A Year of LLM Lessons: Tactics, Operations, Strategy
Eugene Yan's retrospective distills practical lessons from a year of building with LLMs across tactical, operational, and strategic levels.
AI & ML
Frontier models, research results, agents, and developer tooling.
A Curated Reading List for Language Modeling Papers
Eugene Yan's reading list offers a practical starting point for groups wanting to study foundational language model papers.
Hugging Face Launches AnyLanguageModel for Apple Platforms
A new unified API aims to reduce friction for developers working with both on-device and cloud-based LLMs on Apple platforms.
DeepMind Brings AI Image Verification Into the Gemini App
Google DeepMind is integrating image verification directly into the Gemini app, giving users a clear signal when an image is AI-generated or AI-edited.
DeepMind opens Singapore lab to boost Asia-Pacific AI research
The new research center signals a deeper commitment to regional AI talent and partnerships.
Google Research Details Real-Time Speech-to-Speech Translation
A new blog post outlines algorithmic work aimed at reducing latency in direct speech-to-speech translation.
AI Distinguishes Natural Forests From Plantations for Supply Chains
Google Research's AI model helps companies verify deforestation-free sourcing by separating natural forests from other tree cover.
SIMA 2 Brings Gemini-Powered Agents Into Virtual 3D Worlds
Google DeepMind's SIMA 2 points to a future in which AI agents do not just generate text but act, reason, and learn inside interactive 3D environments.
Hugging Face Announces OpenEnv for Open Agent Ecosystem
Hugging Face is inviting the community to help build shared infrastructure for AI agents through a new initiative called OpenEnv.
Sentence Transformers Joins Hugging Face
The announcement on Hugging Face's blog confirms that the Sentence Transformers project is becoming part of the company.
Google Research Unveils Nested Learning for Continual AI
A new machine-learning paradigm from Google Research aims to improve how models handle sequential tasks.
Gemini 3 Announced in Twin Google DeepMind Blog Posts
Two new posts from Google DeepMind cover the same product launch but frame it differently: one for developers, one for the broader AI trajectory.
Google Research Turns Prompts into Custom Interactive User Interfaces
A new Google Research approach generates rich, visual, and interactive UI experiences directly from user prompts, moving beyond text-only AI responses.
Google Research Unveils Quantum Toolkit for Optimization Problems
New tools aim to bring quantum methods closer to practical use in solving difficult optimization tasks.
Google Research details AI health coach design
Google Research explains how it is building a generative-AI personal health coach and the reasoning behind the approach.
DeepMind Paper Maps How AI Visual Organization Diverges From Humans
New research from Google DeepMind examines the gap between machine and human visual organisation, pointing toward more human-like AI perception.
Hugging Face Adds ROCm Kernel Build and Share Workflow
Hugging Face is making it easier to build and share custom ROCm kernels for AMD GPUs.
AI pilot gives Northern Ireland teachers 10 extra hours a week
A six-month pilot with Google's Gemini found teachers gained an average of 10 hours weekly through AI-assisted tasks.