Top 30 stories from Hacker News. Top 10 include comment highlights. Compiled at 20:00 UTC.
777 points by sohkamyung · 183 comments
width=device-width, initial-scale=1' name='viewport'> <link href=' rel='dns-prefetch'> <meta content=' name='cdn-host'> <link href=' rel='icon' sizes='16x16' type='image/png'> <link href=' rel='icon' sizes='32x32' type='im...
What HN said:
rahen: Before WSL, the best ways to run unmodified Linux binaries inside Windows were CoLinux and flinux. http://www.colinux.org/ https://github.com/wishstudio/flinux flinux essentially had the architecture of WSL1, while CoLinux was more like WSL2 with a Linux kernel side-loaded.
scoopr: So, is it like colinux[0], but for pre-NT windows? Neat! Back when I was still using windows (probably XP era), I used to run colinux, it was kind of amazing, setting up something like LAMP stack on the linux side was a lot easier and then using windows editors for editing made f...
andai: Is this person a wizard? To me, this seems an impossible feat. But I wonder how it seems to people who understand how it works? I'm reminded of this joke: Two mathematicians are talking. One says a theorem is trivial.
saadn92: This being on the front page the same day as "Show HN submissions tripled and now mostly have the same vibe-coded look" is nice to see. One person spends 6 years understanding Win9x internals to run a modern Linux kernel inside it.
727 points by Kaibeezy · 262 comments
What HN said:
adamcharnock: Up until a year ago I was regularly using a Massy Fergusson 135 [0] (Perkins Diesel version), made sometime in the 1970s. It was wonderful! So amazing to drive and use. Clunky and heavy, but you really really felt like you were using a machine.
Hasz: I think this is a reaction to the incredibly locked down ecosystem that most of these mfgs are pushing. However, the tech exists for a reason and is not inherently bad, the issue is the lock-in, the lack of choice and interoperability.
jmward01: I want this for cars but to keep the modern powertrain. So an EV without the tracking/touch screens, etc etc. Or an internal combustion engine car that is just simple and efficient (and again, no tracking).
red-iron-pine: Danielle Smith never met a corporate shill she could say no to I predict 6 months before John Deer gets the Alberta UCP on the line and gets a law passed that bans "unsafe tractors" (or the like)
563 points by kaipereira · 160 comments
What HN said:
LPisGood: I saw this video yesterday and considered posting it, but I wasn’t sure if it was appropriate for HN. This channel has another video where it shows how the clean room lab is created starting from a basic backyard shed, and that was truly astounding.
p0w3n3d: 1999. We will have flying cars 2024. LLMs - there will be robots 2026. How to make your own RAM
apatheticonion: I'm not sure this is what they meant when they said they wanted to bring manufacturing back to the USA lol
the-smug-one: So, I get that we charge the capacitor up, and that it leaks so we must recharge it periodically. I don't get two things: 1. How is the value read? Is it reading the leak? 2. How is recharging done? I guess the leak itself (assuming my guess in 1.
455 points by mfiguiere · 234 comments
Qwen Studio offers comprehensive functionality spanning chatbot, image and video understanding, image generation, document processing, web search integration, tool utilization, and artifacts.
What HN said:
simonw: The pelican is excellent for a 16.8GB quantized local model: https://simonwillison.net/2026/Apr/22/qwen36-27b/ I ran it on an M5 Pro with 128GB of RAM, but it only needs ~20GB of that. I expect it will run OK on a 32GB machine. Performance numbers: Reading: 20 tokens, 0.4s, 54.
anonzzzies: I wish that all announcements of models would show what (consumer) hardware you can run this on today, costs and tok/s.
zkmon: On llama server, the Q4_K_M is giving about 91k context on 24GB, which calculates to about 70MB per 1K context (KV-Cache). I could have gone for Q5 which probably leaves about 30K token space. I think this is pretty impressive.
syntaxing: Been using Qwen 3.6 35B and Gemma 4 26B on my M4 MBP, and while it’s no Opus, it does 95% of what I need which is already crazy since everything runs fully local.
344 points by ingve · 266 comments
Take GitHub to the command line
What HN said:
a2128: Why we collect telemetry ...our team needs visibility into how features are being used in practice. We use this data to prioritize our work and evaluate whether features are meeting real user needs.
CMay: If you have 3 of your developers spending 80% of their time in an area of the codebase that gets no usage and you don't see a path forward that realistically is likely to increase usage, it can be a better use of developer time to focus them elsewhere or even rethink the feature.
ryanshrott: > you're going to have to opt out of a lot more than this one setting The opt-out situation for gh CLI telemetry is actually trickier than it sounds. gh runs in CI/CD pipelines and server environments where you may not want any outbound connections to github.
embedding-shape: Love it when a PR is brief: https://github.com/cli/cli/pull/13254 > Removes the env var that gates telemetry, so it will be on by default.
322 points by xnx · 157 comments
An overview of Google’s eighth generation TPUs, built for the agentic era.
What HN said:
himata4113: I already felt that gemini 3 proved what is possible if you train a model for efficiency. If I had to guess the pro and flash variants are 5x to 10x smaller than opus and gpt-5 class models.
pmb: At this point, when you are doing big AI you basically have to buy it from NVidia or rent it from Google. And Google can design their chips and engine and systems in a whole-datacenter context, centralizing some aspects that are impossible for chip vendors to centralize, so I su...
Keyframe: As others have been capturing news cycle eyes, seems to me Google has been going from strength to strength quietly in the background capturing consumer market share and without much (any?) infrastructure problems considering they're so vertically integrated in AI since day one? A...
WarmWash: Whats interesting to note, as someone who uses Gemini, ChatGPT, and Claude, is that Gemini consistently uses drastically fewer tokens than the other two. It seems like gemini is where it is because it has a much smaller thinking budget.
245 points by marklit · 175 comments
Benchmarks & Tips for Big Data, Hadoop, AWS, Google Cloud, PostgreSQL, Spark, Python & More...
What HN said:
himata4113: Florida and most dry / sunny states having little to no solar panels is pretty damn wild. I know in florida you have janky laws stopping you, but below 10kw it's still relatively easy.
CrzyLngPwd: We're off grid and have 7kw of panels, and 40kwh of 48v lithium batteries, with a generator for backup, which is rarely used since we are frugal with electricity and switch everything off when not in use.
noduerme: What's the big deal with having a whole liquid cooled workstation, and why is it important information for me to know what this dude's hardware is? And seriously, is there something about the rig that is necessary to chew through a dataset with a few million rows?
ragebol: Would be kinda interesting to see a histogram of the azimuths and/or tilt angles. In my native Netherlands I'd guess to see that peaking at ~south at say 15-30 degrees, with some lower peaks at east/west combos. Curious to see what it would be in this dataset.
238 points by hubraumhugo · 184 comments
An attempt to detect AI design patterns in Show HN pages
What HN said:
simonw: I expect most side-projects are being built with AI-assistance now. Side projects are typically time constrained - if AI saves you time, why wouldn't you use it? They're also the ideal place to try out new AI tools that your professional work might not let you experiment with.
sunir: Yes, it's the September That Never Ended again. It's fun to complain about the good ol' days, but I'd rather face the world as it is and find the joy in it. http://www.catb.org/jargon/html/S/September-that-never-ended... https://en.wikipedia.
dematz: Nice list of design patterns, but imo a big unmentioned one is a grid of rounded rects https://correctarity.com/roundedrects (maybe what this post calls "Icon-topped feature card grid." ...that might be the official design pattern term)
skyberrys: Don't people just tell you if something is made by AI? It doesn't seem like something to hide. Look, I made something cool using an AI tool. That's great to hear, the thing I'm interested in is the Something Cool, but I do also want to know how, so I can learn how to build Someth...
209 points by zdw · 53 comments
What HN said:
FelipeCortez: 1x5 can also work if you take advantage of subpixel rendering https://www.msarnoff.org/millitext/
ludocode: The 5x5 is very nice, and the 3x5 isn't bad either. Unfortunately neither of them have all of ASCII. The size is also a bit misleading because you need to add spacing so really they need a 6x6 or 4x6 grid. I'm quite fond of Spleen: https://github.
jh00ker: I always loved the 5x6 Pixel font in this classic 90s PC game: https://covertaction.fandom.com/wiki/Cryptography_(Mini-Game... The extra 1 pixel of height for the text in green, in particular, allowed for some cool "italic" styling, especially for letters like E, D, J, U, V
soperj: If the author sees this. I think the lower case t would benefit from a pixel above the cross, similar to how the lower case k goes up one more pixel. It looks a lot like the capital T with how it is now. It is very well done though. Thanks for sharing.
194 points by alfanick · 45 comments
What HN said:
alexcz: It gets even more interesting if you take into account how the satellites know where they are. Around the world there are fundamental stations: https://en.wikipedia.org/wiki/Fundamental_station I had the opportunity to visit one.
delamon: This blog post is also worth noting: https://ciechanow.ski/gps/
keyle: Always makes me laugh when you get some dimwit that claims the Earth is flat, but then uses Google maps in his car. Magic! GPS are amazing. If you understand how they work, and how they reliably know the time etc.
tqi: The time part reminded me of the old WWVB radio time signals. If/when that goes away, I wonder if it will be cheaper to use a gps chip to make "self setting" clocks, or if everything will just be wifi connected.
AI/ML
Other
Programming
Science
Security
Stories and comments sourced from Hacker News public API. Not affiliated with Y Combinator or Hacker News.