Top 29 stories from Hacker News. Top 10 include comment highlights. Compiled at 20:10 UTC.
585 points by bilsbie · 291 comments
What HN said:
petilon: Magnesium supplementation solved my sleep issues. I have seen many doctors, including sleep specialists, regarding insomnia. They all pointed to one source as the reason for the sleep issues: stress. And they all wanted to put me on prescription sleeping pills. I said no to that.
A1kmm: I wonder how much of this is driven by confounding variables they haven't accounted for. They do factor in shift work as a categorical variable, and employment status as a categorical variable not taking into account occupation.
nottorp: I wonder how many of these "surprising" results are markers of other life style decisions. Regular sleeping hours as a marker for a less stressful life? Also ran across this: https://pmc.ncbi.nlm.nih.gov/articles/PMC5105607/ I doubt it's the actual reading.
djoldman: As always with a lot of these: it's not saying causation. You might measure the speed of your car by putting your hand out of the window and notice that the wind force on your hand is strong when the car goes fast.
229 points by ramon156 · 166 comments
and why communication is so important
What HN said:
smugglerFlynn: > My goals for the end of 2027 are as follows: > Stop making stupid mistakes. I want to be able to finish a task fully without missing or skipping a step. One way to do this is to make a plan for everything you do, and only do that thing. Nothing else.
donkers: Learn to be your own best friend. My DBT therapist is making me work through this Mindful Self-Compassion Workbook [0]. It feels hokey at the beginning, but it seems to actually be working.
t43562: Work forces you to see your own motivations and character and you have to manage yourself like you might manage a very valuable employee that you cannot afford to lose. This means you need to see your strengths and understand how you are motivated and try to come up with ways of...
214 points by vimarsh6739 · 63 comments
Our first open-weights model: multimodal, Mixture-of-Experts, with controllable reasoning effort. Available to fine-tune on Tinker.
What HN said:
segmondy: Very nice, multi modal, largest open weight model that supports audio. Would be interesting to see how good the audio capability is. If you want to run locally, checkout https://github.com/danielhanchen/llama.cpp/tree/add-inkling https://unsloth.
ls_stats: America needs its own DeepSeek or Z.ai, a lot of people (myself included) root for open chinese models to win because they have no other choice. Thinking Machines might be it.
ianbutler: It's nice to see a strong long context open weights model that is multi-modal. There are many applications that will benefit from the strength in audio here and until z.ai and co work in visual this could be very strong for general agentic applications, though I see there's a bit...
minraws: For a first model, and given it's open, I am gaining some faith in American Open research labs again... I couldn't test it since it's not on openrouter or something, but even if it's only as good as GLM5.1 that's more than good enough first attempt, I think.
208 points by davidbarker · 174 comments
What HN said:
geraneum: If you’re puzzled as to why this exists, imagine that, out of the goodness of your heart, you donate $230 to OpenAI to support their mission of rear ending the singularity, and receive Codex Micro memorabilia as a token of appreciation.
landr0id: If anyone is looking at this thinking it looks pretty and wants to check out Work Louder's keyboards, let me save you the time. Their keyboards must be made by designers who do not type much because they are both not pleasant to type on and not very high-quality.
snowstormsun: Is it April 1st already?
kevinsync: Wouldn't surprise me if the real purpose of this is to get a physical object on your desk that makes you constantly think about Codex -- either babysitting your currently-running agents when it's lit up and running, or subconsciously bullying / shaming you into using Codex if you...
198 points by theanonymousone · 85 comments
Telegram claims to have 5 data centers (DCs), referred to as DC1~5 in Telegram’s code and documentation. Among them, DC1 and DC3 are located in Miami, USA; DC2 and DC4 are in Amsterdam, Netherlands; and DC5 is in Singapore.
What HN said:
flexagoon: Just like DC5 is often down to the discontent of Chinese users, DC2 is the one serving all Russian and Ukrainian users, so in the more technical Russian-speaking communities "dc2 down" is also a pretty common saying
Peanuts99: The more I learn about telegram, the sketchier it seems.
SpaceL10n: For a hot second there I was really excited to learn about historical telegram "data centers". It's a capital T.
glaslong: The DC3 gap is interesting. I wonder whether they deprecated it because the other EU server had plenty of capacity, or still keep it but only for... "special" account data flow.
180 points by hermanzegerman · 119 comments
What HN said:
jameson: > The EUIPO found that the word "open" would be understood by the relevant public as meaning freely accessible, while the combination with "AI" (artificial intelligence) would be interpreted as referring to products based on openly accessible artificial intelligence.
zkmon: Someone finally asks some sensible questions, about hijacking of the term "open".
summarity: Key difference between the trademark systems here: in the EU system you don’t get a trademark by trading with a specific name and it then being recognized. It’s the other way around: the name must be unique, not confusing, and highly specific.
mrtnmcc: We had a similar result when a big U.S. defense company (Kratos) tried to take our open source project's domain name: open.space The panel ruled in our favor, that their OPENSPACE trademark is probably invalid because it is descriptive. https://domainnamewire.
177 points by rvz · 88 comments
What HN said:
nickjj: I'm not sure I like this idea. Braintree is a legit competitor to Stripe. I'm guessing they have some informal agreement to keep transaction fees about the same but if they become 1 company, what's to stop Stripe from raising fees even more? I'm not a fan of either to be honest,...
elevation: I typically use paypal for paying contractors, but try to "load balance" across a couple competitors because you never know when one of them will flag your account, nor for how long it will be flagged.
charlieyu1: PayPal user for 20+ years and it is time to end that dross
benmorris: I'm not sure how this is a good thing. There are already very limited options to accept payments online, consolidating two of the major ones is not a great thing.
164 points by neomindryan · 92 comments
A storage appliance that shipped before AVX2 existed now runs Google
What HN said:
dwa3592: I have a prediction. By the mid of 2027, we will have >200B MoE models running on basic consumer hardware. I am running Qwen3.6-35B-A3B locally on my 16GB mac with 7-9 tokens/second. Link - https://github.
hagen8: Some ppl don't like to hear it. But I would assume that token costs when using an inference provider are cheaper than electricity of using locally. If we just take into account output token generation for simplicity. With 5tps u get 18k tokens an hour. That would costs around 0.
hparadiz: Here's my report running several different models on a dual Xeon with 256 GB of DDR4 and no GPU. https://gist.github.com/hparadiz/f3596d00a62d8ebb2dadcc46ee5...
throwaway2027: That's quite slow I'm getting 8-12 t/s on a 13 year old CPU. (Speed varies by context size and other settings who knows) https://news.ycombinator.com/item?id=48354801
147 points by soheilpro · 85 comments
Telegram Serverless lets you run backend code for your bot and Mini App directly on Telegram's infrastructure — no servers…
What HN said:
eamag: What are the quotas like execution time, storage etc?
domh: This is cool. I wish Signal had a bot API like telegram's.
imhoguy: Clever idea! Although after reading it briefly I see a need for secrets storage. I've made one Telegram bot hosted on VPS with Docker and cloud LLM. It also interacts with a few other outside services and all credentials are injected via env vars now. Should I push them as `.
raybb: Providing a SQLite db out of the box is a nice touch. I wonder if they're capping it's size in any way.
143 points by dxs · 180 comments
Sharon Brightwell heard her daughter crying down the line, and that was the end of any defence she might have mounted. The voice belong...
What HN said:
saltcured: Sad times are coming for a lot of families and individuals. It isn't just that technology is upending our naive ideas of trust and authenticity. This is, essentially, the broad class of "confused deputy" attacks.
offsign: Sounds like AI is just greasing the wheels of a long established 'grandparent scam'... goes something like this: 1) voice one: young adult calls, sobbing 2) grandparent inquires with a name... "Ben, is that you?" 3) voice one: "Yes grandma, it's me, Ben...
spenvo: This old post of mine may be of interest, where I point out: "After a bit of threat modeling, it becomes apparent that future spearphishing robocalls may not directly con you, but rather “farm” your voice data by asking you benign questions, and use that to train a voice model to...
imoverclocked: So, you answer your phone to the scam and… now they have your voice too. Talking on the phone is now an unmitigated liability.
AI/ML
Other
Programming
Science
Stories and comments sourced from Hacker News public API. Not affiliated with Y Combinator or Hacker News.