HNHacker News
TopNewBestAskShowJobs

kybernetikos

6,302 karma · joined July 11, 2012

[ my public key: https://keybase.io/kybernetikos; my proof: https://keybase.io/kybernetikos/sigs/yHgJmn4qufQOBXtQjZ-3EsFWRdtec9C5wGhHqz3ETYw ]
submissionscomments
kybernetikos··on It's hard to justify buying a Framework 12
I'm not sure that there's a lot of overlap between the target markets. Most schools near me require windows and require a stylus capable touch screen, so the mac is out of the running immediately.

Even if it weren't, the fact that if you're giving a computer to a teen as their first machine to take to class and use every single day, you really, really, really want to be able to separately repair the screen and the ports.

As always, you're paying a premium for the repairability, but if your teen cracks the screen a single time in three years of carrying it to class every day, then you've already saved money.

kybernetikos··on Various LLM Smells
I think there are some factors beyond just skill too - the kinds of tasks you're giving the AI, and how involved you are in ensuring the output is good (via either extensive planning guidance, extensive review/testing, or a combination).
kybernetikos··on Google's Antigravity Bait and Switch
There are decent alternatives. Personally I use newsblur, and I like having a feed that contains only sources I've chosen.
kybernetikos··on The Letter S, by Donald Knuth (1980) [pdf]
You can still hear the music written for the game 'glitch' if you join a slack audio room with nobody else in it.
kybernetikos··on XS: A programming language. Anywhere, anytime, by anyone
yeah, it's like a list of some of the language features I'm most interested in (probably the only thing really missing from my list would be a form of capabilities- so you can restrict library code to a subset of the capabilities you have- this is because library code is inherently not fully trust worthy, and it's infeasible to thoroughly review all library code).

I think interaction with C code would be important for some of the usecases I'd be interested in, but I wasn't sure if there was anything around that.

It's got an odd feel though - normally languages try to use a small set of deep features to provide many higher level features, but this language feels more like a grab bag than a careful design.

I was also expecting some explanation of how the effects system gets exposed in the type system, but I couldn't find anything in the text about that.

And you're right. A GIL feels very suspect.

Having said all that, the most important part of any programming language is its ecosystem, so more detail on syntax highlighting, refactoring (is there an LSP?) and community (maybe there should be more than just test-package on the official package manager site?) would be welcome.

kybernetikos··on A 0-click exploit chain for the Pixel 10
It's not a solution, it's the problem statement.
kybernetikos··on A 0-click exploit chain for the Pixel 10
"Don't click on links" is not a solution, and it's not something people actually do, it's just something they think they do.

Corporate Security will tell you that it's ok to click links to the payroll system or hr or vanta or the 'secure email service' or jira or github or to docusign or the microsoft office document that a partner company sent you or an amazon delivery notification, but not ok to click links in the phishing email that looks exactly like one of those that they sent you.

It's not possible to tell whether a message giving you a link to something is 'sketchy' or not before clicking the link, and any 'security' that relies on people knowing whether a message is malicious or not by magic is broken in the real world.

kybernetikos··on UK sovereign LLM inference
The obvious approach to me for a country seeking sovereign ai would be to hire as many of the recent qwen core team as possible.
kybernetikos··on PipeDream on the Acorn Archimedes
Most of the hardware had rom with its drivers in it. Meant just about everything was a plug and play experience.

And it is true that bit was fast, but once you'd customised the font and replaced all the system icons and set strongedit as your default editor in your!boot, it could take quite a long time to start up.

kybernetikos··on PipeDream on the Acorn Archimedes
I think fireworkz pro was the next evolution of the concept.
kybernetikos··on A recent experience with ChatGPT 5.5 Pro
Current LLM architecture doesn't learn - and you're right this is a huge piece that normal folks fail to understand, since in many ways, it's the opposite of what years of AI research has been trying to create.

However, I think it's important to remember that LLMs are embedded in larger systems, and those larger systems do learn.

kybernetikos··on Your phone is about to stop being yours
It's hard for me to know for sure, but to me it felt like the fact that none of the firefoxOS phones were targetted as devices developers would themselves want to use as their main phone was a big misstep in strategy.
kybernetikos··on Your phone is about to stop being yours
I've still got a firefox OS phone in a drawer somewhere. I was disappointed it got discontinued like so many other mozilla projects.
kybernetikos··on HEALPix
I used S2 to make https://wherewords.id/

It was very pleasant to work with - I spent by far the majority of the project time on the wordlist.

kybernetikos··on Trump fires NSF's oversight board
I think maybe a better analogy would be "the ladder I'm standing at the top of has some faulty rungs near the bottom. I'll set it on fire."

There are lots of things where tearing the system down and starting from scratch is a bad idea, especially if you do it while depending on it and before you have a replacement.

kybernetikos··on Trump fires NSF's oversight board
All systems have problems. What's your evidence that the system in the US is overall net negative? Pretty much the entire rest of the world would have loved to have a scentific system as good as that in the USA. The research output from both government and business was much more extensive and productive of value than the equivalent systems in europe for example.
kybernetikos··on DeepSeek v4
Maybe I need to spell out the step that connects them - how will those companies afford to keep "iterating forward at the frontier" when they probably have a huge crash in their income coming from competition with good enough, but 1/50th the price cheaper and open models.

Iterating forward at the frontier doesn't seem like a sustainable approach if everyone else can catch up with you in 6 months.

kybernetikos··on DeepSeek v4
If you can get something almost as capable for a fiftieth of the price, in most cases you'll do that. You might still send a few tokens to the more expensive option for the exceptional, difficult cases, but that's maybe 10% of the tokens at most. I don't see how it'll be possible to keep spending what anthropic, openai, google etc are spending if they're only going to see the trickiest 10% of tokens.
kybernetikos··on Even 'uncensored' models can't say what they want
I think 'average' is creating a bad intuition here. In order to accurately predict the next word in a human generated text, you need a model of the big picture of what is being said. You need a model of what is real and what is not real. You need a model of what it's like to be a human. The number of possible texts is enormous which means that it's not like you can say "There are lots of texts that start with the same 50 tokens, I'll average the 51st token that appears in them to work out what I should generate". The subspace of human generated texts in the space of all possible texts is extremely sparse, and 'averaging' isn't the best way to think of the process.
kybernetikos··on Even 'uncensored' models can't say what they want
They are a mathematical function that has been found during a search that was designed to find functions that produce the same output as conscious beings writing meaningful works.
kybernetikos··on Even 'uncensored' models can't say what they want
A human in school might try to minimise the difference between their grades and the best possible grades. If they're a poor student they might start using more advanced vocabulary, sometimes with an inadequate grasp of when it is appropriate.

Because the training process of LLMs is so thoroughly mathematicalised, it feels very different from the world of humans, but in many ways it's just a model of the same kinds of things we're used to.

kybernetikos··on Even 'uncensored' models can't say what they want
Neural networks are universal approximators. The function being approximated in an LLM is the mental process required to write like a human. Thinking of it as an averaging devoid of meaning is not really correct.
kybernetikos··on All phones sold in the EU to have replaceable batteries from 2027
My previous phone was refurbished and was great in all ways except for battery life. I have now bought a new phone that I wouldn't have bought if batteries were replaceable.

Having said that, I do like having waterproof phones, and I expect this rule would make that harder.

kybernetikos··on The Legibility of Serif and Sans Serif Typefaces (2022)
It's surprisingly rare for fonts to be careful to distinguish not just Il1| but also 0O 2Z "'' 5S B8. I typically set my system font to something that does, like Atkinson hyperlegible.
kybernetikos··on Miscellanea: The War in Iran
He wasn't particularly scathing about it - in the article it's presented as a decent solution to a difficult problem, just that in his opinion too much was paid for it - but that being so it should have stayed in place.
kybernetikos··on How I write software with LLMs
There's a lot of cargo culting, but it's inevitable in a situation like this where the truth is model dependent and changing the whole time and people have created companies on the premise they can teach you how to use ai well.
kybernetikos··on Ageless Linux – Software for humans of indeterminate age
I don't want to give the impression that I don't find the whole direction of travel concerning, because I do, but as I understand it, the requirement is that the system administrator assigns ages to the users on their system. That seems pretty reasonable to me, and maybe even like a good idea in some scenarios. As far as I know, we aren't talking about software that fights against the interests of the system owner - that's the admin. In fact, I think this might be a feature I would even want.
kybernetikos··on MCP is dead; long live MCP
I don't even fully understand what people are suggesting instead. That we use CLI tools for everything? There are lots of things I do and tools I use that cli would be very inefficient for interacting with.
kybernetikos··on MCP is dead; long live MCP
I assume that this is dependent on app, and it's quite possible that your approach is best in some cases.

In my case I started with something somewhat like Playwright, and claude had a habit of interacting with the app more directly than a user would be able to and so not spotting problems because of it. Forcing it to interact by pressing keys rather than delving into the dom or executing random javascript helped. In particular I wanted to be able to chat with it as it tried things interactively. This is more to help with manual tests or exploratory testing rather than classic automated testing.

My current app is a desktop app, so playwright isn't as applicable.

kybernetikos··on Elon Musk pushes out more xAI founders as AI coding effort falters
I understand that in earth based data centers, 30-40% of the power is spent on cooling. That's in facilities that can cool using conduction and convection to the outside environment.

I don't have any experience in this area, but it seems like for every square meter of solar panel you need about half that in radiator area. And depending on your orbit, these are probably not static things just sitting there, they need to be orientated correctly to work and their correct orientations will change over time.

The worry for me is the level of human maintenance required. The ISS has probably the biggest solar array around, and they send humans out to perform maintainance and repair on it multiple times a year. A decent size data center would need an order of magnitude more solar and radiators than the ISS, and so presumably would need even more maintenance.

← PreviousPage 2 of 34Next →