Marc Tarnutzer

Building @MnemeAI

[Google Gemma @googlegemma]

Want to teach Gemma to master chess?

Check out this awesome community project showing how to fine-tune Gemma 4 12B on your own data, 100% locally!

Running text, images, and audio on just 8GB VRAM makes custom models more accessible than ever.

[Marc Tarnutzer @MarcTarnutzer]

interesting

[Joel - coffee/acc @JoelDeTeves]

I accidentally discovered that Gemma-4-26B-A4B is way better at writing human sounding content than every other model out there - including frontier models like GPT 5.5 and Sonnet 4.6.

I'm not sure why this is - it's kind of crazy how slopified these big expensive models are and for some reason, Google's open source model sounds a lot more natural and follows writing instructions better.

WTF?

[Marc Tarnutzer @MarcTarnutzer]

@unusual_whales now release mythos

[Marc Tarnutzer @MarcTarnutzer]

interesting

[Awni Hannun @awnihannun]

It's very cool that Apple shipped a 20B parameter on-device.

You can't put 20B parameters in RAM at any reasonable precision. To make it work they are using pretty exotic architecture by today's standards.

A small model predicts from the query (or prompt) which experts to load from Nand into RAM. The key distinction from a typical MoE is that you do this once per query and then generate all the tokens with the same experts (instead of switching the experts for every token).

[Marc Tarnutzer @MarcTarnutzer]

great, i missed those!

[Andreas Storm @avstorm]

The Mickey Mouse pointers are back in macOS 27

[Marc Tarnutzer @MarcTarnutzer]

great work!

[Omar Sanseviero @osanseviero]

Gemma 4 MTP just got officially merged into llama.cpp

This means you can use Gemma 4 QAT + MTP for a lightweight + super fast setup. Excited to see what the community builds with it

https://t.co/1te7tgdi2H

[Unsloth AI @UnslothAI]

Google releases Gemma 4 QAT. ✨
You can now run Gemma 4 at 3x less memory with near original performance.

Quantization-Aware Training (QAT) makes it possible to run Gemma 4 26B-A4B on 16GB RAM.

GGUFs: https://t.co/wQgEocxUId
QAT Guide: https://t.co/Nsm1yeGEHx

[antirez @antirez]

Opus 4.8 is a total disaster. The problem is not the model per-se, they have Mythos and can anyway train a better model. The problem is: what is it happening inside Anthropic, at the management level? Since this is a product failure. If there was a technological issue NOT delivering is better.

[Google Gemma @googlegemma]

Meet Gemma 4 12B!

A unified, encoder-free multimodal model designed to bring high-performance intelligence directly to your laptop, and released under an Apache 2.0 license.

Bridging the gap between edge efficiency and advanced reasoning. Here is what’s new with Gemma 4 12B: 👇

[Marc Tarnutzer @MarcTarnutzer]

local ai is the future

[Cult of Mac @cultofmac]

Apple’s next-gen Siri could rely heavily on local AI https://t.co/hW6j7ervHx

[Marc Tarnutzer @MarcTarnutzer]

great new album

[BOARDS OF CANADA @boardsofcanada]

INFERNO, OUT NOW.
https://t.co/r6yNo3gkJN

[Daniel @growing_daniel]

This is what happens when you don’t pay down your technical debt

[Marc Tarnutzer @MarcTarnutzer]

working on some exciting updates for mneme ai, stay tuned

[Marc Tarnutzer @MarcTarnutzer]

one of the most important projects in ai right now

[Georgi Gerganov @ggerganov]

llama.cpp now has an official website: https://t.co/vztdUpdBWL

Our goal is to make local AI accessible to everyone, and improving the user experience is a big part of that. On the new landing page you’ll find a single-line cross-platform installer. The installation provides a single unified `llama` entrypoint which you can use to run/serve models and interface with 3rd-party agentic applications.

While oriented towards simplified user experience, the new `llama` application also provides all the advanced functionality of the existing llama.cpp tooling with which experienced users are already familiar. Also note that all GGUF models that you might have already downloaded with llama.cpp in the past will be automatically available to use without downloading again (they are stored in the common HF cache on your machine).

We have many improvements in the pipeline both at the UX and at the engine level and we plan to iteratively ship new things over the coming months. One of the main focuses will be seamless integration with local-friendly 3rd-party agents (such as Pi). In the meantime, we’ll continue to listen for feedback from the community and adjust accordingly, so keep letting us know what you think and need.

[François Chollet @fchollet]

Einstein on (not) using NL for invention: "The words or the language, as they are written or spoken, do not seem to play any role in my mechanism of thought"

[Lucas Maes @lucasmaes_]

Would you like to join the research effort on JEPA and World Models easily?

After a full year of hard work, we’re excited to finally release stable-worldmodel:

an open-source, scalable platform built to accelerate JEPA & World Model research!

📄: https://t.co/gnxGvens5A

[Marc Tarnutzer @MarcTarnutzer]

great interview

[Loris Cro ⚡ @croloris]

Andrew's interview with JetBrains

https://t.co/7SsmfjksJP

[Marc Tarnutzer @MarcTarnutzer]

Version 1.4 of Mneme AI adds iPad support and is out now! This is a first version for the iPad, so the design will probably change in the future.

[Marc Tarnutzer @MarcTarnutzer]

Just implemented iPad support for Mneme AI, update coming soon! The AI speed will be insane on the newer iPad models.

[Marc Tarnutzer @MarcTarnutzer]

Good morning, Mneme AI version 1.3 just released with new control center, lock screen and action button actions (on iOS 18) and a new widget! Please let me know what features you want to see next!