AI News

Don't be fooled by AI's efficiency. A new UN report reveals that data center electricity and water consumption is set to skyrocket — and more efficient models could actually make the situation worse.

Microsoft has released Visual Studio Code 1.123 with million-token context window support for GPT-5.5 and Claude Opus 4.7, chat session syncing, and parallel agent execution. Here's what it brings for developers.

A new UN report reveals that more efficient AI models don't lead to savings. Due to the Jevons paradox, consumption is rising — by 2030, AI could emit as much CO₂ as the entire UK and use more water for cooling than humanity drinks.

At WWDC 2026, Apple unveiled Xcode 27 with built-in AI agents from OpenAI, Anthropic, and Google. For the first time ever, developers get ChatGPT, Claude, and Gemini in a single IDE — no window switching, no external plugins.

One of the founding fathers of artificial intelligence has raised over 23 billion crowns for a startup that goes against ChatGPT. Instead of more chatbots, LeCun is building world models — AI that understands physical reality. With backing from Nvidia, Toyota, and Jeff Bezos.

Betting odds on Polymarket paint a surprising picture of the 2026 AI race. Anthropic has an 87% chance of claiming the best AI title, GPT-5.6 could drop any day now, and Google is losing steam at 10%. What do the hard data reveal about the future of artificial intelligence?

Google has released its AI-powered hydrological model as open source. The Czech Hydrometeorological Institute played a major role in its development and testing, and was also the first to prepare the model's integration into live production use.

Meta is entering the AI agent race. Its Hatch isn't just about answering questions — it's designed to complete tasks autonomously, from writing code to planning vacations. Priced at $200 a month and integrated into Instagram, Zuckerberg is aiming to give OpenAI and Anthropic a run for their money.

South Korea's RoboGate platform by AgentAI tested seven of the most advanced VLA models in industrial scenarios. The result? All of them failed 100% of the time — including Nvidia GR00T, which had excelled in other benchmarks. Despite this, Nvidia admitted the startup to its Inception incubation program.

Korean platform RoboGate by AgentAI tested seven of the most advanced VLA models in industrial scenarios. The result? All of them failed in 100% of cases — including Nvidia GR00T, which had excelled in other benchmarks. Despite this, Nvidia still accepted the startup into its Inception incubation program.