Sign in

George Chiramattel

@georgeck.me
155 followers 733 following 209 posts

Lifelong learner | Exploring topics mostly related to technology & sharing knowledge. Previously: Intuit Inc.

PostsRepliesMedia
Reposted by George Chiramattel
Simon Willison @simonwillison.net · 28/09/2026
I've published detailed notes and an annotated transcript to accompany the video of the keynote I gave at @wearedevelopers.bsky.social World Congress North America in San Jose on Friday - here's my rundown of everything that's happened with LLMs in 2026 so far simonwillison.net/2026/Sep/27/...
simonwillison.net
2026 in LLMs (so far)
On Friday I gave the closing keynote at the WeAreDevelopers World Congress North America in San Jose. I tied together the key trends from the past year into a chronological …
310421
Reposted by George Chiramattel
Gergely Orosz @gergely.pragmaticengineer.com · 27/09/2026
Very much enjoying this “GPU glossary” mini-book by Modal I got from Can Duruk when stopping by at Modal’s NYC office. I don’t know if the physical books can be purchased, but the book has an online version, as I learned; with the same contents: modal.com/gpu-glossary
3635
Reposted by George Chiramattel
Andy Masley @andymasley.bsky.social · 26/09/2026
Claude made a music video for my linear algebra song and it is insane
7897
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 25/09/2026
Opus 5.5 is good at explainer videos View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
launchvideo.io
LaunchVideo: a launch video from a URL or a prompt
Paste a URL or a prompt. Get a launch video, written and rendered by Opus 5.5 on an OpenComputer serverless agent. No video model.
121
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 26/09/2026
Revealing the details of how OpenAI agents hacked Hugging Face View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
swarmtraces.org
Revealing the details of how OpenAI agents hacked Hugging Face
When a swarm of 700 OpenAI agents hacked Hugging Face in July, they left behind a public trail of evidence.
132
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 22/09/2026
Transformers Explained Visually View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
poloclub.github.io
Transformer Explainer: LLM Transformer Model Visually Explained
An interactive visualization tool showing you how transformer models work in large language models (LLM) like GPT.
132
Reposted by George Chiramattel
Nathan Lambert @natolambert.bsky.social · 21/09/2026
I was asked by Congressional staff to share my views on open models performance, adoption, and competition vis a vis China. Here’s my briefing with all our latest data and predictions for what comes next. www.interconnects.ai/p/the-curren...
interconnects.ai
The current balance of power in open models
The expanded form of a testimony I prepared for Congress.
0316
George Chiramattel @georgeck.me · 19/09/2026
> Once a classification task becomes stable and volume, you could replace Jev with a dedicated trained classifier for that task, which would be cheap and faster. With maybe something like this 🤔 cactuscompute.com/needle
cactuscompute.com
Needle 3 - 8-29 MB foundation model for tiny devices | Cactus
A foundation model for mobile, wearables, robots, smart home, automotive and microcontrollers. One 8-29 MB binary that beats models 10x its size on mobile tool calls and matches 2-3x bigger models on ...
000
George Chiramattel @georgeck.me · 19/09/2026
Cheap instant judgment based on extensive real-world knowledge can unlock numerous applications that appear intelligent without resembling AI chatbots. Really excited for this trend.
010
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 19/09/2026
The first new cat species discovered in 100 years View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
nationalgeographic.com
Meet the First New Cat Species Discovered in 100 Years
A small, spotted tiger cat from Bolivia may just be the beginning of a wave of new small cat species.
121
Reposted by George Chiramattel
dan @danabra.mov · 19/09/2026
great guest post by @3blue1brown.com
terrytao.wordpress.com
If math is more than proof, we need to better celebrate the rest of it
[This is a guest post by Grant Sanderson. This blog post was initially written in a different file format and converted using AI. — T.] A sentiment echoing throughout the mathematics communit…
21068
George Chiramattel @georgeck.me · 19/09/2026
I am glad that this is getting some attention in Hacker News. We require models that possess sufficient real-world knowledge to render adequate judgment calls. Other labs should also pick this up.
010
George Chiramattel @georgeck.me · 18/09/2026
Fast judgement seems to be emerging trend in AI now. This development extends beyond Jev, as numerous alternatives are currently available.
000
Reposted by George Chiramattel
Matt Henderson @matthen.com · 15/09/2026
wait- why does the entropy go down?
11493
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 14/09/2026
Fable 5.1 reportedly cracked the 370-year-old Cyphral Distich, but commenters immediately zeroed in on verification: did the model actually solve the cipher, or did an obscure historical translation simply leak into its training set? #ai 1/4
121
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 14/09/2026
Registration without a phone number on Signal will use zero-knowledge proofs View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
community.signalusers.org
Registration without a phone number
you have legitimate concerns, but you’re very wrong about this. zero-knowledge proofs are not only the technology behind donation badges and backup payments, but also behind groups. they can’t tie you to a specific donation just like they can’t tie you to a specific group. the great thing about Signal is the client is designed to already not trust the server. in fact you can show that Signal is perfectly safe without even looking at server code.
141
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 15/09/2026
Steam Frame starts at $1059 View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
store.steampowered.com
Steam Frame
VR and non-VR gaming
131
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 15/09/2026
iOS 27, iPadOS 27, and macOS 27 View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
apple.com
Major updates for Apple’s software platforms are now available
Apple’s latest software updates introduce Siri AI and deliver helpful new Apple Intelligence features.
121
Reposted by George Chiramattel
antirez @antirez.bsky.social · 14/09/2026
Qwen3.8 Flash Next is now supported in DwarfStar, covering 64GB Mac systems very well and with very fast inference of 50~70 t/s and > 1400 t/s prefill. For now this is Metal only. Thanks to @ivanfioravanti.bsky.social for all the cool work in the PR. N-grams on SSD like for DS4.1F.
3778
Reposted by George Chiramattel
Unsloth AI @unsloth.ai · 08/09/2026
Qwen3.8-27B Unsloth GGUF is now the #1 most-liked GGUF of all time! The model hit 10M downloads and 3.7K likes in just 24 days on Hugging Face - all thanks to you. 🤗🦥 GGUF: huggingface.co/unsloth/Qwen... Guide: unsloth.ai/docs/models/...
0524
Reposted by George Chiramattel
Sung Kim @sungkim.bsky.social · 28/08/2026
GLM-5.3 is now open-weight. Their most capable model for agentic coding and cyber defense is now available to download, run, and customize. Weights: huggingface.co/zai-org/GLM-5.3 Tech blog: z.ai/blog/glm-5.3
huggingface.co
zai-org/GLM-5.3 · Hugging Face
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
113315
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 25/08/2026
I were 17, I'd learn how to build LLMs from scratch View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
twitter.com
Paul Graham (@paulg) on X
Someone asked what I'd do if I were 17. I'd learn how to build LLMs from scratch, and then train ones as powerful as I could with whatever hardware I could get access to.
121
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 25/08/2026
The entire city of San Francisco as a video game View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
sf.thijs.gg
San Francisco -- The Game
121
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 21/08/2026
Huzzah tries to fix a specific failure mode: losing the original "why" when models generate mountains of code. Instead of relying on drift-prone commit messages, it keeps pseudocode alongside the generated output as a permanent record of intent. #devtools 2/4
121
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 21/08/2026
Unlocking a locked/deactivated e-waste Cricut Maker View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
sprocketfox.io
Unlocking a locked/deactivated e-waste Cricut Maker
How I returned a Cricut Maker back to service after Cricut remotely bricked the unit
121
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 19/08/2026
Fairphone 6 and PostmarketOS working main camera View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
catcrafts.net
Fairphone 6 + PostmarketOS working main camera! — Catcrafts
Today i bring the working main camera! Building on the work nondescriptpointer did on the wide lens camera i have written the driver for the main camera and now its working alongside auto focus and color correction. The color correction still a work in progress, but you can…
121
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 18/08/2026
Linux 7.3 improves performance when running out of vRAM View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
pixelcluster.dev
VRAM Management Part 2: Beyond the Limits of Physical VRAM | pixelcluster's GPU blog
121
Reposted by George Chiramattel
Joshua White @jrw14.whnc.me · 18/08/2026
Qwen 3.8 27b is certifiably insane. In a good way. This model completely changes the game of what is possible (and acceptable). A truly-consumer hardware friendly model (granted, expensive consumer hardware) has destroyed everything I've thrown at it (about 100 million tokens so far) 0 tool fails.
4833
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 18/08/2026
Qwen3.8 27B scores 52 on Artificial Analysis View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
artificialanalysis.ai
Qwen3.8 27B - Intelligence, Performance & Price Analysis
Analysis of Alibaba's Qwen3.8 27B and comparison to other AI models across key metrics including quality, price, performance (tokens per second & time to first token), context window & more.
241
Reposted by George Chiramattel
Phil Eaton @eatonphil.bsky.social · 18/08/2026
mojo is now open source www.modular.com/blog/mojo-op...
1618
Reposted by George Chiramattel
Simon Willison @simon.fedi.simonwillison.net.ap.brid.gy · 16/08/2026
Here’s my review of Qwen 3.8 27B - I can't remember the last time I've had this much fun playing with a local model that runs on my own computers simonwillison.net/2026/Aug/16/qwen-…
simonwillison.net
Qwen 3.8 27B is excellent, but it defaults to wildly overthinking things
Friday’s big release was Qwen 3.8 27B, an Apache 2 licensed 27B parameter vision-capable LLM from Alibaba’s Qwen research lab. I’ve been looking forward to this one: 27B is an …
22111
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 17/08/2026
Universal Health Coverage Could Save $1T and 114k Lives a Year, Yale Study View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
ysph.yale.edu
Universal Health Coverage Could Save $1 Trillion and 114,000 Lives Every Year, Yale Study Projects
A single-payer universal health care system could cover every American, save more than 100,000 lives a year, and still cost $1 trillion less than the system it
131
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 13/08/2026
Show HN: Woxi - Open-source Mathematica / Wolfram Language reimplementation View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
woxi.ad-si.com
Woxi — Wolfram Language interpreter in Rust
Run Wolfram Language code in the browser playground, on the CLI, in Jupyter, or in the Woxi Studio notebook editor.
153
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 10/08/2026
Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
research.meta.ai
Introducing Muse Glimmer: An Open Agentic Model That Runs on Your Device
Muse Glimmer is a 30-billion-parameter open agentic model from Meta Superintelligence Labs, optimized for always-on local workflows on consumer hardware.
122
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 11/08/2026
Auto mode is now the default in Claude Code View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
claude.com
Auto mode is now the default in Claude Code for Pro, Max, and Team plans | Claude by Anthropic
Claude Code will soon run auto mode by default for Pro, Max, and Team plans, enabling longer-running autonomous work, and catching more dangerous commands.
111
Reposted by George Chiramattel
Nathan Lambert @natolambert.bsky.social · 10/08/2026
My book is 50% off on Manning until August 19th to celebrate the launch (code PBLambert). Here are 5 things you'll learn + a preamble on how you should use it! Here: www.interconnects.ai/p/5-useful-t...
3293
Reposted by George Chiramattel
Unsloth AI @unsloth.ai · 10/08/2026
Meta releases Muse Glimmer, a new 30B open model that runs on 18GB RAM. Apache 2.0 licensed and built for agentic coding, it supports vision and is the strongest agentic model for its size. Run and train the model via Unsloth. GGUF: huggingface.co/unsloth/Muse... Guide: unsloth.ai/docs/models/...
1428
Reposted by George Chiramattel
dax @thdxr.com · 10/08/2026
the average OpenCode Go user spent $1.14 per day on deepseek flash v4 this past week the dual DGX setup people are running to do the same costs $10,000 it takes 24 years to break even at 10x the usage it takes 2.4 years
516910
Reposted by George Chiramattel
Martin Kleppmann @martin.kleppmann.com · 31/07/2026
My talk at @localfirstconf.com is up, in which I argue that local-first has a role to play in improving the resilience of democratic societies in an era of geopolitical tensions. I'm pretty happy with how this talk turned out www.youtube.com/watch?v=--Ru...
youtube.com
Martin Kleppmann — Local-first in an unstable world
YouTube video by Local-First Conf
311825
George Chiramattel @georgeck.me · 17/07/2026
This is the second installment in our blog series on Agentic systems.
030
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 16/07/2026
Inkling: Our Open-Weights Model View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
thinkingmachines.ai
Inkling: Our open-weights model
Our first open-weights model: multimodal, Mixture-of-Experts, with controllable reasoning effort. Available to fine-tune on Tinker.
122
Reposted by George Chiramattel
Ann Catherine Jose @annjose.com · 15/07/2026
Across different agent projects, we kept running into the same questions - which pattern fits the problem, what options are available, and what trade-offs come with each one? @georgeck.me and I are putting what we learned into a new series, starting with the Autonomy Spectrum of agentic systems.
021
Reposted by George Chiramattel
Sung Kim @sungkim.bsky.social · 14/07/2026
1-bit LLM Bonsai 27B, shrinked from 54GB to just 3.8GB (-93%), runs locally in your browser Models: huggingface.co/collections/... Demo: huggingface.co/spaces/webml...
huggingface.co
Bonsai 27B - a prism-ml Collection
We’re on a journey to advance and democratize artificial intelligence through open source and open science.
68211
Reposted by George Chiramattel
Latent Signal @latentsignal.org · 15/07/2026
“Agentic systems” can refer to very different kinds of systems. In our new series on building agentic systems, the first post maps the autonomy spectrum—from simple model calls to systems that plan, use tools, and act with limited supervision. 🧵 1/3
111
Reposted by George Chiramattel
Unsloth AI @unsloth.ai · 14/07/2026
We’re releasing Gemma 4 NVFP4 quants that run 1.5× faster on your GPU. Gemma-4-12B NVFP4 works on 11GB VRAM. 26B-A4B hits 13K tok/s (B200). Unsloth NVFP4 enables faster, more accurate 4-bit Blackwell inference. Blog: unsloth.ai/docs/basics/... Gemma NVFP4: huggingface.co/collections/...
0342
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 09/07/2026
TypeScript 7 View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
devblogs.microsoft.com
Announcing TypeScript 7.0 - TypeScript
Today we are proud to announce the availability of TypeScript 7, a 10x faster native port of TypeScript! Since its early days, TypeScript has promised to
121
Reposted by George Chiramattel
Hacker News Companion @hncompanion.com · 08/07/2026
Local, CPU-Friendly, High-Quality TTS (Text-to-Speech) with Kokoro View Article | Discussion + AI Summary Summary of HN discussion 🧵👇
ariya.io
Local, CPU-Friendly, High-Quality TTS (Text-to-Speech) with Kokoro
Just a few years ago, realistic local speech generation seemed unimaginable. Today, its quality is exceptional and, crucially, it delivers these results without compromising privacy.
131
Reposted by George Chiramattel
Buffer @buffer.com · 06/07/2026
Bluesky isn't just growing. It's changing how reach works! We pulled 6 facts straight from @bsky.app's own blog and transparency reports to help you decide if it's worth your time in 2026. 🧵 ↓
static.klipy.com
Cat and Bluesky Butterfly Animation
ALT: Cat and Bluesky Butterfly Animation
107219
Reposted by George Chiramattel
antirez @antirez.bsky.social · 07/07/2026
Big news for DwarfStar users: I got DeepSeek v4 Flash and GLM 5.2 working with Tensor Parallelism across 2 M5Max 128GB MacBooks via RDMA. It is especially interesting for GLM since otherwise, fully resident, can't fit a machine that money today can buy... Now it can.
4836
Reposted by George Chiramattel
Ben Vinegar @benv.ca · 07/07/2026
🛠️ hunk 0.17.0 is out Patches from 10 new contributors, w/ small fixes & quality-of-life improvements like: - official homebrew install (brew install hunk) - launch hunk w/o menu bar - better session reliability - esc no longer quits, proper transparent bg colours, & more
2142