News for hackers. Stuff that matters.

deklesen

Re:Mathematics has changed. Let's act like it on Sunday November 01, 2026 @05:42PM (#50330695)
Sometimes I wonder if "art" is also just some form of pattern matching on a very meta level.
Mhh i wonder how this would have evolved, since, you know, for 99.99% of time, humans didn't really get any information other than from their immediate proximity. And if they did (e.g. earthquake), it was very much relevant and should affect them.

Just flaunting my ignorance of biology here.

Re:Muse Wonder on Monday October 12, 2026 @06:10PM (#50063768)
Maybe not in the real world...
Re:Advancing the price-performance frontier with GPT‑5.6 on Thursday July 30, 2026 @01:57PM (#49113407)
Good observation! Kudos
Re:GLM 5.2 Is Out on Saturday June 13, 2026 @03:19PM (#48520490)
... really? are you sure about the timezones? That's kind of odd, isn't it?

Maybe the post was edited afterwards?

Re:Liquid AI reveals 8B-A1B MoE trained on 38T on Friday May 29, 2026 @05:28PM (#48329529)
Yeah, but you are not washing yourself there, I suppose?

The whole twist here is that to wash your car, you need your car, so you cannot go by foot.

The site is broken beyond belief on latest android beta, latest chrome on android.

As in, it doesnt work at all and just shows an error on load.

Re:The AI Great Leap Forward on Wednesday April 08, 2026 @05:18PM (#47696418)
This also seems to implicitly assume that ai models won't get better - a bet I am not willing to make currently..
Mhh... my hunch is that part of this is that all python keywords are 1 token, I assume. And for those very weird languages, tokenizing might make it harder to reason over those tokens.

Would love to see how the benchmarks results change if the esoteric languages are changed a bit to make them have 1-token keywords only.

Re:Launch HN: Cardboard (YC W26) – Agentic video editor on Thursday February 26, 2026 @02:54PM (#47171184)
Nice demo experience!
Re:Nano Banana 2: Google's latest AI image generation model on Thursday February 26, 2026 @02:46PM (#47171095)
For a podcast on this topic (niche pornography and how it was affected by the advent of pornhub and the likes) check out "the butterfly effect"
Re:What Is a Centipawn Advantage? on Monday February 23, 2026 @05:45PM (#47130080)
I learned chess when I was 5, and didnt have a chess computer in the first like 5 years and by then I have progressed quite far.. so i cannot really tell
Re:What Is a Centipawn Advantage? on Monday February 23, 2026 @01:25PM (#47126493)
FWIW, as an avid chess player, I find the "up 5 pawns" has more intuitive signal.
Re:Node Banana on Sunday February 22, 2026 @05:46PM (#47115613)
Can I not just do this in ComfyUI?
Re:Music Discovery on Sunday February 22, 2026 @05:42PM (#47115577)
Actually, I find Spotify horrible for finding new music outside my bubble that i like. YouTube works much, MUCH better for me for this purpose.
Re:Music Discovery on Sunday February 22, 2026 @05:41PM (#47115564)
You seem knowledgeable about this.. Care to test my old project for music recommendation? I built it by looking at co-occurrence of artists in Spotify playlists, which gives me word2vec-style vectors, and then its just kNN.

No login needed, just enter some artist names and see what you get:

https://blog.jonas-klesen.de/artist2vec

Re:Music Discovery on Sunday February 22, 2026 @04:41PM (#47115011)
I entered "if tipper and lsdream had a child".

4 of the 5 links didn't even work. The other sounded nothing like what I described.

Show HN: Earworm.lol – fast universal music links

https://earworm.lol/https://open.spotify.com/track/5BBK1WoaiD3b3sJ2UysLuM

Something I built out of personal necessity. Now that more and more people are rightfully leaving Spotify, I was searching for a simple way to prepend a url to any Spotify/Deezer/Tidal/AppleMusic link and get a universal link which allows visitors to navigate to that track/album/artist on their platform of choice. I found song.link, but it takes >5 seconds to load, even on popular songs. So...

( Read More... | 1 comment | 2 points | on HN )
its written in the post

Show HN: YapType: hassle-free yap-to-text for Linux

https://github.com/jonas-klesen/yaptype

I found that having a quick way to locally transcribe what i am saying is helpful for my workflow, especially when vibe-coding or generally when chatting with AI models, as small transcription errors are mostly not a problem.So I built YapType. It's just <200 lines of code total. On press of a shortcut, it transcribes locally and opens the result in your text editor - simple as that.

( Read More... | 1 comment | 2 points | on HN )