Andrei Nistor
Journalist and media entrepreneur. He founded Știrile Transilvaniei, where he is editor-in-chief, and Digitalio. He runs Story Agents, which builds AI tools for newsrooms, and founded Press Cluster, Romania’s first media cluster. His research looks at how AI models treat Europe’s smaller languages: he proposed the Linguistic Media Equity framework and measured how much more Central and Eastern European newsrooms pay to use AI.
https://x.com/AndreiNistor74 · https://www.linkedin.com/in/andreinis/ · https://doi.org/10.5281/zenodo.23552863
Articles
The AI Act speaks extensively about transparency. For newsrooms in Central and Eastern Europe, transparency should start with something simple: how much the language they work in actually costs.
Does Romania need its own language model?Poland and Bulgaria have already developed national models. Romania has several open-source models created by universities, but no national project. I have weighed the arguments for and against, and then shared my own view.
DiscoveryRSI: Recursive learning for the discovery of valuable hypothesesResearchers propose DiscoveryRSI, a framework that utilizes multi-agent evolutionary search to generate, test, and refine research hypotheses, facilitating a recursive scientific discovery process.
Object-Oriented Harness: Evolving Agent Harnesses through Object-Oriented DesignThe paper proposes OO-Harness, an optimization framework that utilizes object-oriented design principles to create a hierarchy of specialized classes capable of managing heterogeneous tasks through different execution strategies.
The token tax: the same news costs more in Romanian than in EnglishI measured eight public tokenizers on news texts in Romanian, Hungarian, and Slovak. None come close to English, and the press pays the most.
An AI model can write impeccable Romanian and still not understand RomaniaThe linguistic gap in language models isn't visible in grammar. It is visible in facts, in context, and in the question of who has the right to verify. This is where we started when we proposed the Linguistic Media Equity framework.