The best way to enforce a ruleis to make breaking it unreachable.
Avantura is a two-person company based between Belgrade and Bali. We build deterministic software for problems where written rules meet high-stakes decisions.
Five countries,
then a tool.
Ana is Serbian. She started moving the year she turned eighteen: Austria, then China, then Thailand, and now Indonesia. Four moves on a passport that doesn’t open many doors taught her to think in terms of feasibility by default: which connections exist, which visas apply, which combinations of route and timing and paperwork produce a trip someone like her can actually take. Constraint was the shape of every plan, not an exception to it.
Her work took the same shape. English teacher, then technical support, then custom travel agent, building bespoke trips for clients who knew what they wanted but not what was reachable. When she set out to start her own agency, she realised the bottleneck wasn’t the agency. It was the gap between non-experts imagining a trip and the set of trips that were actually possible. Fixing invalid itineraries after the fact was slower than preventing them from being reachable in the first place.
She started building a tool, drawing on Bret Victor’s essays on learnable programming, the idea that a good tool shows the consequences of a choice as it is being made, and that impossible choices should not be reachable at all. A few weeks into the prototype she showed it to Anja, a childhood friend from Belgrade. Anja had spent five years at BlackRock in business intelligence and implementation, then left to try interior design and learned CAD along the way: tools where the work is constraint-first and invalid states literally cannot be drawn. Anja recognised what PRIM was reaching for and joined the same week. PRIM, the engine underneath everything, has been the work of both of them since.
Travel taught the engine what it needed to know. Written rules, feasibility graphs, fatigue budgets, itineraries as state machines. It remains the most built-out adapter in the codebase, and the reason the engine is as general as it is. Somewhere in the middle of building it, Ana realised that motorsport, something she had followed since childhood, had the same shape but sharper. The rules were written. The stakes were high. No one had turned the rulebook into something executable. Travel stayed as the engineering proof. Motorsport became the first commercial focus. Today Avantura runs between Belgrade and Bali, with Anja commuting to Indonesia when it is time to build in the same room.
Three commitments
we operate by.
These are not values. They are the hard constraints on how Avantura is built and how we decide what ships.
Prevent, don’t correct.
The best enforcement is making breaking impossible in the first place. Post-hoc fixes are slower, less trustworthy, and always late.
Rules should be executable.
Written policy and the system that enforces it should not be two different things. A clause of regulation and the code that checks it should point at each other.
Humans decide, engines validate.
We don’t automate judgment. The steward, the officer, the auditor decides. The engine records the decision and checks it against the rules.
If you have rules
nobody has executed yet.
We’re interested in conversations with people working in regulated decision environments. Sport, finance, emergency services, anywhere written rules meet high-stakes decisions.