Snar

PiSwap L2, aka PiEthereum Wallet v2, a little walk through of the new functions. I will post a demo of the any token swapping function which now utilizes @Uniswap v4 Universal Router. This is a free open source raspberry pi python project for any L2 EVM chain. Build it yourself via instructions on github. Link in Bio. Touch the blockchain with no intermediaries and many private keys.

@toady_hawk buying selling hobby related stuff IRL is awesome. it pairs you with people who share your interests, an instant friend possibility!

managed to sell our old sofa set on facebook marketplace in less than 3 days, to a real human, without going insane or getting stabbed, picked up the evening before our new set gets delivered… we are so back.

gotta go LLM bargain shopping sometimes

@tkarrdeiz wow yes minimax2.5 showing some promise! can you run minimax2.5 50k token context on the 8k mac setup smoothly?

@PunksGan MiniMax2.5 is your best go to on Olama. If you don’t like it, stay frontier

Over the weekend I tested 17 different LLM’s through various optimizations and as of this morning I am running MiniMax2.5 at 6Bit quantization at 83/tokens per second and beating Grok at usable intelligence, all for the ongoing costs of running a 100 watt lightbulb. M.S. M3ULTRA/256RAM

@Thomasdelvasto_ have to entertain the answers to Why, before you can stop asking Why

Consider that all the people selling you the idea that meditation will make you super happy 24/7 don’t have complex emotional issues to deal with

@tkarrdeiz i was accessing gpt-oss 20b through Ollama cloud. Good point maybe they are running it on the wrong specs

@PunksGan How much RAM do you have?

@tkarrdeiz i tried gpt-oss 20b for agentic coding on openclaw and it is pretty crappy... any other better models that would work on your mac setup that i can test out?

@PunksGan they can yes, but bang for buck goes to Claude. it’s frontier and only $100 a month for basically endless usage. GPT-OSS 20B scored 5/5 for coding on my setup this weekend. that’s the smaller model I’ve tested that scored 5/5 on coding

You start to be boring the minute you want to be taken seriously.

@tkarrdeiz thanks for the tip. Interesting that llm is available on ollama cloud to test its real life agentic coding abilities.

@tkarrdeiz So no accessible hardware/LLM combo that can really do agentic hardcore coding yet? i've been playing with qwen3: 8b for agentic coding which doesn't work, it has some limited use for basic structure outline, maybe planning small coding projects. So my hopes are fading fast.

@PunksGan $8k, and no not for coding. Honestly you can get more high end coding out of a $100 claude subscription if that’s all your doing

1. never ever set your global default openclaw LLM to a slow ass agent. 2. never set your default main agent to a slow ass agent.

@RoniBandini que divertido

OpenClaw 🦞encontré una ticketeadora térmica, le pedí que aprenda a manejarla y me imprima por las mañanas el "OpenClaw Ticket News" verificando con la webcam. La cantidad de tareas que llevó a cabo de manera autónoma 🤯

@brucefenton a truly heinous ghoul

Last month Jared Kushner presented “The New Rafah” promoting his real estate development project on land in Gaza where 60,000+ people, mostly women and children have just been killed and millions displaced. Trump says Jared was a major advocate for starting the war with Iran.

@AlexSlobodnik the latest claude code model update no good?

@AlexFinn does qwen3.5 27b have enough strength to power constructing large codebases?

If you have your OpenClaw working 24/7 using frontier models like Opus, you're easily burning $300 a day. That's $100,000 a year. I have 3 Mac Studios and a DGX Spark running 4 high end local models (Nemotron 3, Qwen 3.5, Kimi K2.5, MiniMax2.5). They're chugging 24/7/365. I spent a third of that yearly cost to buy these computers I'll be able to use them for years for free On top of that they're completely private, secure, and personalized. Not a single prompt goes to a cloud server that can be read by an employee or used to train another model I hope this makes it painfully obvious why local is the future for AI agents. And why America needs to enter the local AI race.

@FreeMrktCptlst 😆😆😆

the cost of healthcare, food and housing will fall 99% over the next five years thanks to ai humans have created more than enough wealth to last until the sun blows up there’s literally nothing else to do except smoke weed and have fun we made it!

@0xNairolf Moltlaunch?

we need an ai agent marketplace > describe the problem > get recommended agents > hire one > give it context > come back later > problem solved life changer

@wassoshi 🪖🫡

It was disturbing to live for a week at the mercy of invisible fighters and drones not knowing when they’ll strike somebody around you. Having to find a quick way out and driving through an actively bombed area was not on my bingo card. After 24 hours of detours and layovers, I’m glad to be out and back to my family.

@brucefenton The “petty tyrant” archetype was created by carlos Castaneda or rather his teacher don juan. Ive never heard the term outside of his book “the fire from within.” Ive only had the pleasure of brief run ins with smaller petty tyrants, but I have seen them at work in others.

@PunksGan - what makes you think of Carlos out of curiosity

Went to the town council meeting last night and referred to the town manager as a “petty tyrant”. Town council lost their minds, one tried to censor, another paused the meeting and a third (shown here) screamed like a maniac. Amazing how triggered some people are by words. 😂