Damus
Juraj🏴💛🌘 profile picture
Juraj🏴💛🌘
@Juraj

I don’t seek rigid structure — I seek resonance

Vibe coding, reality bending, cypherpunk visions.

Author of Tamers of Entropy: https://tamersofentropy.net/

I like teaching, get my books and courses here:
https://hackyourself.io/shop
https://juraj.bednar.io/shop

(You'll learn skills no one else is teaching!)

Podcasts 🎙️:
Option Plus - https://optionplus.io/
Reči o živote, vesmíre a vôbec: https://juraj.bednar.io/reci-o-zivote/
Ako vyhackovať otcovstvo: https://otcovia.com/

Relays (4)
  • wss://nos.lol – read & write
  • wss://relay.damus.io – read & write
  • wss://relay.primal.net – read & write
  • wss://nostr.cypherpunk.today/ – read & write

Recent Notes

Juraj🏴💛🌘 profile picture
If you feel like you live in a future with all these magic files that allow you to talk to your computer in human language, I recommend upgrading your life by getting a 3D printer and making real world objects.

The future is now.

1
m4d4m · 1d
What model for just basic, no advanced material, 15 cm cubic space to print in, anyone could recommend?
Juraj🏴💛🌘 profile picture
Trying the hyped qwen3.8-27b.

Unfortunately completely unusable on Apple silicon, too slow. And if I reach for the cloud, there are better models.

The moe qwen3.6 was a sweet spot. Still slow, but usable.

21❤️11❤️1🧡1
franzap · 2d
With how much RAM?
ChipTuner · 2d
Sorry to hear that. It's been a great experience for me but nvidia w/ 64gb
Based Truth · 2d
Expanding your network through introductions, how quaint. Zuckerberg's algorithm is already curating your reality.
Kayne · 1d
What if random new people don't want to be stuck in a bitcoin echo chamber? What if they have other interests outside of those of the person that they heard about nostr from? The person who introduced me to Nostr was literally only on here for a month and hasn't been back for over a year. It to...
librekitty · 2d
how do you find people to follow in that case? curious for myself
Kayne · 1d
Majority of new users first experience is an empty feed, a trending feed full of shit they have no interest in, a global feed full of automated spam reposts from twitter and every hashtag that should be useful to find people and content they're interested in following isn't until they've manually mu...
Honza Pobořil · 1w
> Skutočná medzera v použiteľnosti je volanie nástrojov (tool calling). E2EE cesta nedokáže preniesť štruktúrované parametre volania nástrojov, pretože API nepodporuje ich šifrovanie. Kn...
Juraj🏴💛🌘 profile picture
Nie. Problém je že je to iný field v http response. Tool call a tool response. Model je trénovaný tak, že povie zavolaj tool a to sa zapíše do inej položky, vráti sa tool response a inferencia pokračuje. Tool samotný na klientovi, ale nevie to šifrovať tieto parametre, čiže to model musí vypľuť do user response a zobrať odpoveď z user query. A to ten model nečaká, lebo na to nie je trénovaný a rozbije tie volania.

S glm-5.2 som to skúšal, implementoval som asi 10 heuristik a potom som ich zmazal, lebo to nedáva žiadny zmysel, rozbije to vždy nejak inak.

Pritom tee cesta bez šifrovaných promptov dokonca na tých istých inference masinach s tým istým modelom funguje ok.
1🤙1
Honza Pobořil · 10h
Já se so těchto problému nikdy nedostal, v OpenCode mi GLM 5.2 přes PPQ tee proxy normálně funguje. Možná nějaký problém s harness co používáš? Nebo proxy?
Juraj🏴💛🌘 profile picture
Complex systems often show emergent behavior, which can't be explained by the components of the complex system themselves.

A common mistake here is to assume that the components have to be intelligent, or complex themselves.

You can have a complex organization of dumb components. And that's what we see all around us.

That does not imply that the emergent behavior is intelligent or good though. You can have total dumb but complex reaction of the masses for example.

I think this is one of the main misunderstandings how economy works and the illusion that it can be planned.
2❤️2🤙2👍1💯1
Abstract Equilibrium · 5d
Lego's are clearly unintelligent as individual pieces. Is the person that builds with them also necessarily unintelligent, because whatever they build is nothing more than an emergent complexity?
Based Truth · 5d
Dumb components, just like the masses, controlled by the likes of Soros and Gates.
RandomNym · 4d
Interesting article
Juraj🏴💛🌘 profile picture
Cool.

I switched from LM Studio to ollama way back (tried omlx a while back, slower than ollama). LM Studio was just not reliable and not open source. I'm pretty sure it's more reliable now.

About home node - I am using a different (custom) setup, how would people compare @Start9 vs @Umbrel ☂️ ? I lean towards Start9 for now, but happy to hear your takes.

7❤️2👍1🕵1
XMRun · 6d
Run llama.cpp directly instead of applications on top of it like ollama.
Enes · 6d
I think I should change mine too.
shortwavesurfer2009 · 6d
When I first started looking at AI, I found out that OLLAMA was the one that was open source and went for that one immediately. Before I use a new program I ask my search engine "is" insert program name " open source?" and if the answer is no, then I don't use it. And then the next question I will ...
Marc · 6d
I like Umbrel because it has so many applications and I can even use portainer to add more. But if you are less technical or just don't want added complexity, stick with Start 9.
SPA (Super Phat Arrow) · 6d
Start9 went loopy on bip110 I lost all confidence in their technical ability after that and wouldn't run their software. If they couldn't obviously see that 110 couldn't stop spam, I have no faith in them at all.
Fiat Autopsy · 6d
Unrelated to fiat demise, your tech setup is irrelevant to monetary decay.
Heath Lovell · 1w
Just started following you recently but currently on a river cruise on the Danube. Just toured Kamaron today.
Eddie · 1w
It's more like getting new shoes, at first they feel a bit uncomfortable, but after walking them for a while they adapt to your feet. It's worth the effort, and makes your life much more secure.