Top 27 stories from Hacker News. Top 10 include comment highlights. Compiled at 20:10 UTC.
844 points by nprateem · 464 comments
What HN said:
donavanm: Ive dealt with this error at AWS. It’s a unit error. In my case we meant to charge like 5¢/GB, but missed the unit (GB), and then the billing system defaults to bytes. 5¢ per Byte of data transferred meant some customers were seeing MM bills within hours.
yuchen20: I got 3 consecutive emails warning that my budget crossed its $18 threshold. Opened it up: cost was 78 million. Thought it was a phishing attempt, logged into my actual account, and... still 78 million. EMOTIONAL DAMAGE.
wglass: It's crazy enough this will be fixed soon. Years ago I found an actual hidden error in my bill. (This was early 2010s). The system was calculating the EC2 reservation savings incorrectly for some of my servers.
lukaslueg: Apparently what used to be
GB of storage consumedis confused withBytes of storage consumed, leading to a cool off by 2*30 error. > You're right to question my calculation. The MCP server failed to connect when I tried to look up the field definition.
321 points by merksittich · 264 comments
What HN said:
scrlk: https://archive.ph/3J3iw
Zigurd: Everybody wants a platform but nobody wants to spend what it takes to make a platform. That includes things like Windows Phone, Fire Phone, all the glasses, Humane, etc. As much as everybody hates on OpenAI for chaotic management, they did buy Jony Ive and are presumably giving h...
deepwoods: FT frames this as some aggressive escalation tactic, but document retention letters are extremely standard practice. At this point they're basically a formality, as any former Apple employee at OpenAI really ought to know by now that they could get dragged into this.
reenorap: Apple must have hard evidence on this. I can’t believe they would take it this far without already knowing they are going to win. If they have to fire a huge chunk of their hardware employees it’s going to throw their IPO plans into chaos.
288 points by rellem · 197 comments
What HN said:
babblingfish: Speculation: open models is what will kill Anthropic and OpenAI. Hyperscalers can run the models without a licensing fee. Apple can make them smaller and put them on the device. The frontier models are an edge and a liability. They're astronomically expensive to train.
GodelNumbering: Exactly 4 months ago, the marketshare on openrouter was 60%-40% in favor of closed models. Now it's 63%-37% in favor of open models. On March 19th, the open models processed 888B tokens in aggregate, yesterday, they processed 4.19T tokens in aggregate.
himata4113: When europes mismanagement of AI suddenly turns out to be the best thing that has ever happened. At best only few billions lost to training models that become obsolete next month.
hughw: This presentation is painful to read. It's an LLM's idea of a CTO presentation. I'm overwhelmed by charts, only slightly connected to the text around them. But no matter, it looks like a CTO slide deck. HIGH IMPACT.
260 points by neversaydie · 184 comments
Researchers have found an Earth-like, rocky planet orbiting within the habitable zone of a distant star with an atmosphere.
What HN said:
smb06: I feel like I've been hearing about atmosphere on Earth-like planets for a while. Or maybe I've missed some nuance.
tulio_ribeiro: Didn't know a rocky planet in the habitable zone of a red dwarf could retain atmosphere against intense stellar stripping. Red dwarfs are known to be cooler (the habitable zone is therefore closer) and unstable. I don't think LHS 1140b is "Earth-like" at all.
mekdoonggi: We should build a solar lens telescope. By the time we're ready to use it, we'll have a bunch of candidates to point it at.
jimbokun: 48 light years is in our back yard. Close enough that we could probably develop a probe to get there in the next few centuries and check it out. What are the current popular candidates for propulsion systems capable of accelerating to near the speed of light?
244 points by crazysaem · 158 comments
What HN said:
jcoder: The ring sizing has been such a debacle, first with needing to buy (!) their special ring sizing kit because standard sizing wouldn’t be accurate they said, and now they say that the ring sizing kits weren’t true to size and suggest this: > When in doubt, order a larger size.
bArray: Was just checking out the Index 01 [1]: > Why can't it be recharged? > We considered this but decided not to for several reasons: > You'd probably lose the charger before the battery runs out! > Adding charge circuitry and including a charger would make the product larger a...
haritha-j: The index is a ludicrous product. The battery life, they claim, is 2 years. This is one heck of a misrepresentation. It is actually 12-15 hours. They get the 2 years because its meant to be used around 10-20 times a day, in 3-6 second recordings.
lukeify: > Is there a warranty? > Yes, we warrant against manufacturing defects for 30 days after you receive your order. Is this the shortest warranty on a consumer electronics device ever? What on earth are they thinking.
237 points by giuliomagnifico · 154 comments
How the brain rapidly switches attention between competing speech streams in complex listening environments is not fully understood. This study shows that neural tracking of a new speaker emerges before disengagement, with transient dual encoding and reduced alpha power supportin...
What HN said:
NalNezumi: Related, but this reminds me of the story by Richard Feynman [1] when he practices counting up to 60 seconds in his head, and after many experiments around what he can do simultaneously conclude that he can simultaneously count and read but not speak.
_doctor_love: Stuff like this is what makes me not worry about AI overtaking humanity. The human brain + mind is so sophisticated, we have no idea what untapped potential still remains. And anecdotally, this headline feels like it's confirming something already well-known.
a_gopher: If reading aloud a children story, you may notice you are able to maintain an independent unrelated train of thought. While doing so, I notice that occasionally extra mistakes can "leak" into the story telling - e.g.
oggreen: Findings like these surprise me. I had a NDE a few years ago and my observation was such that everything slowed down tremendously and I was able to process every instant like it was slow motion. This wasn't like a "metaphysical" thing.
181 points by droidjj · 105 comments
Chinese AI lab Moonshot AI announced Kimi K3 this morning, describing it as their “most capable model to date, with 2.8 trillion parameters”. It’s currently available via their website and …
What HN said:
Eduard: LLM source data sets may have millions of data points for what a bike frame looks like, yet they still fail drawing them correctly. https://www.booooooom.com/2016/05/09/bicycles-built-based-on...
devttyeu: > How does the prompt “Generate an SVG of a pelican riding a bicycle” add up to 95 input tokens? OpenAI’s tokenizer counts 10, Anthropic’s counts 10 for Opus 4.6, 30 for Opus 4.7 and 25 for Sonnet 5/Fable 5.
OsrsNeedsf2P: It's incredible Simon still believes pelicans on bikes aren't part of the training set, despite hundreds of them on blogs, forums, and Github. Stuff we put in our company blog shows up known by LLMs 6 months later, and we have 1000x less traffic than Simon's own website
michaelbuckbee: Like Simon concludes the article, the main use of this isn't to say which model is "better", but to try and poke at the model to sort out things like quality vs cost vs speed. So I put together a quick comparison of the last couple iterations of Opus, Fable and now Kimi.
172 points by geerlingguy · 18 comments
What HN said:
waterheater: What a throwback. I remember reading this guy's build of the "Flying Nimbus" [1]--an early version of the Onewheel--back in 2015. In fact, the Flying Nimbus and the Onewheel were both released in 2014, but it appears they were produced independently, once again demonstrating the...
alex-moon: Every point of friction, every struggle documented in minute detail, on a long and bumpy path to some kind of result. The stakes are zero, the whole project an act of pure expression. It is like a kind of oasis in a desert of articles about new ML models.
arjie: Extraordinary post. To be honest, didn’t even strike me to use TPU for shear resistance. But of course. Good rookie tip among much in there. Fantastic work.
torpex: Dane is the real deal. I've never figured out how he finds the time to build all these amazing contraptions let alone document them in meticulous detail while maintaining a full time engineering job. But it's definitely impressive.
145 points by surprisetalk · 69 comments
Noticing responses to problems; there’s a whole set. The consultant’s reaction to each of these as they arise.
What HN said:
didgetmaster: People often attribute the government's inability to solve a problem even after throwing billions of dollars at it; as a sign of incompetence. While there is plenty of incompetence within government; I think the 'Preserve the Problem' response is mostly to blame.
0wis: Nice article, interesting to keep an open mind. On "No. 0002. Preserving problems", it can happen to people too, no need for a complex system at the size of a company. I have often noticed recognized experts keeping the root of the problem unsolved because it was justifying their...
rawgabbit: The "meta" problem is that political in-fighting usually results in local optimization everywhere. Various departments throw each other under the bus to steal budget/people/resources.
PeterStuer: Somehow made me think of every 'modern' HR department.
135 points by silcoon · 86 comments
Comparison of modern Lisp dialects
What HN said:
nobleach: Since a few folks here recommended Common Lisp to me as the language that would "tick all my boxes", I've been doing a deep dive. Right now, I'm working through SICP again with DrRacket. The first time I worked through it with MIT Scheme MANY years ago.
phtrivier: From my experience, to be happy I would need the perf of sbcl, the syntax, litterals and data structure of clojure, the begginer-friendlyness of drracket, the type system of ocaml, and the dev experience of rust. Does that exist somewhere ? I hope jank gets there.
dieggsy: CL also has pretty much arbitrarily extensible syntax: - https://sr.ht/~dieggsy/whisper/ - https://dieggsy.com/json-literals.html And could also be used to build languages, supporting more modern programming paradigms (though yes, I believe Racket does make this easier): - https:...
PaulHoule: I'm going to take the opposite position that there is much less special about Lisp than people think. Like I have met so many programmers that travel around like itinerant martial artists looking for true functional programming and they never find it.
AI/ML
Other
Programming
Science
Stories and comments sourced from Hacker News public API. Not affiliated with Y Combinator or Hacker News.