Marcio K

Just open-sourced Gust, our task orchestration system. Like Airflow, but faster and much more efficient, written in Elixir. Check it out! github.com/marciok/gust

@MainzOnX I miss more neuromorphic approaches to AI.

My hot take: scaling has become an overused substitute for algorithmic innovation. We’ve convinced ourselves that parametermaxxing is innovation. It isn’t. Compute should amplify algorithmic breakthroughs, not replace them. AGI won’t come from worshipping bigger matrix multiplies.

This is very interesting! I feel that model orchestration is the next step towards horizontal scaling of models for better intelligence.

@SakanaAILabs This is so cool! Language is not static, and meanings can change from one generation to another. Would be cool to have a translator according to your age/gen :)

Why we built Sakana Translate Blog: #English" target="_blank" rel="nofollow noopener">sakana.ai/translate-rele… Widely used translation tools often miss the mark exactly where Japanese is hardest. They struggle with the nuance of business honorifics, culturally specific concepts, and internet slang. The grammar comes out right, but the tone gets lost. Sakana AI has been researching post-training techniques built around a deep understanding of Japanese language and culture. Our Namazu model series adapts leading global models specifically for Japanese. We realized this strength shines perfectly in translation. Our concept is to translate Japanese deeply. We are not just swapping words and sentence structures. We are carrying across the exact context, tone, and interpersonal distance behind the text.

We released our in-house Japanese, English, and Chinese translation tool, Sakana Translate! Try it→ translate.sakana.ai 🐟 • Translate: Handles long text in real time • Proofread: Tone and phrasing refined, with tracked changes • Ask: Nuanced word choices clarified

@0xSero @grok x.com/i/grok/share/b…

@0xSero @grok organize and list the book titles mentioned in this post.

I’m trying to level up my world knowledge. I’m looking for books that will blow my head wide open. What would you recommend?

@0xSero The Infinity Machine: Demis Hassabis, DeepMind and the Quest for Superintelligence

@Mayhem4Markets 🤞

@MarcioK Heard that back when Linux was first becoming a force in the world of servers. Didn't age well then, don't think it will age well now, either.

🌀 Gust v0.1.35 is out!! 🌀 > `wait_for` option for HITL DAGs. > API & MCP support for resuming "waited" tasks. > Many more fixes and improvements github.com/marciok/gust Ex: 👇

@philipkiely @aiDotEngineer Thanks for still writing books at the age of token generation!

Its Inference Day at @aiDotEngineer World’s Fair! Catch my talk at 1:30 on the Inference Track (room 2016). I’ll cover what has happened in inference since I published the book. And grab your copy of Inference Engineering at the Baseten booth.

@pmarca Eat less to save food.

Who is John Galt?

@0xSero @jun_song For third-world countries, 1K is still a lot. Unfortunately...

@jun_song 1k is the target.

A year ago : we need a $100k hardware to run them 6 months ago : we need a $50k hardware to run them Today : we need a $15k hardware to run them And they are still complaining.

Same here! I enjoy it so much that I've created my own automation tool 🙂 github.com/marciok/gust

@elliotarledge Benchmark (and mix) with my prompt fusion 🙃 github.com/marciok/gust/b…

What are YOU planning on cooking with Fable when its back?

@rasbt The pleasure of opening a new book is unmatched.

@MarcioK Thanks! Writing & reading are two of my favorite activities... not sure what I would do without one day 😅

After 18 months of writing, coding, and experimenting, Build a Reasoning Model (From Scratch) is finally out! My first copies just arrived! 📚 440 full-color pages. Inference scaling, reinforcement learning, and distillation from scratch.

@LottoLabs I have a solution: Rename data centers to Token Farms.

Don’t let data centers have the same fate as nuclear

@dasha_shunina @aiDotEngineer github.com/marciok/gust -> Gust: Scalable, observable, and durable task orchestrator: Superpower for agentic recurrent tasks. -> To discuss: When should we rely on deterministic workflows, and when should we let non-deterministic agents take over?

🚨 Giving away 1 FREE ticket to @aiDotEngineer World's Fair. To enter, reply with: What you're building (or want to build) with AI Why you want to be at AI Engineer World's Fair We'll pick 1 winner from the replies.

@hardmaru I am Brazilian, but let's go Nippon!