Hacker News: jerryzhang66

New comment by jerryzhang66 in "Is possible a language easy as py, fast as C, more secure than Rust?"

jerryzhang66 — Fri, 24 Apr 2026 04:42:45 +0000

I feel rust is same complex as cpp..just a better build/packages system, and with so many unsafe c/cpp wrappers as lib...I learned it and try to use it on a small project,finding its difficult to use, as cpp, actually more easy to use. https://gtk-rs.org/gtk4-rs/git/book/g_object_memory_manageme... this is the thing make me think.

Is possible a language easy as py, fast as C, more secure than Rust?

jerryzhang66 — Fri, 24 Apr 2026 03:27:41 +0000

With the AI being so powerful, we should have a better programming language. This language is named cpluz/cz, signifying that it is generally between C and C++, but with some differences. Because its philosophy is fundamentally different from C++, ours is "simple the best".

My initial plan is as follows:

A. Roadmap: Version 0.1 is a subset of C++23. The advantage is that it can coexist with the existing C++ ecosystem, all programs and libraries can be shared.

Version 1.0beta is a hardbreak. At this stage, tools or AI should be available to completely convert version 0.1 and C++23 code into CZ1.0 + independent compiler tools.

B. Language design: Currently considered: 1. No header; use modules; 2. C++20 + Multidimensional Subscript Operator/if consteval; 3. Use the C++23 library.

The overall goal is:

One-page cheat sheet (+half a page of basic ideas, if you have zero coding knowledge); 10 pages of a full book. This means roughly no more than 20 basic syntax rules. This will allow both AI and humans to easily learn and communicate with it, achieving ease use as Python. And put everything else in the library.

Based on C++, it should achieve similar speed, and even more efficient because it's simpler.

Besides the safety of Rust, we also need to add Trust, which means strict code review before sharing (including test code, which exploited by hackers of XZ).

Given the speed of AI development, a complete course should take about 18 months, which isn't an exaggeration?

Comments URL: https://news.ycombinator.com/item?id=47885142

Points: 10

# Comments: 10

New comment by jerryzhang66 in "Ask HN: Build new AI language model is still a good startup idea?"

jerryzhang66 — Fri, 11 Jul 2025 17:44:45 +0000

I had some small business adventures before, some good result, some not, but overall not a serious businessman or startup person. So I really wonder about this question...though about tech part, I kind of believe in my design of the model/system.

The info of the project: toCommon, an experimental and ambitious project aimed at reducing vocabulary complexity and improving the efficiency of large language models (LLMs). Though built upon a series of theoretical hypotheses, this repository lays the foundation for a potential startup venture pushing the boundaries of LLM design, training, and inference.

"Now Model for DenseGroupAttention is ready to train on, at same time, the edit system for Simple-Word Compound Substitution is in develope"

Ask HN: Build new AI language model is still a good startup idea?

jerryzhang66 — Fri, 11 Jul 2025 17:44:45 +0000

Article URL: https://github.com/horsearcher/ResearchPaper

Comments URL: https://news.ycombinator.com/item?id=44535038

Points: 1

# Comments: 1