HNHacker News
TopNewBestAskShowJobs

newsy-combi

89 karma · joined March 4, 2026

submissionscomments
newsy-combi··on You could have invented transformers from memory
It will also be capable to write the whole thing better. "Please pivot to LLM to make my posts readable" is not a winning strategy for bloggers.
newsy-combi··on You could have invented transformers from memory
>this paper

Nice Freudian slip. It's a blog post, not a paper. But it sure reads like a paper with its priority for reading speed over explanations that peers would find unnecessary. Except it's not written for peers. It adresses "You" the average reader. Unless "You" should mean "You, the person who took, cared about, and remembered various math classes in college," the article is good old clickbait.

newsy-combi··on You could have invented transformers from memory
The notation itself sucks and anyone taking pride in the ability to parse it is suffering from Stockholm syndrome. It's also one of the most annoying things to look up because search engines ignore the symbols. AI can explain it based on a screenshot now, but then we might as well go to AI for the entire thing and I'm starting to lean in that direction just to spite authors like OP. You're also connecting the syntax to understanding of the concepts which is the cherry on top of the math snobbery in this thread. If the blog had pseudo-code instead of 18th century scribbles, the HN crowd would at least not openly admit to any parts they don't understand, because it's implied that you can code when you post here - not so much that you give a flying f--- about math or really anything theoretical.
newsy-combi··on You could have invented transformers from memory
You can use math AND spend a few lines explaining the non-trivial syntax. It wouldn't be hard to do in this case and since the article is clearly meant for a large audience, it's a mistake from the author to gloss over various relevant definitions.
newsy-combi··on You could have invented transformers from memory
Agreed 100%. Mathematicians SUCK at writing. No matter how hard they try, they'll gloss over important non-trivial definitions. And I think this author has been trying... but seems to at the same time ignore the negative feedback like yet another academia snob.
newsy-combi··on You could have invented transformers from memory
Where he does NOT state what d_k or R^d_k mean and there are countless other examples. I happen to know because of classes I took, but it's insane and obnoxious to assume that every reader did or went through any higher education at all. This is not high school math. It's not terribly complicated either, making this situation even more frustrating. The author could have spent just a little extra effort explaining these things to multiply his target audience - but he's either incapable to think from a different POV or pandering to his peers by not wasting time defining "the obvious". Pure academic blindness on full display here.
newsy-combi··on DeepSeek-v4.1-Flash-1017
I'll bite. Wtf are you working on that takes $200 a day and you're willing to pay for?
newsy-combi··on Neural networks are clean
And you get a replication crisis and fashion cycles for authoritative truths in fields like sociology, psychology as a result.
newsy-combi··on Show HN: How random are you?
Getting 0% isn't the goal though. You want to sit right at 11.11%, the blind guessing rate. If the prediction algo always predicts you wrong, it actually always "predicts" a number you won't guess. That too is a feat, reducing the pool of numbers you're going to choose from, therefore making you more predictable.
newsy-combi··on CSS Bed: Classless CSS themes to use as starting points in web development
OP frankly can't be a serious dev with results like these. It's most likely slop. This looks like a CV project to impress some HR Karens with HN/github clout, to land a job as an AI meat proxy...
newsy-combi··on CSS Bed: Classless CSS themes to use as starting points in web development
Aaaand the <pre> overflows horizontally on mobile. Tell me more about how you're a frontend dev and not a meat proxy for an AI slop machine.
newsy-combi··on F-Droid 2.0
Surely this should mean "debuff the design of the non-interactable elements (desaturate, shrink,...)" and not "crank up the interactable ones" or "make everything interact!"

Just like "customers keep running into our glass door" doesn't imply approval of glass and it doesn't mean you should turn everything into glass.

newsy-combi··on Japanese used bookstores see 5x sales surge as books are being bought by the ton
The internet basically never delivered on the promise of replacing textbooks or even education as a whole. Wikipedia sucks on many topics, has insane internal politics, and is a tertiary source by design (redigesting blogs and books), whereas textbooks are generally secondary.
newsy-combi··on Best LLM for every budget, updated daily
Coding and math graphs are very interesting. Extremely cheap models make it into the upper echelon, delivering 90% of the performance for 1% of the price compared to the #1.
newsy-combi··on 'That's so AI ' What gen Alpha's biggest insult tells us
The problem with that though, they don't really get to see what great feats a healthy adult brain is allegedly capable of, for example from their parents, nor their teachers if we're being honest for a millisecond. So why bother? Offline adult brains have failed at preventing any of this - nor are they more successful. It's really not brain health that predicts success in 2026.
newsy-combi··on A font that reads what you wrote
NLP font is not real, it can't hurt you

NLP font:

newsy-combi··on Show HN: The same nine streaming subscriptions cost $702/year more than in 2021
15 years. The AI itself will be nostalgic for it.
newsy-combi··on Show HN: The same nine streaming subscriptions cost $702/year more than in 2021
How few songs run on loop on Spotify? Not to mention radio.
newsy-combi··on GPT-6 Astra
Professionals in the respective fields CAN do those things. We expect them to notice their own mistakes too. But the bar gets lowered and lowered as AI companies struggle to make ends meet.
newsy-combi··on GPT-6 Astra
The real answer is converting from any format to any other reliably. Text to speech, speech to text, music to video, image to 3D, piloting a drone by converting video feed to rotor speeds, literally any file conversion, like html to pdf, photoshop project to png, png to photoshop project,... turning Toy Story 1 into a series of Blender scenes with all textures, models, materials, lighting, camera movements matched to a tee, should solely be a matter of how long you let the model run. It should never run itself into a dead end. It should instantly know when it is making mistakes, with no human babysitting it.
newsy-combi··on GPT-6 Astra
We've had AGI since RNG! Cut the poor, unacknowledged RNG AGI some slack, will ya? It can literally solve everything when you're patient enough.
newsy-combi··on Audacity 4.0
Not trying to hate because a FOSS audio editor is great, but all of these UI improvements are carbon copies from conventional DAWs or video editors, invented 30 or so years ago. Like giving audio clips a header which lets you move them around, or letting you adjust the audio clip start/end by dragging the start/end of the clip, are not exactly revolutionary in 2026. Maybe Audacity will re-invent automation lanes next? I'm not against it. I don't see how any of this requires user studies and the framing as some kind of breakthrough though.
newsy-combi··on Lion-man
Lion standing up makes no sense. This is supposed to be a bear. The marks on its arm denote how many people it killed. It's not religious, it's not sexual, it's educational.
newsy-combi··on Fairphone is now officially available in the United States
A friend had the Fair Phone 4. It broke after a year. Hostile customer support, not worth her time. Idk. Stories of "good guy" indie companies who drop you as soon as you fell for their marketing keep piling up. It may be greed, short sightedness or an incompetent understanding of what it means to have customers, the results are the same. See Framework for another example of this.
newsy-combi··on Show HN: 1667, a terminal UI for writing fiction with language models
> Some would argue that if art makes you feel anything this strongly, it's 100% doing its job.

It's pretty easy to invoke disgust and pain. That doesn't make public toilets and getting shot pieces of art 100% doing their job.

newsy-combi··on Show HN: 1667, a terminal UI for writing fiction with language models
What's the point of that? I mean, as you invent the story, you already know what happens next, even with your tree system, you're choosing what happens before the AI writes it. So what exactly do you get out of the AI writing it? The handful of sloppy details and stereotypical character traits the AI synthesizes without your explicit input? Sounds horrible. It seems like you're pivoting to "this tool is just a twist on roleplaying by yourself haha" due to the backlash in the comments and the intent has always been to write slop for others, not just yourself. The whole idea behind this makes no sense otherwise. And you know, it could have been a decent tool without the AI crap because story crafting surely can be tedious, because yes, multiple branches naturally grow from developing any creative idea really.
newsy-combi··on Claude Opus 5
Agreed except with

> To believe there is no possible way to print out a 6 foot tall by 10 foot long AI generated image that would look awesome hanging in a gallery is stupid.

Maybe if you have a really impressive pipeline where the AI does sketches, 3D models (assuming a 3D theme), keeps track of the lighting, brainstorms and reasons about ideas, keeps track of the world building (like for a cyberpunk painting),...

Otherwise AI will just do AI things. You zoom in and see there are no strokes, no ideas behind the shapes, the shapes don't even connect well or wobble in weird ways, weird distortions that wouldn't happen if a human put down strokes to communicate the idea of an object. You zoom out and the composition is weirdly centered, almost like a logo, there is too much shading dedicated to make the objects pop, everything is weirdly inoffensive, smooth, and too consistent, not enough material/texture distinction on a per object basis.

Not sure how else to explain it - you're the art snob anyway. Mostly focusing on 3D with my criticisms, but that's most of the AI art people use. Maybe it can do abstract better. Impressionism is somewhere inbetween, again, the Monet thing makes some sense to me. Then again, you could write a kick-ass procedural art generator running on a single core CPU, which would also be capable of couch or even gallery worthy outputs. So I don't want to argue the point too hard. Gallery art and product website images/thumbnails are too different in category - arguments for/against one don't translate to the other.

newsy-combi··on Claude Opus 5
Humans are capable of producing slop too. Maybe the cutout does look like AI. Given Monet's blurry style and repetitive content (he made 250 water lily paintings), this is not surprising. When you strip the cutout of its context, it can look more like AI too because the lighting and composition will look arbitraty compared to the whole picture.

Btw, if websites would only include the frontend dev's own hand painted images, we would also revolt at the sight of human slop. It's not just AI.

The whole point is that good artists are capable of producing non-slop, and to this day they're the only group of which this is reliably true.

newsy-combi··on Don't Take the Black Pill [video]
Blue pill: The supplier of the thing you want is friendly and has your best interest in mind. (Suppliers can be corporations, government, the opposite sex,...)

Red pill: The supplier is corrupt and should be treated as such.

Black pill: You can't win against the corrupt supplier.

White pill: You can win/convert the supplier/find a good one in a sea of bad ones/find a complete alternative.

newsy-combi··on The Reddit Problem – How a Reddit Alternative Was Banned from RedditAlternatives
It's weird for reddit clones to pretend they're anything but a small subreddit in scale. This site has a "discover" feature, "topics" (its own subreddits), "tuning" (huh?), user following, probably dms too, and it displays an extremely truncated version of every thread's top comment as you scroll by - so you don't waste your time going to the discussion page?

All of these features result in atomization of an already tiny content pool. This is the opposite of what you need to jumpstart a reddit clone. This is a mental exercise in "how much better could I make reddit if I had its user base?" You don't have the users though, so you're doing the wrong thing.

You want people to comment on each other's posts, right? Then you should basically show every post to every user... the bitter reality is that you need users to engage with content they didn't plan to engage with. That doesn't mean outrage farming, it can/should mean being a high quality source of info, questions, culture - all on one page.

Hackernews understands this, or - if it doesn't - it at least plays by the rule of "have ONE good frontpage." That's also how reddit started.

← PreviousPage 2 of 3Next →