AI News

Alibaba has introduced Qwen3.7-Max, its most powerful AI model for the agentic era. It autonomously programs, optimizes GPU kernels, and can work continuously for 35 hours with over a thousand tool calls. In benchmarks, it beats both Opus-4.6 and DeepSeek-V4.

At Google I/O 2026, the company announced the end of Gemini CLI. On June 18, the tool with 100,000 GitHub stars will be replaced by Antigravity CLI. What does this mean for developers and how will the migration work?

At I/O 2026, Google unveiled Google Pics — a brand-new image editing tool — added voice features to Docs and Gmail, and expanded AI Inbox to all subscribers. Here's what these updates mean for users in the Czech Republic.

Andrej Karpathy, one of the original co-founders of OpenAI and former head of AI at Tesla, is joining Anthropic. He'll lead a team using Claude to accelerate pre-training research for language models — a step toward recursive self-improvement of AI.

At I/O 2026, Google unveiled the Gemini 3.5 Flash model with improved agent capabilities and a visual redesign called Neural Expressive. New additions include the Gemini Spark agent, the Omni Flash video generator, and a 60% price cut on the premium subscription.

Google announced Gemini Spark at I/O 2026 — an AI agent running 24/7 in the cloud. It monitors your finances, writes reports from emails, and books restaurants. Launching in the US at $100 per month. What can it do and when will it reach Europe?

OpenAI and Anthropic have signed joint ventures with TPG, Blackstone, Goldman Sachs, and Bain Capital worth $5.5 billion. They plan to deploy their AI engineers directly into thousands of companies and carve out a slice of the consulting market.

Language models generate text, but they don't understand physics. SandboxAQ introduces Large Quantitative Models — AI trained on quantum-mechanical simulations that doesn't hallucinate. LQMs are now accessible through Anthropic Claude via the open MCP protocol. The first model, AQCat Adsorption Spin, is already live.

American and Japanese researchers have received $6.2 million to develop AI robots that will monitor for errors during ship construction right on site. The system is designed to act as a co-pilot for workers.

Anthropic has revealed best practices for deploying Claude Code in large enterprises. From monorepos with millions of lines to decades-old legacy systems — the key to success isn't the model itself, but what you build around it.