Luc Stepniewski @lstep

Local Laya moggs Jev at @grok 4.7-built Tetris 🧩 An open-weights System One model called Laya, beat cloud-based Jev at playing Tetris by making decisions 11 times faster, running locally on a 16GB MacBook Air! Run AI models locally -> atomic.chat

雨夹雪❄️ @mizorewww

7 days ago

介绍比Jev快50倍，在你设备上跑的laya-mlx! 只在你的设备上占用最高1G内存 Laya是一个开源的类似于Jev的,基于文本输出概率的分类系统 我将其移植到MLX,并且做了一些性能优化! 视频中就是这个模型在我的本地M3Max上玩贪吃蛇 这个模型能够以每秒决策60次的速度玩贪吃蛇！ github.com/mizorewww/laya…

291 1K 14K 4.2M 18K

Je résume (unpopular opinion) : - Dario qui doit introduire Anthropic en bourse sur un marché à sec de liquidité émet l'idée de ralentir sur l'IA parce que Claude, il est trop fort et il va tuer la planète. - Sam qui a le même problème que Dario se dit que c'est pas con parce que ça va finir par se voir que les centaines de milliards annoncés n'existent pas vraiment et qu'il faudra vite trouver autre chose pour refourguer sa dette. - Elon qui doit diluer ses investisseurs sur des mois d'unlock de $SPCX saute sur l'occasion, comme ça Grok aura plus de temps pour Groker en orbite. Bref avant d’imaginer qu’un Skynet s’est échappé du labo, je regarderais aussi les calendriers financiers des 3 gugus. Je ne prétends pas que leurs inquiétudes sont bidon. Je dis que sauver l’humanité et préserver ses intérêts financiers ne sont pas nécessairement deux activités incompatibles. Leur unanimité mérite qu’on regarde les deux.

SaxX ¯\_(ツ)_/¯ @_SaxX_

2 weeks ago

Ça pue ! Ça pue ! Ça pue ! Sam Altman, Dario Amodei et Elon Musk s'accordent à dire qu'il faut ralentir sur l'IA ! On va apprendre dans quelques temps qu'il y a eu une dinguerie qui est arrivée avec un frontier model.

191 362 2K 504K 317

Dario has written that we need to “pace the frontier,” and Sam has agreed. People may be surprised by my response: go ahead. You guys are the frontier. By any reasonable metric — market share, revenue growth, model capability — the two of you have a duopoly on frontier intelligence. You’ve also claimed the lead is widening because of recursive self-improvement. I don’t see what you see in the lab. If the unreleased models are scary enough that you think you should slow down, I support your decision to be responsible. But stop pretending you need anyone else’s permission. Stop pretending antitrust law has to be suspended so you can form a cartel. Stop pretending you need a regulatory approval process that supersedes product liability. Stop pretending METR is independent when it is intertwined with Anthropic’s investors and staff. Stop pretending you need those same evaluators to police competitors who aren’t even at the frontier. Most of all, stop pretending the motivation to slow down is purely altruistic. You face massive product-liability exposure if your products enable a truly damaging cyberattack. The market already punishes models that behave in unpredictable or unauthorized ways. After the Hugging Face episode, it is simply good business for OpenAI and Anthropic to trade some raw power for reliability and predictability. Call it alignment if you want. It is also just giving customers what they want. Pacing the frontier would also create breathing room for a more intelligent conversation about regulation than Bernie Sanders’ “shut it all down.” China is very unlikely to join a global agreement, as you know, and that has to be taken into account as well. So go ahead and pace the frontier. You are the ones setting it. The easiest way not to build superintelligence is for you to agree not to build it. Demanding your preferred regulatory framework as the price of that will look like blackmail of the public and the political system. So just do it. If you do, you’ll buy goodwill for the next conversation. If you don’t, we’ll know this was just another bid for regulatory capture — or an election-season psyop.

@PessimistsArc Right. Dario was already claiming that GPT2 was too dangerous to open source back in 2019. I made fun of them then. Everyone should make fun of them now.

Sharing an #Obsidian vault across multiple computers with #Git gives you full ownership, revision history, and zero subscription fees. But a default `git init` quickly leads to sync friction and merge conflicts. Here is my step-by-step guide to make it seamless 🧵👇

This is the type of post I might regret later, but I really had an urge to rant about the current discourse around pacing the frontier. lucumr.pocoo.org/2026/9/12/pdoo…

Six terminaux ouverts. Un agent de code dans chacun. Une URL de ticket collée dans chacun. Et j'attends. Ce jour-là j'ai compris que je n'étais plus développeur. J'étais un tuyau. Alors j'ai construit autre chose. 👉 yoandev.co/specifier-c-es…

Two years on Mac, same audio mess: * mute disappearing when I switched headsets * loud wake-ups * wrong mic in calls * one mute button per app So I built Sotto: pin devices, keep mute muted, use global mute or push-to-talk. sotto-mac.com

I'm trying out an /implement-spec skill Essentially a multi-agent implementer that: - Takes in a spec and tickets - Does codebase research in a subagent - Implements all the tickets in subagents with maximum concurrency - Reviews the final code against the spec - Cleans up all worktrees Should be able to smash out huge chunks of work autonomously with minimal supervision. github.com/mattpocock/ski…

@itsharmanjot For my RTX 3060: huggingface.co/ornith-ai/Orni…

"rules for thee but not for me"

Mia @MiaAI_lab

2 months ago

Anthropic officially refuses to sign letter supporting Open weight. Instead, they want crackdowns on "industrial distillation", and safety testing for "sufficiently capable" models.

27 21 331 27K 22

Operation Cheepseek: Phase 1 Complete OpenCode Go subscribers now get $30 of usage for $10 Phase 2 initiating...

Introducing Bot Mode for Hermes Desktop. Your agent profiles become a series of named Bots. Each Bot has its own role, model, memory, skills and profile picture; Bots can use any model and even communicate with each other. Build a specialist Bot once to use it forever.

A solo dev made a game about walking his dog Ezio, with his real barks and panting in it, so they could "go on walks forever". 😭 Ezio passed away. Today came the game's final update: a memorial statue and videos of him around the map. The game is called A Walk With My Dog.

@Alibaba_Qwen The entire open source AI community waiting for Qwen-3.8-27b announcement

Hermes Agent can now read anything you throw at it. PDF, Word, PowerPoint, Excel, OpenDocument, RTF, EPUB — all auto-converted to clean Markdown the moment the agent reads the file, all locally! Powered by @firecrawl_dev's new open-source anydoc (Rust). Zero setup, installs itself on first use.

Nicolas Camara @nickscamara_

2 months ago

introducing anydoc now your agents get 100x faster local parsing for pdf, docx, pptx & 10 more formats - sub-5ms md conversion - 500 docx files in 1.7s - top quality across all 13 formats - rust-based - open source already powering @firecrawl /parse github.com/firecrawl/anyd…

293 727 9K 2.5M 15K

Today’s YouTube video goes deeper on how to set up Hermes Agent in Buzz through the gateway. I’ve now got 3 agents on 3 different devices working together! How to Set Up Hermes Agent in Buzz (And Get Agents to Work Together) youtu.be/QnYEeOy86Kg

you might want to run hermes update gents :) we traced wasted turns and tool errors across production conversations and rebuilt the core toolset around what we found. - tool failures now carry recovery hints instead of dead ends - leaner schemas, so less context is spent before your task even starts - file reads handle 4x more per call, fewer round trips on big files - patches diagnose their own misses instead of looping on them - searches that come up empty point the model where to look next - repeat skill loads are deduplicated, a big saving on long sessions smaller & local models now runs better under the harness

Teknium 🪽 @Teknium

2 months ago

Hermes Agent is now dramatically more efficient, especially for smaller/weaker/local models! With the help of @nvidia's Nemo Relay and several other strategies Hermes was able to identify a ton of optimizations beyond just saving tool execution time or memory - but also turns

124 215 2K 245K 971

excellent

Alex Finn @AlexFinn

2 months ago

Buzz from @jack is the future of AI It allows you to command an entire army of AI agents My Codex, Claude Code, Hermes, and OpenClaw agents all work together in one place to do mind blowing work In this video I cover how it works and how you can get 100x productivity from it:

151 136 2K 502K 3K

@steipete Using this to visualise the graphs github.com/pskoett/pmx-ca…