HNHacker News
TopNewBestAskShowJobs

ethmarks

325 karma · joined September 3, 2025

Ethan Marks
submissionscomments
ethmarks··on Tesla has to pay historic $243M judgement over Autopilot crash, judge says
I can't speak for Tesla's FSD specifically, but Waymo did a study on the collision rate of their autonomous cars compared to human drivers: https://waymo.com/safety/impact/. They found that Waymos get into about 81% fewer crashes per mile. Compared to a statistical human driver, Waymo prevented around 411 collisions that would have resulted in any injury, and 27 collisions that would have resulted in serious injury or death. It seems like for Waymo specifically, self-driving cars are demonstrably safer than human drivers. Not sure if that generalizes to Tesla FSD, though.
ethmarks··on AI makes you boring
I guess it depends on whether you're only executing the code or if you're submitting it for humans to review. If your use case is so low-stakes that a review isn't required, then vibe coding is much more defensible. But if code quality matters even slightly, such that you need to review the code, then you run into the same problems that you do with AI-generated prose: nobody wants to read what you couldn't be bothered to write.
ethmarks··on SkillsBench: Benchmarking how well agent skills work across diverse tasks
ZIP files are lossless. If you compress, unzip, and recompress a ZIP file hundreds of times, it'll still be the exact same data as when you started.
ethmarks··on Oat – Ultra-lightweight, zero dependency, semantic HTML, CSS, JS UI library
What makes Astro fast isn't just that it's static, it's that it automatically pre-renders and it doesn't ship a runtime. React can render to static as well, it just also ships the React runtime.
ethmarks··on Lena by qntm (2021)
I read the original version a few years ago and read the new version when it came out, and I thought that the name changes were pretty amusing. qntm kept the story as close to the original as possible while still making it a legally distinct work for copyright purposes. It's like those off-brand Froot Loops called "Fruit Spins" that are juuust different enough to not get into trademark issues. Except in Antimemetics' case, the "knockoff" version was made by the creator of the original, which I think is pretty funny.
ethmarks··on Zed editor switching graphics lib from blade to wgpu
A web port is apparently already on their roadmap: https://zed.dev/roadmap#:~:text=Zed%20on%20the%20Web
ethmarks··on ai;dr
Why bother fixing existing problems if you can just create new problems and then fix those? /s
ethmarks··on Voxtral Transcribe 2
If encoding more learned languages and grammars and dictionaries makes the model size bigger, it will also increase latency. Try running a 1B model locally and then try to run a 500B model on the same hardware. You'll notice that latency has rather a lot to do with model size.
ethmarks··on Voxtral Transcribe 2
They've already done the inverse and trimmed non-coding abilities from their language model: https://openai.com/index/introducing-gpt-5-2-codex/. There's already precedent for creating domain-specific models.

I think it's nice to have specialized models for specific tasks that don't try to be generalists. Voxtral Transcript 2 is already extremely impressive, so imagine how much better it could be if it specialized in specific languages rather than cramming 14 languages into one model.

That said, generalist models definitely have their uses. I do want multilingual transcribing models to exist, I just also think that monolingual models could potentially achieve even better results for that specific language.

ethmarks··on Voxtral Transcribe 2
But it could potentially be even more efficient if it was single-language.
ethmarks··on AGENTS.md outperforms skills in our agent evals
Minification is still a form of compression, it just leaves the file more readable than more powerful compression methods (such as ZIP archives).
ethmarks··on Claude's new constitution
It depends on what you mean by "valid". If a criticism is correct, then it is "valid" in the technical sense, regardless of whether or not a counter-proposal was provided. But condemning one solution while failing to consider any others is a form of fallacious reasoning, called the Nirvana Fallacy: using the fact that a solution isn't perfect (because valid criticisms exist) to try to conclude that it's a bad solution.

In this case, the top-level commenter didn't consider how moral absolutes could be practically implemented in Claude, they just listed flaws in moral relativism. Believe it or not, moral philosophy is not a trivial field, and there is never a "perfect" solution. There will always be valid criticisms, so you have to fairly consider whether the alternatives would be any better.

In my opinion, having Anthropic unilaterally decide on a list of absolute morals that they force Claude to adhere to and get to impose on all of their users sounds far worse than having Claude be a moral realist. There is no list of absolute morals that everybody agrees to (yes, even obvious ones like "don't torture people". If people didn't disagree about these, they would never have occurred throughout history), so any list of absolute morals will necessarily involve imposing them on other people who disagree with them, which isn't something I personally think that we should strive for.

ethmarks··on Briar keeps Iran connected via Bluetooth and Wi-Fi when the internet goes dark
Insider trading on prediction markets is the whole point. They don't exist to provide a fair platform for normal people to make money, they exist to create accurate predictions by providing a monetary incentive for people to be correct. Whether "correct" means that you were just lucky, that you had insider knowledge, or that you were able to influence the result, is irrelevant.
ethmarks··on Meta announces nuclear energy projects
What alternative do you propose that's more efficient?
ethmarks··on Meta announces nuclear energy projects
Accidents, mainly. Solar panels and wind turbines produce far less energy per module than nuclear, so you have to build much more of them. If you build enough of something, the odds that everything goes perfect every single time are quite low.
ethmarks··on Meta announces nuclear energy projects
Thanks for the reply! I think you're arguing with the wrong person in the second half, though. I agree that renewables could potentially be more economically viable than nuclear power[1]. My reply was disputing the "people can die from nuclear therefore we should never use nuclear" argument, not arguing about economic viability. Also I think that broadly claiming that your opposition is "completely irrational" is not a very tactical rhetorical move.

[1]: although since you're basing your claims on the speculative future state of solar technology 10 years in the future, I don't see why the same shouldn't apply to the speculative future state of nuclear power, but that's besides the point

ethmarks··on Meta announces nuclear energy projects
> Nuclear is expensive even after the reactor is build.

Solar panels and wind turbines need maintenance too. And they have much shorter operational lives than nuclear power plants, meaning they'll need to be expensively replaced much more frequently.

> And I wouldn’t call it progress to still rely on steam machines for energy

Could you please explain your objection to steam-based power? Is it purely aesthetic, or is there some inherent downside to steam turbines that I'm not aware of? Also, concentrated solar power systems that concentrate sunlight and use it to boil steam[1] are significantly more efficient than direct photovoltaics.

[1]: https://en.wikipedia.org/wiki/Concentrated_solar_power

ethmarks··on Meta announces nuclear energy projects
Nothing is ever perfectly safe and a lack of perfect absolute safety is not a valid objection. All sources of power have associated risks, even renewables. Wind power has 0.04 deaths per terawatt hour and solar has 0.02 [1]. Nuclear power has 0.03 deaths per terawatt hour (safer than wind), and it's worth noting that almost all of those are from Chernobyl, which was considered unsafe even at the time (they knew about the positive void coefficient). I'm not arguing that nuclear power is perfect, mainly because it isn't. But it's not like all other sources of power are idyllic havens of safety. There are always tradeoffs.

[1]: https://ourworldindata.org/safest-sources-of-energy

ethmarks··on Dell admits consumers don't care about AI PCs
When had Microsoft done this before? For Cortana maybe? I can't recall them ever mandating dedicated physical keys for anything other than the Windows key, but that was over 30 years ago and I assume that's not what you're talking about.
ethmarks··on Ask HN: What are you working on? (Jan 2026)
FYI, your site's admin login panel can be bypassed by deleting the login modal from the DOM. Also, there doesn't seem to be any authentication for your admin tools (like triggering a scrape).
ethmarks··on Apple releases open-source model that instantly turns 2D photos into 3D views
What would your definition of "instantly" be? I would argue that, compared to taking minutes or hours, taking less than a second is fast enough to be considered "instant" in the colloquial definition. I'll concede that it's not "instant" in the literal definition, but nothing is (because of the principle of locality).
ethmarks··on How I protect my Forgejo instance from AI web crawlers
Maybe a stupid question but how can Cloudflare detect what portion of traffic is coming from LLM agents? Do agents identify themselves when they make requests? Are you just assuming that all playwright traffic originated from an agent?
ethmarks··on GLM-4.7
Huggingface[0] says that this is a 358B model. What kind of hardware is necessary to run it?

[0]: https://huggingface.co/zai-org/GLM-4.7

ethmarks··on A guide to local coding models
This is like if an article titled "A guide to growing your own food instead of buying produce" explained that the author was using a four-acre plot of farmland but suggested that that reader could also use a potted plant instead. Absolutely baffling.
ethmarks··on Show HN: HN Wrapped 2025 - an LLM reviews your year on HN
https://hn-wrapped.kadoa.com/ethmarks

Mine seems to think that I'm some kind of detail-obsessed super-pedant. Personally, I think this is ridiculous. "super" is a Latin stem meaning "beyond", which implies that I've transcended the qualities of pedantry. A better term would be 'pluri-pedant', which denotes someone who is exceptionally punctilious while still remaining within the bounds of being pedantic.

ethmarks··on Show HN: HN Wrapped 2025 - an LLM reviews your year on HN
It would have made more sense if it was "your upvoted submissions in 2035"
ethmarks··on Claude in Chrome
No it isn't. At least not on my devices. Try opening gemini.google.com in an incognito window.
ethmarks··on Claude in Chrome
Security risks aside, that's pretty remarkable problem solving on Claude's part. Rather than hallucinating an answer or just giving up, it found a solution by creatively exercising its tools. This kind of stuff was absolute sci-fi a few years ago.
ethmarks··on CSS Grid Lanes
Kind of random but why, in the linked repo, are you using dotnet core for minifying a Javascript file? I'm just curious. It seems like overkill to me.
ethmarks··on CSS Grid Lanes
Pointing out that masonry isn't as good with mixed-orientation content as it is with uniform-orientation content is all well and good, but we still need a way to display mixed orientation content. What alternatives to masonry do you propose?

- If you stretch all images into a uniform aspect ratio, they get all squashed and look terrible.

- If you crop all images into a uniform aspect ratio, you lose potentially the majority of the content in some images.

- If you display all images at their natural aspect ratio and their full size, there will be huge swathes of empty space in between them because they don't pack tightly.

Masonry layouts allow you to preserve aspect ratio without wasting a massive portion of your user's screen space. It's not perfect, but it's the best layout mixed-orientation content that I know of.

If you know of a better method to handle mixed orientations, I'd love to hear it and would gladly rescind by remarks.

← PreviousPage 2 of 6Next →