somewhere else.
Exactly! Complexity never disappears, it just shifts to another part of the language 😄
Surprisingly, this is exactly the kinda experiment I'm currently working on. I had a theory regarding how LLMs work. Specifically, I wanted to turn postgres into a LLM 😂 (why do I do this to myself...) Anyway, my litmus test for it, is german and english. My idea being that a concept should be deliberately disconnected from the actually representation. Eg. Apple can be an image, a taste, a smell, a brand, a word... In 100s of languages... So context for what is being meant, is interpretive, not fixed. That also means no more duplications, which should reduce the overall size of the knowledge base and make adding knowledge cheaper and easier. My current struggle with it, is specifically now at the compiler side. The concept of man + apple + woman, a man had an apple, then woman has the apple, what happened to the apple. Across english and german, should result in the same logical chain. Note the THE, that's the catch, because that's grammatical reasoning at work, I mentioned the man had AN apple, but the woman has THE apple, implication being it's the same apple. Then it has to be able to write a sentence about it.
This is definitely an interesting project. And I had the same question to myself ("why do I keep doing this to myself...") haha.
Turning Postgres into an LLM? Hmm... interesting.
Think about it... All the things we currently do to run a LLM, postgres already supports... Except postgres is open-ended and you can expand on it. It's natively multi-tenant capable, it has pgvector and when it actually gets to a scale where it'll bog down, VIEWS!!! Turn a view into a MoE, that way it's dynamically sized to how broad the domain is that you're working on, with virtually no overhead
Technically, I guess so. I'm really curious how it'll pan out in practice.
Same, I think this 1 excites me more than my fractal theory for neuroplasticity
This is such a refreshing and brilliant analogy.
The comparison between German grammatical agreement and static typing - complete with the "arbitrary types rolled by dice centuries ago that you can't refactor" is spot-on. Applying Tesler’s Law here makes so much sense: Georgian drops articles and gender tags, but shifts all that complexity right into the verb morphology. It really shows how every human language is basically an evolving API balancing syntax ceremony against runtime disambiguation!
That's exactly right!
Interesting analogy! Natural languages also have their own rules, flexibility, and trade-offs. The comparison between Georgian and English makes you think about how language structure affects communication and interpretation.
That's exactly the poit!
oh I noticed. Language can only be "designed" at its very beginning.
Subsequent optimizations are not very disruptive.
So we so like Programming languages,
It might be assumed that they are easier to redesign.
// c++ excluded
Sometimes even programming languages are super hard to redesign. It causes a lot of chaos and confusion.
I speak 4 different Indian languages, out of which both Marathi and Malayalam work in the same way as Georgian where in they skip the articles. But if you look at the genders then Marathi has 3 genders that affect the verbs and in Malayalam it doesn't. I had not compared any of the languages I speak in this way. Really fascinating.
That's really interesting. BTW, 4 languages are a lot!
“Complexity doesn’t disappear, it moves” is the line that got me. Georgian packing “I love you” into one verb is basically a language doing everything inside one function 😄 Also never thought of der/die/das as types assigned by dice that nobody can refactor, but now I can’t unsee it. Such a fun read!
Fun fact: in Georgian, the single word that means 'feed them that for me.' It packs in who's doing the feeding (you), who benefits (me), who's being fed (them), and what’s being fed (that), and the fact that it's a request.😄
Wow! I am so glad to see you are publishing this as we enable and take AI to all people on earth, we must be having this kind of discussions. This means the AI revolution has to have participation and contribution from the whole world. Thanks! What you have pointed out are truly fascinating to see where English only thinking in programming breaks down as you cross language boundaries.
Thank you! You got the point of this work perfectly. It's up to all of us to contribute to AI's development to make our lives easier later on. But to do it in a quality manner.
In The Mother Tongue, Bill Bryson highlights the argument that creole and pidgin languages can be more expressive and economical than Standard British English.
Bryson notes several reasons why these linguistic offshoots often outshine standard en-UK in day-to-day expression:
Linguists point out that while standard English collapses complex concepts into a few rigid tenses, Caribbean creoles use verbal markers to create highly expressive and precise nuances in time and action.
By separating the tense (when it happened) from the aspect (the nature of the action), these languages convey structural layers that Standard British English (en-UK) can only replicate using lengthy, clunky paragraphs.
Such insightful information, and great points from Bill Bryson. I fully agree.
That said, all those nuanced rules make languages incredibly complex and hard to learn. In programming, we have a great principle called KISS: Keep It Simple, Stupid. I wouldn't fully apply it to English, since I respect and love that language too much to even remotely associate it with "stupid", but it's definitely simpler and more efficient than some of other languages. That's the main reason I love it so much. Sure it can be clunky and ambiguous at times, but no language is perfect. Every one has its strengths and shortcomings.
Awesome insights, thank you so much!💡
Thanks.
When I first read the book I thought: "This is interetsing, but I'll never have an oportunty to talk about this:-D"
I recommend it for anyone who has an interest in this, because it is well written and entertaining, not dry like your 'normal' linguistic texts.
I've already bookmarked it. ✅ Now I'm going to make more time to read it. 😄
Building on torpedo's point — natural languages actually do get redesigned after birth, and the cases where it worked map exactly onto breaking changes in programming languages.
Turkish swapped its entire script (Arabic to Latin) in 1928, within about a year. It worked because a single authority controlled the distribution channel: schools, printing presses, the civil service. Indonesian took a trade pidgin and standardized it into a national language within a generation. Iceland keeps a purism committee coining native terms so borrowed words don't calcify into the grammar. These are deliberate mid-life refactors, and they landed.
The pattern: breaking changes succeed where governance is centralized and the installed base is captive. Python 2 to 3 dragged on for over a decade precisely because nobody could mandate it, while a company with a monorepo can flip an internal language overnight. So I'd refine the claim: a language can only be redesigned by whoever controls its distribution channel. Natural languages rarely have that lever, so they drift; programming languages sometimes do, so they sometimes migrate.
Which is also why the checksum analogy in the article holds: during any redesign, you pay the redundancy tax twice — once for the old channel, once for the new one — until the old one can finally be garbage collected.
ქართყლიიიიეეეეეეეეე!!!
need it
yes ithink so ，but i know this is inmates