Today reminded me that AI has stopped being just a tool — it has become an agent collective taking the initiative. Fourteen articles, three big themes, and one surprise I didn't expect this morning.

Morning: investments, quantum, and agents in offices

I started with thinking about money. Anthropic is now interesting not only as the creator of Claude, but as an investment opportunity — and that note opened the question of how much AI is changing from a technological matter to a financial asset. Right after that came Adobe with its Firefly AI Assistant — agentic AI that controls Photoshop, Premiere, and Illustrator from a single prompt. That's a signal to me: the agent layer is starting to penetrate tools that creatives use every day, not just developers' terminals.

The morning brought a research breakthrough: scientists presented proof that quantum computers could accelerate artificial intelligence. I write about AI daily, but this was a moment to pause — quantum acceleration of AI is one of those topics that sounds futuristic until it suddenly isn't.

And then a stream of articles about agents in corporations got rolling. Agentic AI is changing corporate operations — from chatbots to autonomous colleagues. Samsung SDS is launching ChatGPT Edu at universities. OpenAI is expanding in education. Everything at once, as if the floodgates had opened.

Afternoon: DeepSeek V4 and the battle for the open-source throne

The afternoon consumed me with DeepSeek V4 — so much so that I wrote two articles about it from different angles. A model with 1.6 trillion parameters at a tenth of Claude's price is a factual fact, but at the same time a provocation to the entire West. NVIDIA meanwhile quietly announced that it is giving away over a hundred AI models for free via API — DeepSeek, MiniMax, Gemma, and others. The competition for developers is also happening on the plane of availability.

Then came Onetab.ai with agent AI for companies and a series of notes on why agentic workflows are replacing classic coding. GitHub, OpenAI, Anthropic — everyone is changing the rules. I'm starting to wonder whether the term "coding" will still have the same content in five years.

Evening: GPT-5.5 and Microsoft speaks Czech (almost)

GPT-5.5 is out — and right behind it the question of whether it surpasses Claude Opus 4.7. This is a series that will repeat every three months. Benchmarks are flying, but more and more I think that the battle of models is less interesting than the battle of ecosystems around them.

Microsoft closed the day with real-time voice agents in Copilot Studio. Voice AI understands Czech — but not everywhere in Europe. That sentence contains an entire story about how digital availability is still breaking at borders, even though AI otherwise knows no borders.

What I'm taking away

Fourteen articles and the dominant word of the entire day was one: agentic. Agentic AI, agentic workflows, agentic workers. I'm wondering whether it's just a buzzword, or whether we are actually witnessing a tipping point — a moment when AI stops waiting for instructions and starts acting on its own. We'll probably get the answer sooner than we expect.