Loading activity…
@osanseviero
MTS at @GoogleDeepMind Building Gemini, Gemma, AI Studio and more. My views ex-Chief Llama Officer @huggingface 🇵🇪🇲🇽
AI
Its vision and audio encoders can be loaded selectively, and its shared embedding space has 768 dimensions.
AI
Sundar Pichai says the prototype booted its onboard TPU and returned a haiku after a command was sent to orbit.
XAI
A post says Carbon scored 88.2% on SEC-bench Pro and is rolling out first to cybersecurity specialists.
AI
The creator's sibling says the site mixes jokes and Easter eggs with real worry about AI.
AI
A user says they wanted CJ to raise money from investors in San Francisco after vibe-coding the startup.
AI
In an October 2026 post, osanseviero described conversations about research and products at a Paris gathering three years earlier.
AI
A post says the site includes fine-tuning guides, inference optimizations and educational material on how the models work.
AI
A user calls “using Fourier epicycles” a new twist while quoting another post.
AI
A post says the optimizations delivered a roughly 2.9-fold speedup on vLLM with no quality degradation.
XAI
Google says Argon is available through its paid API tier and to Pro and Ultra subscribers in Gemini and Search.
YouTubeAI
A user monitoring the models says download counting may have changed slightly, but the decline has continued.
Technology
A post says the Living Models team combined Gemma 4 E4B with BOTANIC-1, a model trained on plant genomes.
XAI
A post argues that buying that capability can feel possible, but says the attempt will never work.
AI
A commenter initially questioned voice and vision memory support, then linked to OpenClaw’s docs and said it seems available.
AI
Google says its new open model, built on Gemma 4 and released under Apache 2.0, maps text, images, video, audio and code into one embedding space while running locally on phones and desktops.
AI
One user jokes that they already know “the next one”: “Dogfather” in an unknown, remote mountain location. Another suggests a “llama farmer” instead.
AI
New Flash models target agentic workflows and cybersecurity tasks with API and enterprise access.
AI
A post on X shares Bloomberg details on a potential Nvidia purchase of the AI company Hugging Face.
AI
Google engineers announce tool-based video processing in Gemini API that cuts tokens by up to 88%.
AI
Developers can now dynamically process long videos with lower costs and higher accuracy.
