Top 30 stories from Hacker News. Top 10 include comment highlights. Compiled at 20:10 UTC.
860 points by edent · 390 comments
My client was a utility company, and they had a big problem...
What HN said:
OskarS: As a non-web dev, I have a question about this part: > There was a sad coda; as is the way of contract work, I moved on. I explained what I had built to my replacement, that it always worked even without javascript. He was appalled and said, “but that’s a lot more work for us.
faangguyindia: Most of my apps are now simply HTMX + Go + SQLite. I've found it's enough for most projects. One of my sites is image heavy and serves 10 TB of traffic per month. For this, I use the following setup: 1. S3 (I wanted reliable data storage) 2.
graypegg: I haven't heard much about in a while, but the HTML Triptych proposal [0] is still something I hope to eventually land in browsers. HTML forms speaking to REST endpoints are a good pattern.
arnorhs: This post is good, and it's a great example of taking a problem and solving it with the appropriate tech with the right amount of depth. It really helps to have full domain knowledge of your customers as well.
462 points by raffael_de · 283 comments
Exklusive Einblicke und individuelle Angebote: Erleben Sie mit Mercedes-Benz das Maximum aus digitaler Live-PR. Exclusive insights and individual offers: Experience the maximum of digital live PR with Mercedes-Benz.
What HN said:
miohtama: Mercedes acquired Yasa (UK) couple of years ago and now getting up to the speed in the production. Here is a nice video that explains axial flux motors with a factory visit https://youtu.be/B2Hl4c1iZK0?si=VfDYARyuaPVj1nKm They are so, so, small.
AndrewDucker: It would have been awesome if that article had, at any point, explained what an electric axial flux motor was, and why anyone might want one.
s08148692: Very cool. Good to see more axial flux motors in the wild - will be interesting to see if they become the new standard in future. With smaller material costs the cost to manufacture at scale could actually become lower than radial I expect radial will still dominate for at least...
zackmorris: I want to do an engine swap in my 1980s Toyota pickup (like on Back to the Future) from a 100 hp 22r to a 150-250 hp fuel injected inline 4 or turbodiesel to raise the thermodynamic efficiency from 20-25% to ~40% to nearly double fuel economy.
365 points by eries · 300 comments
What HN said:
lebovic: I worked at Anthropic, and I wouldn't attribute much to the structure itself – so I'm wary of using it as a positive example here. I do attribute a lot to specific people.
nradov: It will be interesting to see how many of those companies remain "incorruptible". Your new book seems a bit like a sequel to Jim Collins's 2001 book Good to Great. Several of the "great" companies including Circuit City, Fannie Mae, and Wells Fargo later ran into serious problems...
mehulashah: Eric - I've worked for NASA, ATT, IBM, HP, Amazon, and Google, not to mention a couple of startups that I started in between. None of them (except the startups, but they were brief) stayed true to their original mission.
ngriffiths: My biggest struggle with this question is that "going bad" sometimes coincides with not just financial incentives, but also more people getting value out of it. For example Spotify gradually shifting from "we make it easy to curate and share playlists" to "we make them for you to...
291 points by levkk · 149 comments
PgDog is a connection pooler, load balancer, and sharding proxy for PostgreSQL. Scale Postgres horizontally without rewriting your application.
What HN said:
BowBun: I really wish they'd acknowledge the prior art and name that they've taken inspiration from - https://github.com/postgresml/pgcat Don't pay a startup for your DB proxy, you should own that layer yourself inside of your infrastructure.
eikenberry: > The reason DBs like Mongo or Dynamo exist is because Postgres has a scaling problem. I've used Postgres at a few places and the #1 problem was always high availability, not scaling.
codegeek: "Why Us" => "I ran Postgres at Instacart, where we scaled the company 5x in April of 2020. The biggest problem we had was making Postgres serve 100,000s of grocery delivery orders per minute" Couldn't be a better why us :)
chrisvenum: I am trying to gain a basic understanding of this: Right now I have a 4TB DB on one large box. Is the idea that using a proxy tool like PGDog I could spin up 8 smaller boxes handling ~500GB each and then one medium box for the proxy? Right now I have a project that has very heavy...
199 points by tonyrice · 114 comments
What HN said:
z2: This all feels like a race where the model companies try to solve doing work locally in a way that doesn't suck, before the major operating systems companies figure out AI integration into their OS that doesn't suck.
nathanyz: The VM itself is for Claude Cowork which does all work within the VM sandbox. That doesn't help answer why they spin it up immediately and don't have a way to disable it though. Just the "why it exists" question.
literatepeople: I didn’t get a screenshot of this, but I just found a really pointed example of Anthropics lack of craft / rush to build. If you open Claude on Windows, and click Dispatch (under cowork) to start that up, it will tell you that you need permissions windows doesn’t have.
tom1337: I won't understand why Cowork isn't simply opt-in. It also installs a ~10GB vm bundle which you cannot remove https://www.reddit.com/r/ClaudeAI/comments/1rlc71n/claude_de...
194 points by meetpateltech · 41 comments
An overview of DiffusionGemma, an exceptionally fast text generation model with up to 4x faster speeds.
What HN said:
vineyardmike: Recently I had switched to OpenCode to try out many of the Non-US-Frontier-Labs models. My unexpected favorite model to use was Mercury (a diffusion model). Not because it was “smart” but because it was stupid fast.
samuelknight: Some of these comments miss the advantage of diffusion. This is will have a big impact on edge devices, such as your phone or the GPU in your computer. An LLM's decoder computes tokens one-at-a-time because attention has to account for each previous token.
simonw: NVIDIA are hosting a free endpoint for this one, details at https://build.nvidia.com/google/diffusiongemma-26b-a4b-it - you have to create an account and (I think) verify a phone number too. (I got it to draw a pelican: https://tools.simonwillison.
beklein: A good visual explanation of how text diffusion models like DiffusionGemma work: https://newsletter.maartengrootendorst.com/p/a-visual-guide-...
176 points by nielz_r · 42 comments
Trending AI research papers with code, datasets, methods, and evaluation leaderboards.
What HN said:
jeffreysmith: I played a somewhat unusual role in this whole story. I was the guy who acquired in the original Papers with Code and managed them after they joined Facebook/Meta. It was super sad to see FB/M abandon the original mission of what PwC was building towards and let the original comm...
nielz_r: Hi, Niels here from the open-source team at Hugging Face. Like many others, I was a huge fan of paperswithcode, a website which allowed to easily find the state-of-the-art (SOTA) across any domain of AI, from computer vision to language models to time-series forecasting.
Sharlin: Shame about the name, it feels better suited to a more general curated repo/hall-of-fame of papers in any field that come with easily rerunnable code to reproduce the paper’s results, or try out different datasets, or similar.
imadr: Is no one tired else of these repetitive, obviously Claude-made webdesigns?
155 points by momentmaker · 56 comments
In 1872 Japan had a single railway line. Today the map carries more than 9,000 stations. Drag through 150 years and watch the network fill in — an interactive data map of every Japanese train station with a recorded opening date.
What HN said:
chamomeal: Super cool but I did get this error while scrolling the timeline on safari/iOS Application error: a client-side exception has occurred while loading jivx.com (see the browser console for more information).
helterskelter: Fun bit of trivia: Japanese train stations each have their own custom stamp, and if you have a piece of paper or a notebook they'll stamp it for you at the ticket booth (this is a fairly common thing to do there).
alternator: Cool idea! Not sure if this was created with LLM help, but I suspect so? Not because the page is buggy (it is, though, crashed on my iPhone), but because they make data visualization so accessible.
kalleboo: Never seen this one before: "[Error] SecurityError: Attempt to use history.replaceState() more than 100 times per 10 seconds"
152 points by kisamoto · 81 comments
SBB Resale ist Ihre Wiederverkaufsplattform für hochwertige Occasion-Materialien.
What HN said:
mplanchard: You can pay Amtrak to haul your train car around[0], so you’ve just got to figure out a way to get the car from Switzerland to the US, and then you can really get around in style. [0]: https://www.amtrak.com/privately-owned-rail-cars
mwexler: I just want the clocks. Mondaine tries, but they aren't the same. That 58.5 second rotation then pause is quite clever. https://en.wikipedia.org/wiki/Swiss_railway_clock
xattt: A similar option is available in North America (1). Very niche, and it’s run by Larry Paikin, 93-year-old father of legendary Canadian journalist Steve Paikin. (1) http://www.locomotives.ca
Svip: If I was filthy rich, I'd buy a plot of land near a railway line (that is at least attached to the main lines), build my own siding, and buy one of DSB's IC3 MUs[0], maybe also an IR4 MU[1], so they can together ride on electrified and non-electrified tracks.
129 points by tvissers · 101 comments
Blue41 helps regulated organizations monitor AI agent behavior, detect manipulation and misuse, and prove that sensitive workflows stay within safe boundaries.
What HN said:
EnglishRobin96: This line really stood out to me. > It may look like ordinary text, but when it is placed into an LLM context window, the model may interpret it as an instruction rather than as data. I feel like as long as this is the case, we'll never have secure LLMs.
nticompass: > There is no single control that solves indirect prompt injection There is, actually. It's called removing the AI agent. Done.
bilekas: Putting AI anywhere near people’s finances without even being asked while being responsible for those finances is some next level negligence imho.
reddalo: Good job AI, after we managed to almost fix SQL injections everywhere, you made them come back!
AI/ML
Business/Tech
Other
Programming
Science
Stories and comments sourced from Hacker News public API. Not affiliated with Y Combinator or Hacker News.