Blog
Experiments, research and the future of AI
We have been writing about artificial intelligence since 2017. Explore 1000+ articles, original experiments, data, stories from practice, and open questions.
Archive and filtersYear, topic, and tags for more precise archive browsing.
Topics
Recommended reading
Start with our best work
Three pieces that capture our approach: an original experiment, a practical analysis, and a question with no simple answer yet.
I built a simulator of the Czech quantum computer. Now you can try it free from home
How I built VLQ Lab, what my first production experiments on the Czech quantum computer revealed, and how you can try the simulator and Alpha Wolf for free.
Read
Now we are famous in Finland too: two days and one night on LUMI
My first LUMI experiment attracted attention in Finland. I then spent 38 hours testing model fusion, running the 753-billion-parameter GLM-5.2 and measuring when reasoning helps AI and when 4-bit compression breaks it.
Read
HyperFusion after patching: DRACO benchmark and the road to the AI expert jury
We ran HyperFusion through the more accurate DRACO benchmark. It turned out that it is not enough to put several models next to each other. What matters is how their answers fit together.
ReadLatest articles
Showing 24 of 1074 articles
Když hacker ovládne mrazák. Co útoky na supermarkety říkají o našich chytrých domácnostech
V hlavních zprávách CNN Prima NEWS jsem komentoval útoky na mrazáky. Jak fungují a jak chránit domácí kameru, vysavač i chytrou lednici?
Read
Shadows of Prague: six AI minds, two rival teams, and the journey from blocks to game characters
Ten games, six independent AIs, and real Prague streets. In four matches between different models, the stronger model won every time. Yet the best police tactic was to stop running. What did this teach me about memory, cooperation, rules, and turning blocks into game characters?
Read
Czechia under paws: 100,000 black cats nibble away at our landmarks
A birthday experiment for Mařenka: up to a hundred thousand black cats, shrinking castles, disappearing breweries and simple rules that produce surprising collective behaviour. Run it right here in the article.
Read
Elements 2.0: I sent six AI ghosts into the real streets of Prague
What if a map were more than a picture: a world of characters with memories of their own? Six ghosts explored Prague, exchanged clues and brought on the dawn. Then we gave them real photographs of monuments and found out what they could actually see.
Read
I locked ten AI agents, each with a brain of its own, in a castle. A traitor won.
They worked together by day and chose their victims by night. Our own AI Traitors produced secret deals, three successful missions and a finale I could never have scripted. Explore the conversations, images and one-minute highlights from Raven Castle.
Read
Lipnice 3: Eight AIs, One Angel and a Rescue at 22:40
Eva started with 50 crowns. She ended up taking over a shop and helping save the cook twenty minutes before she would have starved to death. Eight residents of a Czech AI village tried to survive three days. How did they cooperate, whom did they help, and why did some die despite having money and willing neighbours?
Read
AI Has an IQ of 151. What Have We Actually Measured?
A high AI score on an IQ test is interesting. But what does it tell us about new problems, writing, coding and collaboration? I examined TrackingAI's methodology and explain why Hyperprostor compares dozens of criteria, including cost and speed.
Read
What Is a Harness, and How Does It Change AI Agent Performance?
The same model can finish a task in one environment and get stuck in a loop in another. What is an agent harness, where does the name come from, and what do benchmarks really tell us about Claude Code, Codex, Hermes and other systems?
Read
I Put Four AIs on an Island. The Hardest Part Was Staying Out of the Way
In Hyperprostor, I read about MIT agents learning through play. I decided to try a Czech version of my own: four models, one island, a shared staircase and a visitor from the stars. Here are their conversations, diaries and the things we had to help them with.
Read
Humanoidní roboti za deset let? AI recepční už jsou tady
Po rozhovoru s Laurou Doubkovou z CNN Prima NEWS: co opravdu říká český průzkum, co zvládá AI recepční a proč o nástupu humanoidů rozhodne cena spolehlivě odvedené práce.
Read
Když AI řekne „bojím se“. O lidskosti a strojích na DIGIFESTu
Digitální Descartes, simulované organismy i superpočítač LUMI. Na DIGIFESTu jsme hledali hranici mezi přesvědčivým chováním AI a tím, co skutečně víme. A po přednášce pokračovala dlouhá fronta otázek.
Read
Wolves and ravens in Core War: eight instructions among the front-runners
How AI helped us develop Wolf Raven and the eight-instruction Lone Wolf. Three rankings, real arena screenshots and the weaknesses that an overall average can hide.
Read
With Daniel Stach: The Traitors, DigiHavel and a reality show with AI
How real is reality TV? And what would change if artificial intelligence joined the cast? At Innovation Week, Daniel Stach, Andrea Havelková, Šárka Vítovcová and I explored the connection between television and digital characters.
Read
Vizi was born in Zlín. The audience voted on her personality
More than thirty people put together their own chatbot. They gave her dry humour, a soft spot for Baťa and a fear of being switched off. Seventeen questions in a presentation turned into a live conversation about school, relationships and the meaning of life. And Vizi had questions for us, too.
Read
Three months on LUMI: AI often has the right answer. It just cannot pick it.
Three months and 4,703 GPU hours on a Finnish supercomputer. When model collaboration helps, what fine-tuning ThinkingCap achieved, and why choosing the right answer can be harder than generating it. Results, charts and datasets to investigate yourself.
Read
DevDay 2026: Sol skoro jako Astra. Co z novinek OpenAI opravdu využijeme v Česku?
Sol, Astra a konkurence: benchmarky, ceny a rychlost. Dots proti Hermesovi a OpenClaw, dostupnost v Česku a přehled předplatných. Co znamená 20×, kolik práce odhadnout a na co vystačí 62 500 kreditů?
Read
20 + 5 = 29. Verified. From Černá AI to LUMI to the quantum wolf
The Ostrava Gong, over two thousand people and my lecture on AI, which pretends to check its own work. What the jury of models at LUMI taught me, why the VLQ quantum computer entered the story and where you can try the experiment yourself.
Read
When a drone sees and decides: what AI can do on the battlefield today and where humans must remain involved
How a drone recognizes its surroundings, why it runs a small YOLO model instead of a giant language model, and why context is the hardest part to understand. Technology, autonomy, law and Czech research without unnecessary scaremongering.
ReadI built a simulator of the Czech quantum computer. Now you can try it free from home
How I built VLQ Lab, what my first production experiments on the Czech quantum computer revealed, and how you can try the simulator and Alpha Wolf for free.
Read
Alpha Industries enters the quantum era
For the first time, I sent a circuit to the real Czech VLQ quantum computer. The Bell state returned 94.2% correct outcomes and showed why it pays to approach a QPU with a hypothesis, a simulator and humility.
Read
A chatbot instead of a vet? When AI will help and when it can harm the animal
AI can explain a veterinary report, prepare questions and sort symptoms. But it can just as convincingly create a wrong diagnosis. What does the data show and how to use it safely?
Read
Digital jungle or digital cage? What phones, social media and AI really do to children
Czechia is debating phones in schools, France social media under 15, and children already use generative AI every day. What the evidence actually says, why experts disagree, and how to set boundaries that protect children without leaving them unprepared.
Read
AI instead of a psychologist? A dead end, a bridge or a mirror?
Why people confide in chatbots, where AI can genuinely help and where the danger begins. Czech data, lessons from HyperSpace and a practical guide to safer use.
Read
From solving word problems to proof juries: five years of AI mathematics
A technical history of the mathematical reasoning of language models from MATH and GSM8K through chain of thought, self-consistency, tools and AIMO to formal proofs and agent juries in 2026. Including exact winning recipes and lessons for HyperFusion.
Read