SuperV1234

5 ms·

61.
by SuperV1234 6mo ago
> i will reject any contribution that i do not understand Fair. > that developer will have to earn my trust like anyone else What does it take to "earn your trust"? > LLM contributions are non-deterministic, which means t
62.
by SuperV1234 6mo ago
You are still dodging the question -- what is the problem with not being able to determine in advance what the LLM will do? Even then, you can clearly see that the LLM will try its best to follow the instructions. The result might not be 10
63.
by SuperV1234 6mo ago
It seems pretty solid to me. Where is the nonsense?
64.
by SuperV1234 6mo ago
I'm glad you find it entertaining. Now read the actual points made by the AI with an open mind and a critical mindset, instead of dismissing them because they were not written by a human being. The point I'm making is that this po
65.
by SuperV1234 6mo ago
What is the actual problem with the lack of determinism? Why is auto-completion a problem if it's done with AI?
66.
by SuperV1234 6mo ago
My claim that LLMs can benefit the end product and create quality contributions does not imply that the person behind the contribution is less capable/creative/smart than someone who doesn't use LLMs. But it seems that the Zi
67.
by SuperV1234 6mo ago
That's not what the article is about.
68.
by SuperV1234 6mo ago
What point are you trying to make? Be explicit.
69.
by SuperV1234 6mo ago
You could extend that argument to any tool used by the developer, like a linter, sanitizer, the IDE itself, or even auto-completion. Why target LLMs specifically? The more I think about it, the more nonsensical it is. - What if I do everyt
70.
by SuperV1234 6mo ago
No, it doesn't.
71.
by SuperV1234 6mo ago
I don't believe that even the weakened version of the argument works -- it is based on an assumption, not fact. Why would a contributor that uses AI assistance have fewer chances to be trusted? I'm not talking about AI slop, but a
72.
by SuperV1234 6mo ago
https://claude.ai/share/f38ee8a6-56f1-408a-a536-211eb34c7045 I mostly agree with the assessment. IMHO: hard, inflexible rules like these are always deeply rooted in biases and personal convictions, not in facts. The su
73.
by SuperV1234 6mo ago
You shouldn't use what's demonstrated in the article in a random commit in the middle of business logic. But even in a real-world project, using this technique might have value as part of an internal implementation detail of a use
74.
by SuperV1234 6mo ago
It's quite simple: statements such as "In the real world I would think trying to do any of the things discussed in this article should be an automatic commit rejection on any project." really piss me off. I find it ironic tha
75.
by SuperV1234 6mo ago
Many, if not most, useful libraries, languages, and tools have weird hacks and corner cases in them that look esoteric and insane from the outside. The reasonable engineer understands why they exist. The engineer with a hate boner or negati
76.
by SuperV1234 6mo ago
> would only ever be put into a production code base by the most amateur of architecture astronauts, shortly before their startup fails? Yep, definitely failure of understanding. :) > For real though, defend constexpr two-stepping as
77.
by SuperV1234 6mo ago
Literally installed and used Warp for the first time ever, asked the agent "/agent can you open a tab in the terminal for each build folder" and my account got disabled for AI immediately due to a "Terms of Service viola
78.
by SuperV1234 6mo ago
Didn't know you could use C arrays to get static compile time reflection, guess C was really ahead of its time!
79.
by SuperV1234 6mo ago
This particular issue is a niche corner case of C++26 reflection, which -- like reflection in other languages -- is a massively useful feature. In the real world, failing to understand what you're reading and eagerly generalising to th
80.
by SuperV1234 6mo ago
C++ critics are great at analyzing niche corner cases of the language and generalising those to the entirety of its feature set. This does not look like a productive way to get things done.
81.
by SuperV1234 6mo ago
You cannot do anything meaningful with a container of arbitrary types, it's just bad design. If you want to apply the same operation on all of them, then they share some API commonality -- therefore you can use polymorphism or type era
82.
by SuperV1234 6mo ago
If you are interested in a hybrid solution, I've been working on a fork of SFML which uses SDL as the backend, among many other improvements: https://vittorioromeo.com/index/blog/vrsfml.html
83.
by SuperV1234 6mo ago
Incredibly dumb and nonenforceable policy. What matters is human review and correctness. You're never going to be able to prove that a contributor didn't ask an LLM to help them make some changes, or review/optimize changes t
84.
by SuperV1234 6mo ago
Boxed. Killed all my excitement.
85.
by SuperV1234 7mo ago
It depends on what you are instantiating and how often you're doing so. Most people write templates in header files and instantiate them repeatedly in many many TUs. In many cases it's possible to only declare the template in the
86.
by SuperV1234 7mo ago
Update: I had originally used a Docker image with a version of GCC built with assertions enabled, skewing the results. I am sorry for that. Modules are actually not that bad with a proper version of GCC. The checking assertions absolutely c
87.
by SuperV1234 7mo ago
Please check the article again -- I made a mistake in the original measurements (the Docker image I used had GCC compiled in debug mode) and now the (correct) times are ~50% faster across the board. Not free, still need to audit, but much b
88.
by SuperV1234 7mo ago
The data was incorrect due to an oversight on my part (the Docker image I had used compiled GCC with internal assertions enabled). Including <print> is still very heavy, but not as bad as before (from ~840ms to ~508ms)
89.
by SuperV1234 7mo ago
I've updated the article to say "translation unit" the first time "TU" is introduced. The data was also incorrect due to an oversight on my part, and it's now much more accurate and ~50% faster across the board
90.
by SuperV1234 7mo ago
Update & Apology: I've fully updated the article with new benchmarks. A reader pointed out that the GCC 16 Docker container I originally used was built with internal compiler assertions enabled, skewing the data and unfairly penali

More ›