AI News

Artificial intelligence edged closer to a once-unthinkable milestone this week: cracking open mathematical problems at PhD level in under an hour. While Fields Medal winner Timothy Gowers stress-tested GPT-5.5 Pro, Anthropic rivals were teaching their Claude model to "dream" — and quietly inked a computing deal with Elon Musk. What else shook the AI research world in the past seven days?

Chinese scientists unveiled Hanyuan-2, a quantum computer featuring a dual-core architecture with 200 neutral-atom qubits. It consumes less than 7 kW, and its modular design paves the way for scalable quantum computing.

The MIPI Alliance has launched a new Physical AI BoF working group to standardize interfaces for humanoid robots. Intel, Samsung, Bosch, and NXP are among the participants. According to Yole Group, the market is set to grow 56% annually and exceed $6 billion by 2030.

Imagine your kids receiving a printed slip every morning with the weather forecast, schedule, and a joke — all generated by artificial intelligence. An inspiring DIY project shows how to connect AI agents with a thermal printer and smart buttons.

Lenovo and NVIDIA have unveiled a solution that enables companies to deploy autonomous AI agents into their workflows in just seven days. This marks a fundamental leap in enterprise technology efficiency and business data automation.

American startup Perceptron has unveiled the Mk1 model, which understands video better than GPT-5 and Gemini, yet costs a fraction of the price. How does it work and why should Czech companies take notice?

AgentKanban, a new tool, transforms isolated AI conversations into shared developer workflows. It combines Kanban boards with a VS Code extension, automatically capturing context from GitHub Copilot and enabling real-time team collaboration. A free basic version is available, with a professional tier at $10/month.

Every query to ChatGPT consumes electricity, and by 2030 data centers will surpass the entire energy consumption of Japan. Startup Panthalassa is coming up with a bold solution — autonomous computing platforms on the open ocean powered by wave energy. How realistic is this vision?

What's behind the LLM acronym? Understand the principles of large language models, compare GPT-5.5, Claude Opus 4.7 and Mistral, and find out what they mean for Czech businesses and everyday users.

As AI agents begin making independent decisions in live production environments, traditional monitoring breaks down. Honeycomb responds with Agent Observability — a tool that uses OpenTelemetry to expose every step of an agent workflow in real time. For developers, the critical takeaway is clear: the basic version is free, and the company guarantees full GDPR compliance backed by a European data center.