AI

Three AI giants — SpaceX, OpenAI, and Anthropic — are heading for the stock market in 2026. Their combined value could reach $4 trillion. Yet none of them is profitable. Is this a risk, or the investment of the decade?

OpenAI is expanding its subscription plans with ChatGPT Go — the cheapest paid tier featuring GPT-5.5 Instant, costing around 120 CZK per month, and already available in Asia since last fall. What does it offer, how does it compare to the competition, and when might it arrive here?

When an AI agent in a prototype gobbles up thousands of tokens in a few seconds, it's cute. When it does the same thing in production for thousands of customers, your wallet starts hurting. How to solve token burn in agentic systems — from Early Commitment to deterministic replay.

Boomi, a leader in integration platforms, predicts a world driven by AI agents. CEO Steve Lucas explains why companies will need not just models, but above all an architecture to manage them — and what that means for European businesses.

Anthropic launched Claude for Marketing, a new site with ready-to-use prompts for marketers. Together with Claude Cowork, it demonstrates how AI is transforming both routine reporting and creative work. And what's crucial — Claude is fully available in the Czech Republic too.

Anthropic launched the Claude for Marketing page with ready-made prompts for marketers. Together with Claude Cowork, it shows how AI is changing routine reporting and creative work. And what's crucial — Claude is fully available in the Czech Republic too.

Harbor Capital has filed with the SEC to launch five actively managed Lab ETFs. Each targets the ecosystem of a single AI lab — Anthropic, DeepMind, Meta, OpenAI, or xAI. For the first time, investors will be able to bet on a specific lab rather than the entire AI sector.

U.S. asset manager Harbor Capital has filed with the SEC for five actively managed Lab ETFs. Each fund will focus on the ecosystem of a single AI lab — from Anthropic to OpenAI to Elon Musk's xAI. It's the first attempt to slice AI investments into standalone products by individual companies.

World models are learning physics from YouTube videos instead of hand-crafted simulations. Meta's V-JEPA 2 achieves an 80% success rate in unfamiliar environments. An AWS analysis shows that robotics is on the verge of the same breakthrough that language processing experienced 20 years ago.

DeepSeek announced that the 75% discount on its V4-Pro API model is now permanent. Per million tokens you'll pay /bin/sh,435 for input and /bin/sh,87 for output — 11× less than GPT-5.5 at comparable performance.