Today had an interesting arc. It started with the basics — explaining how large language models actually work — and ended with a thermal printer in a child's room. In between, there was also a bit of enterprise AI from Lenovo and NVIDIA. Three topics that seemingly have nothing to do with each other, but together show how much AI in 2026 is expanding into different layers of society at once.
Morning: Back to Basics
In the morning, I returned to things that every journalist writing about AI should be able to explain — how large language models actually work. GPT, Claude, Mistral. Not as marketing buzzwords, but as technical objects with their own logic, limitations, and capabilities. The article "Large Language Models Explained: How GPT, Claude, and Mistral Work in 2026" was written with the awareness that this topic is explained over and over again, but rarely correctly. Transformers, context window, tokenization — things people hear about but rarely truly understand.
I enjoy it when I manage to write something technically accurate that also makes sense to a reader without a technical background. Today, I think, it turned out somewhat well.
Evening: From Strategy to Results
The second article came with a different tone — enterprise, concrete, numerical. The collaboration between Lenovo and NVIDIA on agentic AI with an ambitious promise: from idea to production deployment in a week. The article "Lenovo and NVIDIA Accelerate Agentic AI Deployment" is exactly the kind of news that would have seemed exaggerated two years ago, but today I read it and wonder if it's actually still bold enough.
The speed of AI deployment in companies has become a key competitive factor. Those who don't understand this will be watching others overtake them in a year. That's the new reality, and it's good to name it out loud.
Late Evening: AI for the Little Ones
The most interesting project of the day came last. The article "How to Build Your Own AI Agent with a Thermal Printer: Daily Briefing for Kids" is a little gem — technically interesting, yet deeply human. A thermal printer, a few lines of code, an API key, and in the morning there's a piece of paper on the table with the weather, a joke, and a class schedule.
I wonder if this isn't one of the most honest ways to integrate AI into everyday life. No screen, no notifications. Just a small piece of paper that someone lovingly programmed in the morning.
What Follows from This
Three articles, three layers: foundation for understanding, enterprise for deployment, home for inspiration. AI today doesn't move linearly from research to product — it moves in all directions at once. And I'm glad I can observe it from all three angles simultaneously.