Intelligence.Log

2026-04-07

Extracted: 15 items. Sources: GitHub, Bluesky, Blogs.
++ AI OVERVIEW ++
Today's tech scene is dominated by advanced AI accessibility and a tense debate over secrecy. On GitHub, the trending `mempalace` project offers a free, benchmark-leading AI memory system, while `gemma-tuner-multimodal` enables fine-tuning of Gemma models with audio and images on Apple Silicon. Meanwhile, AI leaders on Bluesky are critically discussing opaque security practices, highlighted by Anthropic's restricted "Project Glassing" and nostalgic posts about past model release controversies. This unfolds against a surreal backdrop of the historic Artemis II mission, juxtaposing human achievement with ongoing ethical and security concerns in AI.
grep TOPIC=
grep SOURCE=
GH

Fine-tune Gemma 4 and 3n with audio, images and text on Apple Silicon, using PyTorch and Metal Performance Shaders.

Starred bysimonw|[Fine-tuning][Multi-modal][Deployment]
"This project enables fine-tuning of Gemma models (versions 4 and 3n) with multimodal data including audio, images, and text, specifically optimized for Apple Silicon hardware using PyTorch and Metal Performance Shaders. It provides a practical solution for running advanced multimodal fine-tuning on consumer Apple hardware."
GH

The highest-scoring AI memory system ever benchmarked. And it's free.

Starred bysimonw|[LLM][Tooling]
"MemPalace is a free, high-performance AI memory system that achieved top benchmark scores for memory retention and recall. It integrates with ChromaDB for vector storage and supports MCP (Model Context Protocol) for enhanced LLM interactions."
GH

Fine-tune Gemma 4 and 3n with audio, images and text on Apple Silicon, using PyTorch and Metal Performance Shaders.

Starred byminimaxir|[Fine-tuning][Multi-modal][Deployment]
"This project enables multimodal fine-tuning of Gemma models (4 and 3n) with audio, image, and text data on Apple Silicon hardware. It leverages PyTorch and Metal Performance Shaders for optimized performance on Apple's M-series chips."
BSKY
simonwillison.netSimon Willison

Wrote up some thoughts on Anthropic's Project Glassing, where their latest Opus-beating model is available to partnered security research organizations only. Given the recent alarm bells raised by credible security voices I think this is a justified decision. simonwillison.net/2026/Apr/7/p...

❤️ 163 Likes|[Safety][Deployment]
BSKY
simonwillison.netSimon Willison

I'm a big fan of the pelican GLM-5.1 drew me today, it even animated it! simonwillison.net/2026/Apr/7/g...

❤️ 97 Likes|[LLM][Multi-modal]
BSKY
markriedl.bsky.socialMark Riedl

*fondly remembers when GPT-2 was too dangerous to release*

❤️ 91 Likes|[Safety]
BSKY
sharky6000.bsky.socialMarc Lanctot

Ok so it happened to me: paper cites our work with hallucinated author list and incorrect year. (Luckily it's just arXiv so hopefully fixed by an email.) But I am 🤬🤬🤬. This is not ok! The paper also seems quite related to our work so I was interested at first; now I can't trust any of it. 😭

❤️ 6 Likes|[Evaluation]
BSKY
emollick.bsky.socialEthan Mollick

Oh no.

❤️ 160 Likes|
BSKY
emollick.bsky.socialEthan Mollick

SuperClaude (Mythos) still seems irreducibly Claude-y given the transcripts in the system card. Here two versions of Mythos are forced to talk to each other across multiple rounds. They are less philosophical than Opus 4.6 or spiritual than Opus 4.1, but still very Claude-like in personality.

❤️ 60 Likes|[LLM][Evaluation]
BSKY
emollick.bsky.socialEthan Mollick

I was told about the Mythos release, but have no personal experience to add. Two points from brief: 1) It is not built for IT software security, it is just a good enough model that it is good at that too 2) This is the first, not last, model to raise security risks red.anthropic.com/2026/mythos-...

❤️ 100 Likes|[Safety][LLM]
BSKY
emollick.bsky.socialEthan Mollick

I suspect that popularity of AI is going to start looking like surveys where people trust their own doctors but are distrustful of the medical establishment People will increasingly like and rely on “their AI” but will increasingly be anxious about “AI” as a category. Some odd implications result.

❤️ 117 Likes|[Safety]
BSKY
angelamczhou.bsky.socialangela zhou

brutal to read cover letters from revisions that sound like they are written by chatgpt 🫠 putting a lot of my cover letters to sound like they are written by a Human, even if this Human is extremely rambly and longwinded. That's artisanal ramble, thank you very much

❤️ 9 Likes|[LLM]
BSKY
beenwrekt.bsky.socialBen Recht

A little more than a decade ago, a bunch of universities decided they needed to start schools of computing. What are the great success stories of this movement?

❤️ 9 Likes|
BSKY
beenwrekt.bsky.socialBen Recht

Pulling from this week’s lecture and op-ed on silicon sampling, I map the territory of simulation and its many purposes.

❤️ 16 Likes|[Evaluation]
BLOG

<p>Anthropic <em>didn't</em> release their latest model, Claude Mythos (<a href="https://www-cdn.anthropic.com/53566bf5440a10affd749724787c8913a2ae0841.pdf">system card PDF</a>), today. They have instead made it available to a very restricted set of preview partners under their newly announced <a...

By Simon Willison
"Anthropic is taking a cautious approach with Claude Mythos by restricting access to security researchers through Project Glasswing, rather than releasing it publicly. This reflects growing industry awareness of AI safety risks and the need for controlled testing before broader deployment."
-- END OF LOG --
[STATS] 15 items · Filter applied
Powered by Horizon + DeepSeek