Hacker News Reader

smokelon The end of Microsoft Research's AI Frontiers group

This refers specifically to "AI Frontiers", a research group within Microsoft Research that focuses on fundamental advances in AI. There is no end in sight to real AI frontiers.

Perhaps "The End of Microsoft Research's AI Frontiers Group" would be a less ambiguous title.

smokelon Show HN: HackerTV – HN's most-recommended Youtube videos as 24/7 TV channels

I can highly recommend the board game, not the programming language, Go.

smokelon Show HN: HackerTV – HN's most-recommended Youtube videos as 24/7 TV channels

Is someone working on HackerTldr? Three sentences that describe the target page would save me a lot of time.

Edit: HackerTV looks great, well worth the click :)

smokelon Show HN: Subvocal – Talk to your computer without making noise

Are there any health risks with this method?

Edit: mostly concerned about RF near my eyes for 8 hours a day.

smokelon Show HN: HackerBooks – HackerTube, but for books mentioned on HN

Are you referring to the "Top all-time" list? That list seems to be mostly filled with popular books that do not go into too much detail, which is required for a decent treatment of probability and statistics.

There is a separate topic "Data & Statistics" that has quite some relevant titles.

Also, I think the interest in probability theory has only recently grown with the advent of generative AI. I'm expecting more interest in the future.

Edit: also, the site seems to miss some nice gems, most likely because they are not traditionally published books: "Probabilistic Artificial Intelligence" is freely available on Arxiv, and was discussed on Hacker News in March 2025 [1].

[1] https://news.ycombinator.com/item?id=43318624

smokelon HackerTube: YouTube built from videos recommended by HN

I suspect they know perfectly well when they can't make money from someone, so they deliberately serve up boring crap to save on hosting costs.

smokelon Genie 4

Simplicity is only relevant to humans (and other animals that evolved into trying to forecast their survival). Most of the universe seems to prefer chaos. So no, from a universal perspective, there is no reason for reality to be simple or elegant.

smokelon JetBrains Thinkrail

It seems that Air is a strategic product, and Thinkrail is more of an experiment.

While they are both "agentic development tools", it seems that Air is mostly concerned with managing multiple agents, whereas Thinkrail is about spec-first development.

smokelon Today, we hacked what we'd set out to hack: the iPhone

> Vote with your wallet

If that would work, I'd be writing this on my iPhone Mini.

smokelon OpenShell – A safe, private runtime for fleets of autonomous AI agents

Well, a lot of people haven't been running NodeJS libraries in containers for ages, which seems even more risky.

smokelon Ask HN: How to deal with shipping AI-written code you don't understand?

> Hacker News isn't interested

I wouldn't be too sure about that :)

smokelon Neural networks are clean

Isn't it ironic that an article on AI explainability is itself so hard to understand?

smokelon Chapter 1 of "Ways of Seeing" by John Berger (1972)

Interestingly, this is not copyright infringement, or at least not for the text being reproduced here. The original book (1972) mentions "The text of this book is not copyright" [1]. The image reproductions could invite copyright trolls though, so be warned.

[1] https://pengpedia.org/edition/ways-of-seeing-john-berger-471...

smokelon Classic PC demoscene productions running natively in the browser

Second Reality ran comfortably with only 640KB RAM using the Gravis Ultrasound.

But it did require a 386, which typically had 2 MB or more of RAM back then.

smokelon Classic PC demoscene productions running natively in the browser

I gave up once full screen video playback was possible from CD-ROM. After that, doing (typically worse) non-interactive animations in real-time, seemed a bit pointless.

But watching Crystal Dreams II on my phone actually made me feel young again :)

smokelon Mistral Large 4

What was your prompt? Most of these seem to be related to metaphors for "ideas" or thinking, or having a bright moment.

"Zephyr" and "breeze" might be related to forgetting everything, starting fresh.

So by this way of naive reverse engineering I would imagine your prompt to be "Forget everything and think about a random word". That would prime the LLM to come up with these?

smokelon Our approach to EU text provenance rules

Why use watermarking, and not simply add a signature?

smokelon Agents don't need memory, they need documentation

Unfortunately, that is not how many software development projects work. A customer may request certain things, or the software might be part of a larger system.

Consider working on software for a coffee machine. Why is pin 42 (GRIND) activated every now and then? Would you really want to document that in Git commit messages?

smokelon Agents don't need memory, they need documentation

Code typically documents the "what" and "how", not the "why".

Why something exists, and how it connects to the outside world may be documented in comments, but more often than not it isn't.

smokelon FTL: A new operating system for clouds

A problem with this approach is that it would put a large burden on the application developer (or development system) to support other devices (or services) than initially planned.

Of course, it would be possible to add new drivers only when necessary, but that would also allow for security problems.

So, in theory it might work, but in practice it would require quite a bit of thought.

smokelon Kolibri: A Sovereign Open-Weight Model

Would you mind substantiating these claims a bit?

smokelon With most information hidden, the game Stratego had stumped AI until now

This approach also works for Hanabi, which is a very interesting game. You can't see your own cards, but the other players can. I bought the game because someone on a reinforcement learning podcast [2] mentioned it, and actually played it multiple times.

[1] https://en.wikipedia.org/wiki/Hanabi_(card_game)

[2] https://www.talkrl.com/episodes/jakob-foerster

smokelon With most information hidden, the game Stratego had stumped AI until now

This puts the earlier "Mastering the Game of Stratego with Model-Free Multiagent Reinforcement Learning", 2022 [1] in some perspective. Apparently the "mastering" in 2022 wasn't quite there yet. Four years later, the new approach seems to actually be better than humans.

[1] https://arxiv.org/abs/2206.15378

smokelon Dutch computer museums (2022)

"SAMENGEST TEKEN" is most likely an abbreviation for "SAMENGESTELD TEKEN". I guess it's a translation of "COMPOSE KEY" and they ran out of rub-on letters somewhere.

smokelon Pi 1.0

Why do these tools have such bugs? With AI it should be trivial to fix, no?

Or is it a non-trivial bug that requires a lot of refactoring, and could introduce a lot of new bugs? That would require careful review from a human.

The latter may well be a reason why I don't see extreme productivity gains in larger brown-field projects.

(Disclaimer: I see enormous benefits in one-off greenfield projects.)

smokelon Pi 1.0

Full screen hides distractions from other applications. Seems reasonable as something minimal.

Minimalism is a difficult subject. In art, minimalists tried to strive for something that is universally minimal. But if you look in nature for straight lines or perfect circles, you end up disappointed. Turns out minimalism found things that were minimal with respect to how some humans think about minimalism. For all we know, pure chaos may be more universally minimal than an empty vacuum.

smokelon MicroLLM Lab – Try 7 tiny LLM's in the browser

No, that is not what SLMs are useful for. They have very little knowledge, and typically lack reasoning skills.

Useful applications include sentiment analysis, text classification, entity extraction, etc.

They certainly can be useful, but you shouldn't compare them with LLMs such as Opus or Fable.

smokelon Show HN: A CC0 museum of retro 3D tricks you can paste into a page

The plasma effect, for example, uses vertex shaders to calculate in the order of 8 sines per pixel, and then adds some squares and square roots.

Originally, for a plasma effect, one would calculate a sine based value only for each row and each column. For a 320x240 pixel effect, only 320+240 "complicated" calculations were necessary per frame. To combine these at the pixel level, simple adds and bit shifts were enough. This was necessary, because it was entirely impossible to do so many calculations (or even memory lookups) per pixel on old hardware.

The actual effect is nice, but it has little to do with the old demoscene effect :)

smokelon Show HN: Building a Markdown editor for Mac, iOS and web

This question is somewhat more general than this specific application, but how can an app guarantee that files will remain local, also in future updates of the app?

To put it simply, an automatic update of the app might start uploading the notes to a server. I don't think there is any security boundary in iOS, Android, or any other platform preventing this, right?

Other than that, congratulations on what seems to be a nice app!

smokelon There are no "rogue" AI agents

That seems like a divide and conquer attack. The actual problem is that the electorate system leads to a two-party equilibrium.