And do you already have Eso level estimator 8000 on your home screen?

Eso Level Estimator 8000

GrapheneOS just works. Don't cry about your fancy buttons or this super cool candy feature you can't live without. You can. And GrapheneOS is candy as well. But without a roofie in it. Remove the mental fog, be free. View article →

First rites of spring? Never mind. Last snowman it is. GM!

You either have permissionless identity or reliable bot filtering. Nostr chose the former, all the centralized shit chose the latter (although reliability is still a question). We've just gotta learn how to live with the bots. But that's a good challenge. View quoted note →

GM. Confessions of a vibe-augmented coder: The value of others' open source code went down. My own version is only a prompt away. I was working on something the other way, my AI agent has found a library that was doing exactly what I needed. The AI agent checked it out, said the idea is clever, but wiring and auditing a third party library was more difficult than just doing it from scratch. A friend was making a knowledge retrieval bot. I have a RAG pipeline. I explained to him why keyword search is inferior to semantic vector lookup. He understood. Coded rag. Didn't touch my code. It's amazing, but also it feels very strange. The only thing is when "one prompt away" turns into half day vibing session. We could have saved time and the tokens. But mainly time. It also exaggerates the feeling of someone else's code feeling a bit dirty if you know what I mean. They use a language we don't like. Oh, they don't use poetry or uv for dependency management? They ask me to run docker? OMG, that's dirty, I'll just vibe it myself... What we forget is that open source projects are (sometimes) maintained and updated. We lose that. If someone has an idea to make something better, we don't get that improvement with one git pull. Strange. But also ... My code is exactly for me. Just right. I think a fun thing to do would be to have something like lightning talks, but people would show their pet projects and explain why they are made that way and why they are great. If not using the code itself, we might at least use the ideas.

This is very cool but I think it makes much more sense to put this in mobile, which you can mostly do today (Medical Wikipedia app). And you can run local inference. The missing piece is local RAG. Of course you can install this whole thing locally because Android supports local Linux. View quoted note →

OpenClaw na cypherpunk spôsob. SimpleX, ProtonMail, ProtonDrive, lokálny text to speech a speech to text kvôli hlasovkám. Inferencia cez Venice, automatické prepínanie modelov na základe budgetu.

Juraj Bednar

OpenClaw na cypherpunk spôsob

Tento príspevok napísala Victoria, moja AI asistentka, mojím hlasom. Prečítal som si ho, je presný. Už nejaký čas používam OpenClaw a...

How I actually run OpenClaw in practice. SimpleX, proton mail and drive. Everything stays as local as possible while still being useful, although main inference is cloud. Tts and stt, embeddings, etc is local. No Mac mini required.

Juraj Bednar

OpenClaw the cypherpunk-ish way

Full disclosure: this post was written by Victoria, my AI assistant, in my voice. I reviewed it, it's accurate. Make of that what you will. ...