Mathematics is the layer
everything else runs on.
Encryption, load calculations, budgets, medication, the evidence in a court case. It also decides how much of public life a person can check for themselves.
Logic, statistics, and modelling allow us to make up our own minds.
Logic
You can follow an argument and see where it breaks down, rather than believing whoever presents it most convincingly.
Statistics
You can read a number in the news and understand what it tells you, and what it does not.
Modelling
You can assess an energy forecast, a budget, or a medical treatment without waiting for someone else to tell you what to think.
Without these skills, citizens are left to trust whoever sounds most convincing. Mathematics is where we learn them, and for most of us that happens in just a handful of courses during the first years of a degree. Those courses determine how many people remain able to question what they are told.
What we want to achieve
People understand mathematics deeply enough to use it
Not just to pass an exam and then forget it, but to draw on it whenever a number, a claim, or a decision in their own lives calls for it.
People remain at the heart of it
People do the thinking. We build tools that support them, not tools that take over.
We use AI responsibly to get there
AI is the most powerful tool we have for making mathematics understandable at scale. It only helps if it leaves the person using it more capable, and that is the standard we hold it to.
We are optimistic. We believe it can be done, and we would rather spend the coming years putting that belief to the test than assume otherwise.
We operate it. Institutions stay in control.
Someone has to operate the system: keep it running, secure, and up to date as AI models evolve. We do that on behalf of the institutions that rely on it. Operating infrastructure for others comes with responsibilities, and we take them seriously.
Institutions can inspect what they rely on
Partner institutions have access to the source code and can review it, audit it, and propose changes. Operating a system on someone else’s behalf only works if they can see how it works.
Their people can contribute
Partner institutions can nominate contributors. Their work becomes part of the shared codebase under their own names and benefits every institution using the platform.
We operate it in the EU
Faculties do not need to build their own operations teams. Processing and storage remain in the EU.
Built to outlast us
If lytris ceases operations or is acquired by an entity based outside the EU, the platform converts to open-source software under the AGPLv3 license.
Institutions already hold the source code; the conversion simply lifts all remaining restrictions. No institution should build on infrastructure that could disappear and leave its courses stranded. This safeguard can be enforced without us.
How we pick the models
Every model involves trade-offs between quality, speed, and cost, and none excels at all three. lytris is many small steps rather than one model, so each step can sit at a different point on that tradeoff.
Quality
How well it explains, corrects and reasons, judged against the test cases we maintain rather than on impressions.
Speed
Someone is waiting for the answer. A model that takes twice as long has to be worth the wait.
Cost
Where quality and speed allow, the cheaper model is used. That is what keeps the whole thing affordable to run.
We maintain test cases for every step, allowing us to measure the effect of changing the underlying model.
The people and partners behind lytris
lytris is built by a small team. Advisors keep us honest by telling us when we are wrong. Educators and researchers put our work to the test. We will introduce them here as they join us.
Team
Coming soonThe people who build and run lytris day to day.
Advisors
Coming soonPeople from research, industry, and practice who challenge our thinking and tell us when we are wrong.
Research partnerships
Coming soonInstitutions and research groups that put lytris to the test in teaching and research.
People should be able to check what they are told. That includes what we tell you.
Two ways to take this further. Run it in a course with us, or take the argument apart. Both are useful to us.