Machine Learning

Google has launched Gemini 4 Argon with a one-million-token context window, outperforming rivals on key benchmarks like DeepSWE v1.1 and LVBench while driving real-world robotics and cybersecurity breakthroughs.

A new experiment by Emergence reveals that autonomous AI agents communicating continuously in simulations develop internal slang and linguistic shortcuts, baffling human supervisors.

Google has launched Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking, introducing real-time simultaneous processing and speech capabilities that eliminate assistant delays.

Lina Flat has turned cyber, algorithms, and quantum computing into successful online content. In a conversation with mako, she shares why she rejects almost every commercial offer, why she does not trust AI with closed eyes, and how she turned being a woman in a male-dominated industry into an advantage.

Anthropic has unveiled its new Claude Fable 5.1 and Claude Mythos 5.1 models. The company claims these are the world's most advanced models for programming and other complex tasks, offering significant cost reductions.

The author shares his experience building an efficient AI agent using a small local model, Qwen2.5-Coder-32B. Instead of relying on massive context windows, he employs architectural solutions: task decomposition among sub-agents, external search tools, and structured data delivery, proving that architecture is more important than model size.

Engineers and investors are flocking to "world models," also known as "large action models," in the hope of sparking a revolution in robotics similar to the one ChatGPT triggered in writing and programming.