minc — A minimal language for building native software
1 points by tumdum
1 points by tumdum
Some of the facts (closed-source yet a stub GitHub, landing page has pricing, released yesterday, AGENTS.md in the download's root directory) and a somewhat vague feeling from the marketing copy (lots of em-dashes, lots of "no X, Y, Z") are pinging my LLM radar.
This commentary in the bundled translation of zlib cites an RFC section with §, which is (IME) rare in human-authored code but frequently used by LLMs:
// ---------------- fixed Huffman codes (RFC 1951 §3.2.6) ----------------
// Codes are packed MSB-first on the wire. We precompute each code already
// bit-reversed so emission is a single bw_bits call (which is LSB-first).
The marketing copy has some claims that are independently reasonable but in combination are clearly so outrageous that a human would feel compelled to justify them, such as:
The entire compiler self-compiles to fit on a double-sided 3½" floppy disk.
[...]
On most well-formed code the generated binary will have similar performance to MSVC/Clang.
yet I see no benchmark results anywhere on the site. There are benchmark sources in the download but I'm not brave enough to run this binary as-is.
I don't understand what's the point of this.
To be clear, I have absolutely nothing against making new languages, whether for fun or for experimentation, or for some obscure use case. (I'd love to do that myself!) If this was on someone's personal website or a README on github, great. But this has a nice website and docs - this feels serious, feels like a product.
The selling point seems to be: it's smol, nicer than C, and much simpler than Rust or Zig. That's cool, but not super compelling given the obstacles it has to overcome: it's new, unproven, not open source, and there's just one person behind it -- so it should be assumed, until proven otherwise, to bitrot in a few years max.
I don't see how anyone would choose to use this product. If the author cares about people using this project, he should either make it open source, or find a niche where it's perfect, build a real company around it, and sell it. If the author does not care about people using it (i.e. it's just done for fun/learning/personal needs), I don't understand the point of making a polished product-like website.
this feels serious, feels like a product.
You can't estimate projects this way anymore: One hour with Claude and you have a serious and professional looking website for any toy project.
Maybe if this is true, but this kind of projects will have large tail of issues with codegen or parsing. For example like this. https://github.com/SpacesOfPlay/minc-dev/issues/3 and I don’t event start anything serious in fuzzing.
Even after serious fuzzing you cannot assume that everything more or less reasonable issues are fixed
what's the point of this
I think the pricing section explains it: https://minc.dev/#download . It's not like it's a very viable business plan, but it does not hurt to try.
Some notes from glimpsing over the website:
Well the title was exciting, I assumed this was somebody's personal project. I wasn't expecting the overly flashy website and it's not even open source.