#LLM
Curated coverage on «LLM»: releases, research, benchmarks, open-source.
Nvidia in talks to buy Reflection AI or raise its stake, FT reports
Financial Times: Nvidia is in early talks to acquire open-model developer Reflection AI or to invest more in it. The deal could fall through.
Japan orders tighter cyber defenses after data leaks affecting millions of customers
Japan's National Cybersecurity Office told agencies, local governments and companies to tighten defenses after a wave of attacks exposed the data of millions of…
Microsoft Decision-1: the model that picks answers from options 35x faster than GPT-6 Sol
Microsoft has released Decision-1, built on Alibaba's Qwen3.5-9B. The model picks from fixed options, running 2.5x faster than H2O-Lightning-4B v1.1 and 35x faster…
Booz Allen Hamilton's AI Agents Succeeded in 8 of 8 Test Attacks on an Industrial System
Karpathy Estimates Only About 5,000 People Have Access to the Most Advanced AI
SIA: China Supplies the US Just 5% of Chip Equipment, but Over 90% of Some Rare Earths
Satya Nadella Urges Companies to Treat AI Models as an Insider Risk and Install an "Emergency Brake"
Yandex Blocks Some 360 and Disk Features After Drone Attack on Third Data Center
After a drone attack on its third data center, Yandex shut off some Yandex 360 and Disk features in an emergency move to keep its infrastructure from critical…
Asbis bets on data centers and AI to sustain a 476% stock rally
Warsaw-listed electronics distributor Asbis Enterprises is shifting sales toward data centers and AI infrastructure as it tries to hold a 476% share-price rally.
There Are More Programmers Now, Even Though AI Promised to Replace Them Within a Year
AI Solved 370 Math Problems, Mathematicians Panic, Musk Seeks $40 Billion: This Week's Highlights
Vibe-coding in Rust: three macOS utilities and five browser projects
Qwen3.6-35B-A3B hits 109 tok/s on an RTX 5080 with an RTX 3060 on PCIe Gen2 x4
APEX Compresses Diffusion Latents 128x and Beats a 256-Token Model on ImageNet-512
APEX orders latent channels from structure to detail and denoises them asynchronously. On ImageNet-512 it wins at 64x and 128x compression, and the code is already…
SWIM saves compute: audio-video generation gains up to 30.7% over Best-of-N
SWIM combines particle search with Metropolis-Hastings under a fixed budget: on JavisBench-Mini, +30.7% over Best-of-N on JavisDiT++ and +17.0% over EvoSearch on…