> Ü is heavily inspired by C++, but doesn't have its downsides. Also it was influenced by Rust, but only slightly and thus is way easier to use in comparison to Rust. Any possible coincidence with design and features of other programming languages is unintentional.
Blessed are the humble, for they shall be humbled.
Many here probably want to know if the language actually works as intended and solves problems mentioned in it's README. I can absolutely say yes, it works!
Since November 2025 I develop a project written (almost) fully in Ü. It's a video game. It's not finished yet, but I plan to release a mostly-working version in several months, including making its source code open. So, stay tuned.
I already have ~45000 lines of code in this project. All this is manually-written (no LLMs involved). Writing Ü is nice, and code mostly works if it compiles. No single time I needed finding/fixing memory-related issues (typical for languages like C++). No single time I faced a crash in random place, if something crashed, it was a safety check. The standard library covers basic needs. unsafe code is used only to interact with SDL2 and OpenGL functions.
I spend my time mostly developing this game project. But occasionally I fix bugs and make small improvements in the language itself. This explains why there are so little commits in last months compared to time before November 2025.
Just a suggestion... you might want to include an example on your main README.
As someone building a language myself, I'm interested in the other languages actively in development...
But you start with an info dump, no examples, and then a table of features - where the first feature is not something anyone would pick a language for.
You claim to be a memory safe language... And those are buried in the middle of the list. You want to highlight that, and say how you accomplish it. You say you have no GC, but no mention of Affine Ownership or Ref Counting. You talk about thread safety, but no mention of how.
You need to show WHY anyone should care about your language, what problem it's solving, and what that looks like as fast as possible.
In your comparison table, you leave out Go and include Odin - that seems like a mistake. Go punches FAR above its weight class. Dismissing it because it "comes with a heavy runtime" is likely to get your project dismissed, no offense. Odin is - essentially - experimental.
People's attention is fleeting.
Everyone and their mother is building a language or two...
Some things I want to know right away:
1) what stage are you at (honestly, not wishfully)?
2) what problems have you ACTUALLY solved instead of INTEND to solve at some point in the future?
3) how thorough is your testing, what do you have, how much, what's the coverage by category?
4) this seems like a performance language - I want benchmarks. If you don't have a good concurrency story, you better have something, and you better have convincing benchmarks that it actually works, otherwise - why is anyone from Go or Rust or Zig or Nim or Crystal or Swift or even Java/Kotlin/Scala or C# going to think about switching?
More than 10 years of development, the core aspects of the language are pretty stable and are unlikely to be changed in future. New features can be added, especially in the standard library.
> what problems have you ACTUALLY solved instead of INTEND to solve at some point in the future?
The main problem is memory safety. It's already solved and not planned to be solved. You already can't shoot your leg with typical memory-related errors.
> how thorough is your testing, what do you have, how much, what's the coverage by category?
I have a lot of tests, several thousands of test-cases covering each language feature, including tests for specific compilation errors, tests for compilation into actual binary code, many tests for each standard library feature, tests for the build system. And of course I have a self-hosted compiler, which proves that everything works as intended in actual code.
> this seems like a performance language - I want benchmarks
Nice suggestion, probably it's worth to adding some benchmarks.
> it actually works
It does. There is just nothing there in the language, which can degrade performance significantly. The same LLVM library is used as in C++ or Rust compilers, no GC is involved, runtime safety checks are sparse. In many cases the result binary code is identical for identical C++ or Rust code or at least closely matches it.
My own rough measurements between two compiler generation (first one written in C++ and second one written in Ü) show nearly identical performance.
> why is anyone from Go or Rust or Zig or Nim or Crystal or Swift or even Java/Kotlin/Scala or C# going to think about switching?
Go and Java-VM based languages are garbage-collected, which is problematic in some areas (like video-games). Rust is fine, but is sometimes too complicated. Zig is just a better C with no memory safety. Nim isn't known for me.
The README makes me never want to use this. A career of seeing the same style from shit vendor salespeople has made me assume that under the hood is nothing but snakes.
Whenever you are publishing something, whether it be an app, a programming language, ... the examples or screenshots must come first. The reality is that unless you have made a very compelling case beforehand (like with a well crafted screenshot or example), most people will not read your list or description below. Put the examples at the start of the README, they're the selling point of your language not the list of features
For instance, this was the first ever version of the rust-lang.org website:
I remember reading it 14 years ago and it was pretty effective at captivating users towards the language. Notice its terseness and the example at the bottom: a reader will first look at the example, then the bullet points above if they're interested, and then the rest of the docs
> Ü uses RAII for memory and resources management (no GC is involved), but manual memory management may be still used in unsafe code.
Saying "using RAII for memory management" is insufficient - with just RAII, you cannot even assign a class into a passed-in variable. The language designer _must_ make make more choices to get a useful language - maybe affine types, or linear types, or prohibit many C++-like idioms, or maybe just good-old refcounted shared pointers (but I'd argue this is a form of GC...)
> Ü is memory-safe and race-condition-safe, as long as no unsafe code is involved at all or as long as unsafe code is correctly written.
How is this achieved? The docs mention in passing that there is some sort of thread-safe immutable structs, but it is not really clear what's the overall picture and how they interact with non-trivial code. And the examples have nothing on thread.
Well, if this is supposed to be something like C++, and if "new" is dynamically allocating memory, then simply don't do that - this kind of thing has been solved years ago.
Why does DeepL translate "Panzerschreck" to "Panzerschreck"? Is that actually a word understood by English native speakers? I know the words "Panzer" and "Schreck" (normal word of used in the appropriate context), but didn't exactly know what a Panzerschreck is.
Well, a fitting translation seems to be "Bazooka". However, "Panzerschreck" sounds like an old fashioned German word (not too surprising I guess) that wouldn't be used anymore today. The typo which I think might be intentional makes it kind of silly.
You are partially right. I cherry-picked features which I personally find important. But since they are so important for me, I bothered designing/implementing them in the language in a right way.
It's sad that the compose key isn't prevalent. Combining " and u into ü makes a lot of sense, however people thought that we wanted 40 different slightly different keyboard layouts instead of a universal one. Despite how annoying it might seem to get, it's not that bad really.
On a random day last month there where 17 commits, and commits happen daily by the same author. It’s true dedication.
Browsing through them requires understanding / translating Slovenian. Translation has never been cheaper, just recently the author has switched to English.
When I started development on the language, my English knowledge wasn't that great, so I preferred writing commits in Russian, but using Latin alphabet instead of Cyrillic to avoid changing keyboard layout (I hate doing this).
Something is wrong with that commit count. There are a total of ~4k commits over nine years [1].
It looks like the author revived the project recently with the help of AI, but the majority of commits are from 2017-2024. Github might be counting all the branch activity from agents.
From the description of the graph you kindly provided:
Contributions per week to master, line counts have been
omitted because commit count exceeds 10,000.
So I am confused as to the "total of ~4k commits over nine years" determination, unless this is in reference to commits Panzerschrek has made spanning the last two years (not nine).
I have thoughts about some kind of attributes for class fields and maybe classes. But such change isn't the first in the list of features to be done, there are more important things to do.
Blessed are the humble, for they shall be humbled.
>code examples are linked as a footnote at the bottom of the page (and need to open each code file individually to view them)
This project apparently does not want to succeed.
(Which is odd given the amount of effort invested in it!)
The language itself seems quite nice though.
Many here probably want to know if the language actually works as intended and solves problems mentioned in it's README. I can absolutely say yes, it works!
Since November 2025 I develop a project written (almost) fully in Ü. It's a video game. It's not finished yet, but I plan to release a mostly-working version in several months, including making its source code open. So, stay tuned.
I already have ~45000 lines of code in this project. All this is manually-written (no LLMs involved). Writing Ü is nice, and code mostly works if it compiles. No single time I needed finding/fixing memory-related issues (typical for languages like C++). No single time I faced a crash in random place, if something crashed, it was a safety check. The standard library covers basic needs. unsafe code is used only to interact with SDL2 and OpenGL functions.
I spend my time mostly developing this game project. But occasionally I fix bugs and make small improvements in the language itself. This explains why there are so little commits in last months compared to time before November 2025.
* https://news.ycombinator.com/item?id=45769161
> See examples (https://github.com/Panzerschrek/U-00DC-Sprache/tree/master/s...) to get a general idea what Ü is and how it looks like and feels. Read the documentation (https://panzerschrek.github.io/U-00DC-Sprache-site/docs/en/c...) for more details. If you have some questions about the specific mechanism providing safety, read the corresponding chapter (https://panzerschrek.github.io/U-00DC-Sprache-site/docs/en/r...). If some topic isn't clear for you after reading the documentation, contact me and I can answer your questions.
As someone building a language myself, I'm interested in the other languages actively in development...
But you start with an info dump, no examples, and then a table of features - where the first feature is not something anyone would pick a language for.
You claim to be a memory safe language... And those are buried in the middle of the list. You want to highlight that, and say how you accomplish it. You say you have no GC, but no mention of Affine Ownership or Ref Counting. You talk about thread safety, but no mention of how.
You need to show WHY anyone should care about your language, what problem it's solving, and what that looks like as fast as possible.
In your comparison table, you leave out Go and include Odin - that seems like a mistake. Go punches FAR above its weight class. Dismissing it because it "comes with a heavy runtime" is likely to get your project dismissed, no offense. Odin is - essentially - experimental.
People's attention is fleeting.
Everyone and their mother is building a language or two...
Some things I want to know right away:
1) what stage are you at (honestly, not wishfully)?
2) what problems have you ACTUALLY solved instead of INTEND to solve at some point in the future?
3) how thorough is your testing, what do you have, how much, what's the coverage by category?
4) this seems like a performance language - I want benchmarks. If you don't have a good concurrency story, you better have something, and you better have convincing benchmarks that it actually works, otherwise - why is anyone from Go or Rust or Zig or Nim or Crystal or Swift or even Java/Kotlin/Scala or C# going to think about switching?
> what stage are you at
More than 10 years of development, the core aspects of the language are pretty stable and are unlikely to be changed in future. New features can be added, especially in the standard library.
> what problems have you ACTUALLY solved instead of INTEND to solve at some point in the future?
The main problem is memory safety. It's already solved and not planned to be solved. You already can't shoot your leg with typical memory-related errors.
> how thorough is your testing, what do you have, how much, what's the coverage by category?
I have a lot of tests, several thousands of test-cases covering each language feature, including tests for specific compilation errors, tests for compilation into actual binary code, many tests for each standard library feature, tests for the build system. And of course I have a self-hosted compiler, which proves that everything works as intended in actual code.
> this seems like a performance language - I want benchmarks
Nice suggestion, probably it's worth to adding some benchmarks.
> it actually works
It does. There is just nothing there in the language, which can degrade performance significantly. The same LLVM library is used as in C++ or Rust compilers, no GC is involved, runtime safety checks are sparse. In many cases the result binary code is identical for identical C++ or Rust code or at least closely matches it.
My own rough measurements between two compiler generation (first one written in C++ and second one written in Ü) show nearly identical performance.
> why is anyone from Go or Rust or Zig or Nim or Crystal or Swift or even Java/Kotlin/Scala or C# going to think about switching?
Go and Java-VM based languages are garbage-collected, which is problematic in some areas (like video-games). Rust is fine, but is sometimes too complicated. Zig is just a better C with no memory safety. Nim isn't known for me.
For instance, this was the first ever version of the rust-lang.org website:
https://web.archive.org/web/20111226082307/https://rust-lang...
I remember reading it 14 years ago and it was pretty effective at captivating users towards the language. Notice its terseness and the example at the bottom: a reader will first look at the example, then the bullet points above if they're interested, and then the rest of the docs
Saying "using RAII for memory management" is insufficient - with just RAII, you cannot even assign a class into a passed-in variable. The language designer _must_ make make more choices to get a useful language - maybe affine types, or linear types, or prohibit many C++-like idioms, or maybe just good-old refcounted shared pointers (but I'd argue this is a form of GC...)
> Ü is memory-safe and race-condition-safe, as long as no unsafe code is involved at all or as long as unsafe code is correctly written.
How is this achieved? The docs mention in passing that there is some sort of thread-safe immutable structs, but it is not really clear what's the overall picture and how they interact with non-trivial code. And the examples have nothing on thread.
They borrowed heavily from Rust here.
What exactly do you mean by " assign a class into a passed-in variable"? Please post some code illustrating what you are talking about.
If this one is too easy, make function return "w" as well, or make it add a widget to two different lists
Well, a fitting translation seems to be "Bazooka". However, "Panzerschreck" sounds like an old fashioned German word (not too surprising I guess) that wouldn't be used anymore today. The typo which I think might be intentional makes it kind of silly.
On a Mac "Ü" is typed "option-u shift-u".
But in last several years I prefer committing frequently, basically each time I do a change which compiles successfully.
And it's all manual changes, no LLM (vibe-coding) is involved.
Browsing through them requires understanding / translating Slovenian. Translation has never been cheaper, just recently the author has switched to English.
Example random day: https://github.com/Panzerschrek/U-00DC-Sprache/commits/maste...
When I started development on the language, my English knowledge wasn't that great, so I preferred writing commits in Russian, but using Latin alphabet instead of Cyrillic to avoid changing keyboard layout (I hate doing this).
It looks like the author revived the project recently with the help of AI, but the majority of commits are from 2017-2024. Github might be counting all the branch activity from agents.
[1] https://github.com/Panzerschrek/U-00DC-Sprache/graphs/contri...
"Some may call this junk. Me, I call them treasures."
Edit: also the name is ungoogleable.