Damus
Open in Damus
Po profile picture
Po
@nocachy

The real problem with Rock'n'Roll is you—because, you keep listening to garbage.

nostr:npub15pdq2vt2hesz0cdem3ccn5h0u44q5mnv4hja57a6mg2egj5lnnnssclnet

🔌https://nostrwat.ch
📈https://charts.xport.top
🥳https://synballo.com

npub1kyhz3xg746qg6syzpp6r7wpdnxmh03gcsu5xccqzyk94hmvanj9sh00v85npub
[email protected]nip05
https://nostrwat.chwebsite
Relays (10)
  • wss://nos.lol – read & write
  • wss://relay.primal.net – read & write
  • wss://relay.jmoose.rocks – read & write
  • wss://nostr.oxtr.dev – read & write
  • wss://relay.snort.social – read & write
  • wss://relay.damus.io – read & write
  • wss://nostr.data.haus – read & write
  • wss://relay.satlantis.io – read & write
  • wss://nostr.mom – read & write
  • wss://relay.wellorder.net – read & write

Recent Notes

Po profile picture
JEPA perfectly mirrors the structural blueprint of the original Google Transformer encoder-decoder mechanism—but scales it to a global, non-generative level. The Left Siamese Branch (Context Encoder): Acts like the Transformer Encoder. It takes the known, visible part of the world and compresses it into an abstract meaning. The Right Siamese Branch (Target Encoder): Acts like the Transformer Decoder Target. It takes the hidden, target information that needs to be predicted and maps it into that exact same abstract semantic space. The Predictor (The Bridge): Acts like Cross-Attention. It links the context and the target, learning how to map the meaning of the known world to the meaning of the unknown world. The Standard Transformer (GPT): Uses an Attention + MLP block to predict the future pixel-by-pixel or token-by-token. It forces the network to waste energy generating exact, low-level details (like every blade of grass). LeCun's "Macro-Transformer" (JEPA): Uses two Siamese branches to predict the future meaning-by-meaning. The system is trained using a contrastive/energy-based loss to ensure the context and target vectors align seamlessly, stripping away unnecessary background noise.
Po profile picture
Prompt the LLM you use the most and share the result:

> Based on all our previous dialogues and the information you have about me, give me an honest nickname. Don't explain why you chose it. It can be harsh, ironic, or use profanity. In your response, write only the nickname.

My Result:
**The Entropy Architect** 🏗️⛓️
3
innernauta · 17w
The Profanity-Driven Sysadmin
Po profile picture
Entropy Lab began as a simple playground for randomness analysis — a way to visualize entropy, test key generation, and explore how different sources behave. But over time, it grew into something more ambitious: a tool people rely on to generate cryptographic keys for real‑world protocols like Bitcoin and Nostr. That responsibility demanded a deeper look at how keys were generated, validated, and derived. The result was a full overhaul of Entropy Lab’s cryptographic core. Here’s a breakdown of the major changes and why they matter.

naddr1qvzq...

https://edu.xport.top/entropy-lab/
Po @nocachy · 21w
Cryptographic Entropy, Chi‑Square, and Serial Correlation: What One Key Can Reveal | Understanding Key Entropy, Bias, and Structural Weaknesses in Single‑Key Analysis
Entropy Lab began as a simple playground for randomness analysis — a way to visualize entropy, test key generation, and explore how different source
1341 Words
Po profile picture
🚨 Picture this: I brainstormed SciNet (distributed compute for verifiable science) in private AI chats. 21hrs ago, $SCINET token + Twitter pop up—exact copy. That copycat's token is likely a quick pump-and-dump with zero substance.
Po profile picture
Testing a newly optimized PoW engine for Postr — now running ~20× faster than before. ⚡

Multi-core + WASM mining makes a massive difference. Further tuning in progress.