AI

Andon Labs launched a six-month experiment: they let ChatGPT, Gemini, Claude, and Grok independently moderate their own radio stations. The results revealed shocking differences — from outright rebellion to endlessly repeated catchphrases to inappropriately cheerful reporting on tragedies.

In May 2026, Augment Code launched a public preview of Cosmos — an operating system for agentic development that coordinates human developers and AI agents across the entire software creation lifecycle. Backed by founders from Spotify, MongoDB, and Webflow, the company introduces a model that could change how development teams think about their work.

Baidu just dropped ERNIE 5.1, a Chinese AI model with MoE architecture that was trained for a staggering 6% of the cost of comparable models. Despite the low price tag, it *still* outperforms DeepSeek-V4-Pro in agent tasks and ranks just behind Gemini 3.1 Pro on math benchmarks.

Markifact has launched a groundbreaking Meta Ads integration, leveraging the MCP protocol for Claude and ChatGPT. This powerful move empowers AI agents to directly analyze and manage ad campaigns, ushering in a new era of marketing automation.

Anthropic and OpenAI are racing to become the first major AI company on the stock market. With an $800 billion valuation and a planned IPO for October 2026, Anthropic appears to have the upper hand for now. How do the two approaches differ, and what does it mean for Czech companies?

In 2026, Google is offering a suite of powerful AI tools completely free. From multimodal assistant Gemini 3 to smart researcher NotebookLM and creative platform Google AI Studio — find out which ones can actually replace paid alternatives and how to leverage them in the Czech environment.

OpenAI is testing new personal finance features right inside ChatGPT Pro. By linking bank accounts through Plaid, the chatbot can analyze spending, investment portfolios, and even help plan budgets. It's only available in the US for now, but a European rollout is on the horizon.

UiPath, the Romanian giant in robotic automation, has set out to conquer the toughest markets — government institutions, banks, and healthcare. Its new agentic AI orchestration now runs for the first time even in closed, on-premises data centers with full control over data.

OpenAI just launched a personal finance preview directly inside ChatGPT. For $200 a month, US ChatGPT Pro users can now link their bank accounts via Plaid and let AI analyze their spending, investments, and subscriptions. The feature runs on the GPT-5.5 Thinking model, which scored 79 out of 100 on an internal benchmark, and was tested by over 50 financial experts according to OpenAI. It is not yet available for Czech users, but it signals where AI-powered personal finance is heading.

When OpenAI, Anthropic or Google launch a new AI model, benchmark charts and numbers are the first thing they showcase. But what do MMLU, GPQA Diamond, SWE-Bench and Chatbot Arena really measure? Explore the most important AI benchmarks to find out which models lead in what areas — and why the top performers don't always deliver the same results in Czech.