HNHacker News
TopNewBestAskShowJobs

GolDDranks

3,917 karma · joined April 13, 2015

[ my public key: https://keybase.io/golddranks; my proof: https://keybase.io/golddranks/sigs/_muJdrzbFuPzV0PR4d_Vl6gZe03xTPAGwnPRaUJuFkA ]

GitLab: https://gitlab.com/golddranks/ GitHub: https://github.com/golddranks/ Twitter: https://twitter.com/golddranks

meet.hn/city/jp-Tokyo

Socials: - github.com/golddranks

Interests: Programming, AI/ML, Data Science, DevOps, Education, Hacking, Hardware, Open Source, Networking, Philosophy, Science, Web Development, Japanese Language

---

E-mail: pyry.kontio@drasa.eu

submissionscomments
GolDDranks··on Show HN: Lazy Tetris
Yes, legal trolling is unfortunately a thing.
GolDDranks··on Show HN: Lazy Tetris
Which is crazy, because the shapes are just an idea, and you can't copyright ideas. They also couldn't be a trademark infringement (IANAL, though) if you don't use them in your branding.
GolDDranks··on What even is a small language model now?
(To share a recent personal experience about Markov models: I bootstrapped recently a HMM with hand-assigned weights. It was around 15x15 class transitions, 225 weights. That's small. Or rather, microscopic. Then I ran it against real data, and picked up examples of wrong classifications, and made them auxillary training data. Of course, it was not a language model, language model is impossible to fit in such a small space. It was a model of transitions of chapter "types" in novels, where types are something like "Epilogue" , "Prologue", "Chapter 23", "Table of Contents", "Afterword" etc.)
GolDDranks··on What even is a small language model now?
A traditional Markov model trained (rather, just "fitted") on tokens or words is a small language model.
GolDDranks··on Claude 4
Re: Prompting to ask. Thanks, I'll try that. And I'm gonna try version 4 as soon as I can.
GolDDranks··on Claude 4
Two problems:

1) Writing a high-performance memory allocator for a game engine in Rust: https://github.com/golddranks/bang/tree/main/libs/arena/src (Still work in progress, so it's in a bit messy state.) Didn't seem to understand the design I had in mind, and/or the requirements and goes on tangents and starts changing the design. In the end, coded the main code myself and used LLM for writing tests with some success. Had to remove tons of inane comments that didn't provide any explanatory value.

2) Trying to fix a Django ORM expression that generates unoptimal and incorrect SQL. Constantly changes opinion whether something is even possible or supported by Django, apologizes when pointing out mistakes / bugs / hallucinations, but then proceeds to not internalize the implications of the said mistakes.

I used the Zed editor with its recently published agentic features. I tried to prompt it with a chat style discussion, but it often did bigger edits I would have liked, and failed to share a high-level plan in advance, something I often requested.

My biggest frustrations were not coding problems per se, but just general inability to follow instructions and see implications, and lacking the awareness to step back and ask for confirmations or better directions if there are "hold on, something's not right" kind of moments. Also, generally following through with "thanks for pointing that out, you are absolutely right!" even if you are NOT right. That yes-man style seriously erodes trust in the output.

GolDDranks··on Claude 4
I sincerely hope you'll manage to find a job!

What I meant was purely from the capabilities perspective. There's no way a current AI model would outperform an average junior dev in job performance over... let's say, a year to be charitable. Even if they'd outperform junior devs during the first week, no way for a longer period.

However, that doesn't mean that the business people won't try to pre-empt potential savings. Some think that AI is already good enough, and others don't, but they count it to be good enough in the future. Whether that happens remains to be seen, but the effects are already here.

GolDDranks··on Claude 4
After using Claude 3.7 Sonnet for a few weeks, my verdict is that its coding abilities are unimpressive both for unsupervised coding but also for problem solving/debugging if you are expecting accurate results and correct code.

However, as a debugging companion, it's slightly better than a rubber duck, because at least there's some suspension of disbelief so I tend to explain things to it earnestly and because of that, process them better by myself.

That said, it's remarkable and interesting how quickly these models are getting better. Can't say anything about version 4, not having tested it yet, but in a five years time, the things are not looking good for junior developers for sure, and a few years more, for everybody.

GolDDranks··on Rust 1.87.0 and ten years of Rust
Edit time's already over, sorry for the confusing wording: *not only do I enjoy writing Rust
GolDDranks··on Rust 1.87.0 and ten years of Rust
Happy tenth anniversary for Rust! I have been using Rust from 0.11, circa 2014, and I'm so happy to see it as successful as it is today!

These days I can in all truthfulness say that I don't only enjoy writing Rust –something that was true already 10 years back, despite frequent frustrations – but I am also productive and feel that the language provides what it set out to provide. That's something that certainly _wasn't_ true 10 years back. A decade of iterative improvement is nothing to sneeze at.

I hope stability without stagnation will continue in the coming decades. May Rust 1.174 in year 2035 be a thing of beauty!

GolDDranks··on Ok Rust, You Have a Readability Problem
I wonder if the author bothered to test their code (or used AI to generate the examples). The lifetime elision one simply doesn't work. Rust doesn't have an elision rule for two non-self input lifetimes.
GolDDranks··on 10 Years of Stable Rust: An Infrastructure Story
Note that this is written by Graydon Hoare, the original author of Rust. He hasn't taken part in the project for years (I think he left before 1.0), so it feels quite special to see a write-up by him.
GolDDranks··on $20K Bounty Offered for Optimizing Rust Code in Rav1d AV1 Decoder
As a resident of Japan, I thought the exact same thing. (I'm also a citizen of an EU country, which would permit me to participate, but most of my colleagues couldn't.)
GolDDranks··on Zed: High-performance AI Code Editor
I've been recently using Zed, and I'm in love. It's the best editor experience I've ever had. Everything's snappy like notepad.exe, but it's also quite featureful, and quickly becoming even better.
GolDDranks··on Strings Just Got Faster
The thing is, even it being used just in collections can lead to DOS, if the attacker can control the string contents, and selectively choose strings that your hash table stops being a hash table and turns into a linked list.
GolDDranks··on I just want to code (2023)
> Coding is another tool

Maybe to you. For some of us, it's also a form of self-expression, a hobby or a lifestyle.

In my experience coding for money and coding for fun is a very different experience. I know my worth, but I am also free to do whatever I want out of my working hours.

GolDDranks··on Solidjs: Simple and performant reactivity for building user interfaces
I haven't used Solid yet, but it is commonly said to have "fine-grained" reactivity, compared to React, and better performance.

I also think that Svelte, having a compiler, seems like that it would have an edge in theory.

GolDDranks··on Solidjs: Simple and performant reactivity for building user interfaces
At some point, the creator of SolidJS, Ryan Carniato did a stint helping developing Marko.js, a fine-grained reactive / compiled / a bit HTMX-like system, and blogged about that. It's a framework that originated from eBay. I'm now quite interested about the next version of Marko.js, Marko.js 6, and I'm gonna try it for my new project.

I'm interested in fine-graned reactivity so Svelte and SolidJS have been on my radar, but the HTMX-style approach appeals to me.

Here's something to munch on:

https://dev.to/ryansolid/marko-compiling-fine-grained-reacti...

https://dev.to/this-is-learning/marko-for-sites-solid-for-ap...

https://www.youtube.com/watch?v=Z1dd09rxny4

GolDDranks··on Verus: Verified Rust for low-level systems code
Test verify that the code works on specific inputs. Formal verification checks that it works on every input.
GolDDranks··on Verus: Verified Rust for low-level systems code
Via an unreliable, linker-based hack.
GolDDranks··on Verus: Verified Rust for low-level systems code
I think that for a guarantee as central as non-panicking, there ought to be eventually some kind of support in the core language.

(Just throwing ideas here, but there could be `#[never_panic]` for simple cases where the compiler can clearly see that panic is not possible, or error otherwise, and `#[unsafe(never_panic)]` for more involved cases, that could be proven with 3rd party tools or by reasoning by the developer like normal unsafe blocks.)

For more complicated guarantees, it's harder to see if there's enough common ground for these tools to have some kind of common ground.

GolDDranks··on Astronomers confirm the existence of a lone black hole
Nobody really knows for black holes of these sizes, because we don't have a theory of quantum gravity, but I think that answer according to general relativity would that the curvature is steep enough that escaping would require going faster than c, the universal speed limit, so it doesn't happen.
GolDDranks··on Astronomers confirm the existence of a lone black hole
My understanding is that they can't be _ultramicroscopic_, but they can be quite small. According to the Wikipedia:

> A primordial black hole with an initial mass of around 10^12 kg would be completing its evaporation today

And according to this:

https://www.omnicalculator.com/physics/schwarzschild-radius

Mass: 10e12 kg → Schwarzschild radius of 1.5e-14 m which is smaller than a hydrogen atom (5.3e−11 m).

I wonder what would be the mass that'd keep a black hole in thermal equilibrium with the current background radiation...

(Edit. whoops, minus sign in a wrong place made the calculations haywire, fixed now.)

GolDDranks··on Show HN: "Is This Tech Dead?" A snarky autopsy engine for your dead frameworks
Tried Knockout.js, Mithril.js, Marko.js, Rocket (Rust), none to be found.

Also, the site seems to have abysmal performance.

GolDDranks··on Whistleblower details how DOGE may have taken sensitive NLRB data
That's chilling. I always thought HN as a kinda level-headed corner of the internet.
GolDDranks··on Whistleblower details how DOGE may have taken sensitive NLRB data
> Typically, stuff like this gets flagged off the front page within minutes.

Why would that be, because it's too "political" for tech news? Or are there actual DOGE sympathies within the HN population?

GolDDranks··on Zig's new LinkedList API (it's time to learn fieldParentPtr)
I think it matches really well for the Zig ethos of "artisanal, minimal code". More than often, a type is strongly tied to how it is being used in the code base. In that sense, having it be tightly coupled to a data structure isn't much of a problem. The opposite isn't true, and the data structure is independent of the embedding parent data. The implementation of that data structure itself can still be presented as a library, and the "generic" parts carefully tested.
GolDDranks··on Zig's new LinkedList API (it's time to learn fieldParentPtr)
If you write a function that takes a _generic_ linked list as a parameter, you'd have a function that refers to just the linked list subrecord, and does only linked list operations to that and the linked nodes.

If you want to write a function that takes a linked list of a specific type as a parameter, you just take in a value of that type. The linked list is baked in, so you can get to the other nodes, and because the type is known, you can get back to the parent type from the linked nodes with fieldParentPtr. How to do that _safely_? I don't think that Zig embraces any Rust-like safe/unsafe dichotomy, so you don't.

GolDDranks··on Rust to C compiler – 95.9% test pass rate, odd platforms
If the transpilation itself is bug-free, why not? For static guarantees, provided we transpile Rust code that already compiles on a normal Rust compiler, the guarantees are already checked and there, and the dynamic ones such as bounds checking can be implemented runtime in C with no problems.
GolDDranks··on Bilinear interpolation on a quadrilateral using Barycentric coordinates
Okay, so the same idea that I spitballed in a sibling thread:

> Btw. wouldn't it be possible in modern pipelines to remap or "disfigure" the texture when converting to triangles, so that it counters the bias accordingly? Ah, but that bakes in the shape of the quad, so it can't be modified runtime or it will get distorted again, right.

How does that work with animated meshes?

← PreviousPage 4 of 29Next →