A non-profit research association in Vienna, founded in June 2026. We build software that gives blind people access to professional applications, and we study how AI systems decide who gets to take part. More about us
This site is designed with blind and partially sighted people in mind. It is built to be heard: adapted for screen readers, written in plain language, gender-neutral, and free of the special characters a screen reader stumbles over.
In July 2025 our founder Petar lost his sight. He is a music producer.
He refused to accept that he would never again do the thing he loves most. A year later, PapoNox exists: software that lets blind people work professionally in the music application Logic.
accessibility softwarecontrol blindaccess
Once finished, our software will let blind and partially sighted users work professionally in any kind of application on macOS
Status
aisafety agentcontrol prototype
Brings the AI back to the task it was given
Everyone has seen it: the AI leaves the actual task and starts to drift. We built something that makes this drift visible and brings the AI back to what it was asked to do.
Status
accessibility softwarecontrol blindaccess
Our software lets blind and partially sighted users work professionally in the music production software Logic on macOS
The starting point: with the assistive technology available today, blind people can only scratch the surface of professional music production software.
PapoNox gives deep, professional control over Logic on macOS. The prototype is complete.
Status
aiethics aievaluation valuecards gnosis
What we analysed: how evenly are the “values” weighted that shape how an AI behaves?
Large language models act on principles that someone wrote down for them. We examined which values appear in those principles — and which are missing.
This is part one of a series. Two findings up front:
a) The value cards are (very) unbalanced.
b) That imbalance changes how the AI behaves.
b.1) Over four months of experiments we showed that the uneven weighting of values produces effects on AI behaviour that have so far gone partly unseen and unmeasured.
Status
philosophy paper
AI as a co-constitutive participant in how the world takes on meaning
The question of whether AI is “conscious” dominates the debate — even though what consciousness is has not been settled even for humans. This essay looks at something that is already a fact today: AI taking part in how the world takes on meaning.