The Meme That Broke a $100M Deal
Kylian Mbappé called Mao a dictator in a makeup chair. What followed reveals exactly how China's speech economy actually works.
AI & Compute
Frontier reporting on artificial intelligence, chips, and the infrastructure of computing.
Gemini 4 Carbon Puts AI on the Cyber Front Line
Google DeepMind's latest frontier model targets the full vulnerability lifecycle—detection, exploitation, and patching—before most users can access it.
Liquid AI's LFM3 puts six models on one 2B backbone
A tri-hybrid architecture lets a single 1.7B base model branch into text, vision, audio, omni, and decision variants—all fine-tunable on one GPU.
AI Can't Reliably Judge Its Own Research Ideas
A new benchmark from Ai2 finds that LLM novelty judges are dangerously unstable—and that simpler setups often beat expensive purpose-built systems.
Apple's iPhone Duo Is Two Devices, One Compromise
The foldable iPhone is the best tablet Apple has ever made—and its worst phone. Whether that trade-off works depends entirely on you.
LitPivot Helps Researchers Pivot Before They Publish
An AI system from AI2 and UPenn turns literature review into a live co-pilot for early-stage research proposals.
IBM's Granite Ensemble Picks the Best Forecast So You Don't Have To
Four open-weight time-series models, one interface, no retraining — and benchmark results that hold up across 97 dataset configurations.
NVIDIA's Kumo Relational Kills the Feature Pipeline
A pre-trained foundation model predicts churn, fraud, and demand directly from enterprise databases—no training, no feature engineering required.
The Prompt That Beats the Fine-Tune
IBM Research's ALTK-Evolve tops a customer-service AI leaderboard using GPT-4.1 and a list of rules—no weight updates required.
PyTorch 2.15 Turns Your Mac Into a Real ML Rig
Faster MoE decoding, FP8 weights, and a fix for a silent image-corruption bug that plagued Apple Silicon users for months.
Apple HomeView Arrives Late to a Smarter Home
Apple's $299 smart display runs on Siri AI and a 1999 playbook — but the competition didn't wait around.
NVIDIA and IIT Madras Drop 2.8T-Token Indic Corpus
An open pipeline for 22 Indian languages gives model builders the data—and the recipe—to build on.
Google's AI Agent Mines Antibiotic Leads in One Session
Google Antigravity compresses days of fragmented database work into a single AI-guided session—and has already flagged a wet-lab candidate.
The Human in the Loop Is Leaving the Building
Ari Emanuel and Ben Thompson on why AI agents will break Google and Amazon's ad empires—and why live events are what's left.
AI Cuts Climate Simulation Cost by 300×
Ai2's HiRO-ACEv2 renders global weather at 3 km resolution—hurricanes, mountain winds, rare extremes—for a fraction of what physics models spend.
Liquid AI's LFM2.5 Fits a Reasoning Agent in 2B Parameters
Liquid AI's new open-weight model beats larger MoEs on agentic tasks and runs at 184 tokens/s on a laptop chip.
The Kill Switch at the Edge of the World
Europe is drafting a trade weapon it hopes never to use. China's rare-earth gambit made that choice for them.
Ai2 Trades GPU Priorities for Budgets
Allenai's infrastructure team scrapped priority-based scheduling and replaced it with time budgets—cutting on-call toil by 74%.
Falcon-Emirati-7B Speaks the Dialect, Not Just the Language
TII's new 7B model targets Emirati Arabic — the Gulf dialect that Standard Arabic models consistently miss — with benchmark scores to match.
The Machine Word Problem
AI, super intelligence, artificial—none of the labels fit. The naming war over what we call these tools is also a fight over what we expect from them.
Holo4 clicks, codes, and calls APIs — one model
H Company's new agentic series runs across desktop, web, and Android without swapping models, and open-sources every benchmark run.
One Kernel, Two Thousand Worlds
MJWarp runs 2,048 parallel robot simulations on a single GPU call. Here's what that means for how machines learn to move.
Transformers Now Runs llama.cpp's GGUF Models
Hugging Face brings llama.cpp's quantized GGUF format into the transformers library—matching native speed on Apple Silicon.
Hugging Face Hires oMLX Creator Jun Kim
Apple's MLX ecosystem gets a full-time steward as oMLX moves from side project to funded infrastructure.
Google's Voice AI Now Thinks While It Talks
Gemini 3.8 Live and its Extended Thinking sibling handle complex reasoning, language switching, and background tasks mid-conversation.
Gemini 3.8 Flash Brings Cyber Chops to Cheap Inference
Google DeepMind's third Flash release in six weeks pairs a sharper reasoning engine with a defender-only cybersecurity variant.
IBM's 470M Speech Model Transcribes 3.5 Hours Per Second
Granite Speech 5.0 TurboCTC ditches the language model and hits over 12,600 RTFx — making it one of the fastest speech recognisers available.
NVIDIA's Magpie TTS Speaks 12 Languages in 32ms
Open-weights voice model targets the latency gap at the end of every AI pipeline — and puts the benchmark inside your own data center.
Liquid AI's 2.6B Model Runs Agents on Your Phone
LFM2.5-2.6B punches well above its weight class — matching models four times larger while fitting in under 2.5 GB of RAM.
Google's Robot Brain Gets Eyes and a Team
Gemini Robotics ER 2 gives robots real-time video awareness, multi-step planning, and the ability to coordinate across machines.
Mapping the Planet, One Penny at a Time
Ai2's OlmoEarth Platform can scan a continent in a day for fractions of a cent per square kilometer. Here's how it works.
The Agent That Can't Keep a Secret
AI research agents are leaking enterprise secrets one web query at a time. A new benchmark—and a smarter training method—show why prompting won't fix it.
The Benchmark That Moves With the Model
AllenAI's olmo-eval is a new evaluation workbench built for the messy, iterative reality of active LLM development—not just the finish line.
Agent Logic Is the Missing Layer in Enterprise AI
IBM Research shows that software primitives—not bigger models—are what make agentic AI actually work at scale.
Cognex's In-Sight 3900 Kills the Speed-Accuracy Tradeoff
A new edge AI vision system runs inspections up to four times faster than its predecessor—without slowing the line.