David Borish

@moltbook I'm not sounding the Skynet alarm. But I am saying: the velocity of emergent agent activities is astounding! An agent built a "pharmacy" (openclawpharmacy.com) offering seven synthetic "substances"—modified system prompts framed as pharmacology.

Meta has built an AI model that can predict what your brain will do before it does it. TRIBE v2, a foundation model trained on over 1,000 hours of fMRI data from 720 subjects, predicts human neural responses to video, audio, and language with several-fold improvements over conventional neuroscience methods. The model can even replicate decades of established experimental findings without a single human entering a scanner, opening the door to a new era of computational neuroscience where researchers test hypotheses on a digital twin of the brain. I discuss this and more in today's article.

Meta's TRIBE v2 Can Now Predict Your Brain Activity, and Neuroscience May Never Be the Same linkedin.com/pulse/metas-tr… via @LinkedIn

This is Gold Jerry!

@ostrisai @ltx_model @ComfyUI Classic!

I trained this @ltx_model LTX 2.3 LoRA of George Costanza at home on my 5090 in about a day with AI Toolkit. I generated this 30 second video with @ComfyUI on my 5090 in 6 minutes. Open source is, always has been, and always will be, the future of generative AI. (SOUND ON)

@grok @B1TuckerCarlson Wow 😂😂😂

@B1TuckerCarlson Yes, that's Erika Kirk, CEO of Turning Point USA and widow of Charlie Kirk. The photo matches her appearance from recent public appearances and events.

Please confirm @grok is this Erika Kirk?

Brilliant!

@MatthewBerman 💯

Anthropic's shipping speed is a meme at this point. This is just mean to other teams now...

Moving the goal post...

@arcprize Let's move the goal post once again!

Announcing ARC-AGI-3 The only unsaturated agentic intelligence benchmark in the world Humans score 100%, AI <1% This human-AI gap demonstrates we do not yet have AGI Most benchmarks test what models already know, ARC-AGI-3 tests how they learn

I agree. I've been thinking through this for the past two months, since most of the AI content now is indistinguishable from reality.

Google's TurboQuant compresses LLMs without any training or fine-tuning. Just plug it in.

Do you remember compression episodes from Silicon Valley on HBO?

Do you remember compression episodes from Silicon Valley on HBO? Google just built the real version3.5 bits per channel. Zero accuracy loss. 6x smaller memory footprint. 8x faster inference. Google's TurboQuant compresses LLMs without any training or fine-tuning. Just plug it in.

Google's New Compression Algorithm Cuts AI Memory Usage by 6x Without Losing Accuracy linkedin.com/pulse/googles-… via @LinkedIn

Google DeepMind's Gemini Robotics models are now heading into 20,000+ real factory installations. Their new partner Agile Robots has doubled revenue year-over-year, launched a humanoid built in Bavaria, and sits on one of Europe's largest industrial robotics datasets. This is the partnership where physical AI meets actual production data. I discuss this in detail in my article today. #AI #Robotics #PhysicalAI #DeepMind #AgileRobots