Our experiments
Slovenská verziaLabZone Labs
Our team runs small, timeboxed experiments: one idea, a short build, and an honest result, whether it works or not. We use public or test data only, and we write down what worked, what didn't, and what we would reuse on a client project. Each experiment is a small working proof of something we can build for you as a prototype on your data.
How we run them
One experiment, one question.
We start with a hypothesis and a kill criterion, and end with a demo to the team.
Hard timebox.
A short, fixed build time. Client work always comes first.
Kill early.
If the first attempt shows the idea can't work, we stop and write down why. A documented dead end counts as done.
Honest numbers.
We publish what we measured and under which conditions, including the caveats.
Experiments
W45 · 2–6 Nov 2026In progress
Rewrite or review?
Local AI vs cloud on extracting data from 60 documents, plus a study of whether reviewing prefilled data saves time.
- TypeScript
- Qwen via Ollama
- Gemma
- Claude
- preregistration
W44 · 26–30 Oct 2026In progress
Ask the statistics
An assistant that picks the right table from 10 Statistical Office of the SR open-data tables and shows a chart with its source.
- TypeScript
- Qwen via Ollama
- Chart.js
- STATdata API
We'll show this on your data.
We build a working prototype on your data in 2–4 weeks. If you continue, the prototype price is deducted from the first development invoice. If not, you pay only a fixed fee for validating the prototype.
We store questions without IP addresses for 30 days to evaluate and improve the experiment. We use your IP address only briefly to limit the number of requests and do not store it. Please don't include personal data in your questions.