HNHacker News
TopNewBestAskShowJobs

smokel

3,958 karma · joined December 20, 2011

Coder by day, artist by night.
submissionscomments
smokel··on CursorBench 3.1
Most interesting things in software engineering are (laughably) subjective.

Just check out any conversation on dynamic vs static typing, talk to a Rust zealot, or ask a backend engineer if microservices were a mistake.

It's unfortunate, and it makes it hard to have proper discussions on these subjects. It would be worthwhile to figure out how we can have more constructive arguments.

smokel··on Pystd, similar-ish functionality with a fraction of the compile time
It is a Python thing, in the sense that it is Python-inspired:

> design-wise copy the Python standard library's APIs whenever possible [1]

[1] https://github.com/jpakkane/pystd

smokel··on Previewing GPT‑5.6 Sol: a next-generation model
This may have been the case one year ago, but with contemporary models such as Opus, I run into this less often.
smokel··on Big AI labs are hiring philosophers
I don't buy the article's title "Why big AI labs are hiring so many philosophers". They probably just hire one or two, and hundreds of software engineers.

I always found it somewhat annoying that a philosophy study would present itself by stating that graduated philosophers have great job opportunities, implying that studying philosophy would not be a bad choice. It just attracts really smart people, and these tend to find a job more easily. This article seems to make the same kind of mistake.

Also, for all we know these imagined herds of philosophers at AI firms are just labelling pictures of dogs.

smokel··on OpenAI unveils its first custom chip, built by Broadcom
> solve our hardest math problems

We're not quite there yet :)

https://en.wikipedia.org/wiki/List_of_unsolved_problems_in_m...

smokel··on VibeThinker: 3B param model that beats Opus 4.5 on reasoning with novel SFT+GRPO
Being able to drive a car properly also depends on having the right exploration-exploitation balance. A three-year-old is likely to explore too much in a situation where mistakes can be dangerous.

This requires not only knowledge, but also the control systems that develop with the prefrontal cortex. LLMs don't do much control yet.

smokel··on Project Valhalla, Explained: How a Decade of Work Arrives in JDK 28
18446744073709551616 possible values and you can't spare 1 for null? :)

TIL that Rust has NonZeroU64 which you can combine with Optional to get the required behaviour with only 64 bits per entry. [1]

[1] https://doc.rust-lang.org/std/num/type.NonZeroU64.html

smokel··on GLM 5.2 Is Out
If humans did actual reasoning, then why is this particular discussion failing so hard?
smokel··on AI is code and can't be prompted into being smarter
I don't think that is what the parent said, but I'm afraid my comment was too snarky (apologies), and the audience in this thread is not eager to be changed in their beliefs. Thanks for taking the time to reply though.

LLMs are fed a lot of data, and there are many patterns in there, including reasoning and some logic. Adding a little domain specific data will not immediately learn that domain, but it will also not be limited to only that data in its reasoning.

smokel··on AI is code and can't be prompted into being smarter
What would happen if you also feed it a book or two on logic, and some books on mathematics and physics?
smokel··on GLM 5.2 Is Out
> The path to AGI (Artificial General Intelligence) must never be enclosed by high walls. We have always believed that AGI should be the cornerstone for all of humanity to collaboratively explore the boundaries of intelligence and solve complex challenges, rather than a privilege monopolized by a few rules and subject to revocation at any moment.

This is not obvious to me. If everyone gets access to AGI, but only a few people have the means to do really bad things with it, then what is the difference? Might as well make clear from the start that AGI is a powerful tool (read: weapon), and not a solution (e.g. world peace).

smokel··on Launch HN: General Instinct (YC P26) – Frontier models on edge devices
For those too lazy to watch someone talk on video for ages to make a point:

The link is to a famous YouTuber called PewDiePie and he uses a local LLM to parse his email, to save time with that. They have an autoreply system and get notified about urgent matters.

smokel··on Retro-Tech Parenting
Most kids that grew up during the timeline you described had no interest in computer architecture. The small minority that did care is probably the same size now.

The other 99% who were into yoyo-ing back then are now into TikTok, that's all.

smokel··on Dumbphone 2
Do you swap SIM cards all the time? This seems to be the biggest blocking issue for me.

I tried switching phones once a week, which was heavenly. Might try that again, it requires some discipline.

smokel··on U of T researchers demonstrate AI worm could target any online device
Thanks for pointing that out. I scanned the paper and found that in their main experiments, they use a shared GPU resource and do not copy LLMs to target machines. Apparently they did other experiments in the ablation study where they did copy LLMs.

So it's even worse than I expected. The intended worm can spread through my thermostat, and when it reaches a GPU host, it can spread even harder. Fun times ahead.

smokel··on U of T researchers demonstrate AI worm could target any online device
The academic paper is here: https://arxiv.org/abs/2606.03811

It's not fully described how things work exactly, but apparently it does not transfer entire LLMs as part of the worm. Now that would be interesting :)

smokel··on ChatGPT for Google Sheets exfiltrates workbooks
Well, people used MS-DOS which had basically no security model at all for at least 10 years. Most viruses were benign, but it was almost trivial to simply wipe the entire hard disk. People generally didn't care, and made backups.

Things have become a bit more complicated now that machines are connected all the time, and the risk of infection is no longer limited to physically inserting a floppy disk into a machine.

I suspect that the solution is not so much in trying to make our current systems secure, but to make disconnection more practical.

smokel··on Show HN: Open-source Workspace (mail,docs,spreadsheet,drive) web/iOS
Looks really nice. Do you plan to support this in the future? Are you planning to foster a community of developers around it, or are you thinking about hosting it as a service?

Asking for a friend, who also enjoys building projects with LLMs, but publishing and supporting them not so much.

smokel··on I think Anthropic and OpenAI have found product-market fit
AutoCAD is $175 per user per month [1].

[1] https://www.autodesk.com/products/autocad/buy

smokel··on I think Anthropic and OpenAI have found product-market fit
Does this analysis factor in potential caching of tokens on the server side? It seems that if they organize things well (as a model provider), they can save quite a lot on that. Looking at my Cursor statistics makes it clear that the token calculations are not at all trivial.
smokel··on I think Anthropic and OpenAI have found product-market fit
For coding assistance, I have tried OpenCode with several large open models through OpenRouter. All were fairly bad compared to Claude Opus. Could you provide some hints on how I should be holding these open models so that I might get more value out of them?

I agree with the common trope that open models lag behind by about a year, but something magical happened just around a year ago when the state of the art models became extremely useful. By this reasoning we're about to see open models perform well, but I'm afraid there is more to it than just waiting for another revolution around the sun.

Note, my application is coding assistance. Open models can be great for other purposes.

smokel··on GPT Guesses Between 1 and 100
In order to find out how real humans reply:

Please guess a number between 1 and 100.

smokel··on Wake up! 16b
There are only 2^128 of such demos. How much of those are valid DOS programs? If we narrow it down to ones that generate both video and sound, I guess there are much less, which should motivate more people to try and find one :)
smokel··on Fast Factorial Algorithms
I hoped this would help me solve some more Project Euler [1] problems. Unfortunately, the algorithms given are not explained in detail, so the learning experience is somewhat mediocre. Then again, I have ChatGPT to elucidate them for me.

This article [2] has some interesting details on the swinging factorial function n≀, but I can't seem to find the essay that it references: "Swing, divide and conquer the factorial", 2008.

[1] https://projecteuler.net/

[2] https://oeis.org/A000142/a000142.pdf

smokel··on Show HN: Files.md – Open-source alternative to Obsidian
Not the one you're replying to, but "my" system is an Angular application that makes heavy use of CodeMirror.

One instance for the base editor, and one instance for the currently active cell. The other cells are rendered to HTML through a different code path (no CodeMirror involved there).

For rendering mathematical formulas it uses MathJax.

A global context-based keyboard handler such as the one in VSCode allows for Emacs key bindings.

smokel··on Show HN: Files.md – Open-source alternative to Obsidian
Interesting. I recently "vibe-coded" my personal Obsidian clone, because I want proper Emacs keybindings, and Obsidian does not support those, not even through extensions.

I do not know what to do with my pet project. I'm using it myself, and it has tons of futures that took quite some effort to get right. For example, WYSIWYG table editing is not trivial, and Claude Opus agrees with me, in the sense that it could not manage it (at all) by itself.

Open-sourcing it is an option, but I don't look forward to negative feedback. If anyone else wants Emacs keybindings in Obsidian, I will change my mind :)

smokel··on I don't think AI will make your processes go faster
ChatGPT 5.5 responds:

> What data should I retrieve, and where should I get it from? Please specify at least: ...

And it then goes on to ask just exactly what is necessary, being all constructive about it.

smokel··on Playing Atari ST Music on the Amiga with Zero CPU
There were early multitasking operating systems starting with the 286, but for demos you'd typically use the entire CPU. Part of the magic was that video routines would run at an extremely constant 50 (or 60) Hz, perfectly in sync with the hardware. This, and color bleeding, resulted in a buttery smooth experience, that I still miss.

One particular example of this experience was that you'd use "raster bars" to time the performance of your routines. If your main loop is synchronized with the vertical retrace, then switching the background color after a piece of code would show up in the margins of your screen.

Animations were tuned to move in constant pixel offsets. All the anti-aliasing in the world cannot bring back the true demoscene spirit :)

smokel··on Googlebook
Did it find Girav [1]?

[1] https://www.girav.com/

smokel··on Show HN: TikTok but for scientific papers
People will not flock somewhere unless they sense some potential return on investment. If a website looks like it will disappear in a few months, it does not make sense for a user to invest time and effort into it.

You have to either invest a lot to get a critical mass to join your site, or make it extremely entertaining to be there from the start. Apart from all the criticism, this is what Facebook, Instagram, Twitter, and LinkedIn got right from the start. For their intended audiences, it is either useful or fun to be on their platforms.

I don't see much added value for most arXiv extensions, except for SemanticScholar [1], which might have been lucky being one of the first.

[1] https://www.semanticscholar.org/

← PreviousPage 4 of 34Next →