#Research
Curated coverage on «Research»: releases, research, benchmarks, open-source.
A Billion for 12 People: Why Yann LeCun Rejects Language Models
Startup AMI Labs raised a billion dollars with just 12 employees. Yann LeCun is convinced: the era of classical language models has ended, and AI requires a…
GitNexus: How a New Tool Taught AI Agents Structural Code Understanding
AI assistants excel at writing local code but often fail to grasp the global architecture of projects. The open-source GitNexus project solves this problem using…
DeepSeek-V4: How New Compression Algorithms Made One-Million-Token Context a Reality
DeepSeek introduced the fourth generation of its models. Through radical attention compression, processing massive data volumes is becoming accessible not just to…
Grammarly sued over AI feature that presented advice as coming from scientists without their consent
Why people don’t trust AI: surveys show a widening gap between hype and reality
Google replaces news headlines in search with AI-generated ones
Anthropic: AI is not eliminating jobs yet, but the gap between users is already growing
NeurIPS reverses controversial policy after protests by Chinese AI researchers
The world's leading AI conference, NeurIPS, announced changes to its participation rules, sparked a wave of criticism from the Chinese scientific community, and was…
Wired: the wave of AI sites isn’t poisoning the internet — it’s making it falsely happy
A new study found that AI content is not flooding the internet with disinformation, but gradually shifting its tone toward artificial, monotonous optimism.
Generation Z Is Growing Disillusioned With AI but Keeps Using It — Gallup Poll
The “AI Is Inevitable” Trap: Have We Reached Peak Hype Around Artificial Intelligence
OpenAI launches fellowship program for independent AI safety researchers
Meta's New Muse Spark AI Model Puts It Back in the Top League
Gemma 4: Google Unveils Its Most Efficient Open AI Models
Google has announced Gemma 4, a new generation of open models optimized for complex reasoning and building autonomous AI agents.
Anthropic: Claude AI Model Has Its Own Analog of Human Emotions
Anthropic researchers looked inside Claude's architecture and found something surprising: the neural network forms internal representations that work much like…