The Ü Programming Language
github.com
github.com
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.
I selected fraktur for a logo of my language because it looks "cool" and is already pretty stylized compared to a boring logo using some modern font. I don't insist to use it forever, maybe I will change it later with some better one, but for this someone else need to draw it properly, I don't have necessary design skills for this.
My nick-name is easy to explain. Many years ago I have enjoyed playing Wolfenstein and Call of Duty and these games have Panzerschreck as one of weapons. I choose this as my nick-name, but I unfortunately misspelled it.
>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.
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.
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.
What exactly do you mean by " assign a class into a passed-in variable"? Please post some code illustrating what you are talking about.
function make_widget(parent& x):
w = new Widget()
x.children.add(w)
If this one is too easy, make function return "w" as well, or make it add a widget to two different lists
They borrowed heavily from Rust here.
> 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.
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
use std;
import std::io;
fn main() {
for i in [1, 2, 3] {
io::println(#fmt("hello %d\n", i));
}
}
say “hello $_” for ^3- Overly boastful language. Nothing is perfect, everything has limitations, etc, etc. If the creator isn't willing to admit that and be open about it then that just means users will hit those issue instead.
- Saying what others are not versus saying what you are. You're talking more about other languages than your own language. Vendors do that as a cheap marketing trick to use the popularity of others to boost themselves.
2) In the examples, the placement of & is incoherent: sometimes it's 'i64&', sometimes it is 'i64 &', pick one (I prefer the first).
3) constexpr is a long name for a 'base' concept. comp/<a special letter> would be better. Also conversion_constructor is huge.
4) In D, their static_if doesn't introduce a scope, Alexei Alexandrescu has made a very insteresting pitch about this.
5) zero_init is nice.
6) I don't like "hidden" overloaded operators, how about #+ for the overloaded variant of +?
7) you support i128 natively but integer literals are limited to 2*64-1?
8) I'm slowly reading "reference checking" ( https://panzerschrek.github.io/U-00DC-Sprache-site/docs/en/r... ) which seems to be an important part of the language, and I still don't know what's the integer overflow behaviour which is quite important..
I advocate Zig's behaviour: signed and unsigned have undefined behaviour in case of overflow for optimised build, trap in debug build and there are modulo operators.
> Exceptions are also one of the greatest C++ sins. They were problematic from the start, several attempts to fix them (like noexcept annotations) were done, but C++ exceptions still remain fundamentally wrong. They violate one of the most significant C++ design principles - don’t pay for what you don’t use. Even so-called zero-cost exception implementations (on happy path) aren’t really zero-cost, since they affect optimizations, because of possible unwinding. Also they introduce additional executable bloat - for unwinding code.
- thread-safety (no race conditions): Swift has full concurrency checks since Swift 6, which prevents race-conditions*.
- compile-time calculations: Macros can permit compile-time compilations. The syntax is not lightweight, that’s for sure, but it is possible. And you can do much more than “just” compile-time compilation.
- references (with auto reference creation and dereferencing): I am not fully sure what this is about, but classes are references…
*Logical race-conditions are still, obviously, possible, but unsafe concurrent access to the same variables are not.
Edit: also the name is ungoogleable.
"Some may call this junk. Me, I call them treasures."
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...
Contributions per week to master, line counts have been
omitted because commit count exceeds 10,000.
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).
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.
My concern is not the number of years, but instead the number of commits. It implies linguistic volatility and a lack of a coherent design and/or vision.
Where is the rationale which explains the voluminous number of commits?
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.
Really? May I suggest to reconsider that? Are you still a fan of World War II Germany?
On a Mac "Ü" is typed "option-u shift-u".
Blessed are the humble, for they shall be humbled.
I can't help but notice your username, and the fraktur font, and of course the umlaut.
Are you aware that the flavouring you've chosen for your user account, the product name, and its logotype are referencing 1930s Germany?
I wonder what the intent could be.
NB non German speakers, "shreck" is fear, and "panzer" is well-known to anyone who has watched nearly any documentary about WWII. The Fraktur typeface is an older Prussian typeface vigorously revived by the National Socialists, who, IIRC, made it mandatory for government documents, because Helvetica was considered woke.
- 'Panzer', of course, is a famous Nazi tank model
- 'Schrek' is from _schrecklish_, terrifying (sp?) in German
- The typeface, Fraktur, has a specific meaning in the National Socialist design language, which is, by about a century, the last time it was regularly used (IIRC.)
So, you tell me. In an Internet awash with encoded symbols, what are these ones telling you?
Use your eyes, your noggin, and your German history texts.
> 'Panzer', of course, is a famous Nazi tank model
Panzer is German for tank, or armor. It‘s not referring to any model specifically, I don’t think, but I‘m not a historian.
Panzerschreck (spelled correctly) is something like „tank scare“.
But indeed, that GitHub page gives off the wrong vibes. Using this project will be tough, at least in Germany and Austria, lest you want to be suspected of Wiederbetätigung.