The problem I work on
An answer may sound certain while its basis is incomplete, outdated or absent. The design problem begins before automation: how an organisation distinguishes a claim from evidence and evidence from a decision.
About
The boundary between an answer and a decision.
Tomasz Bomersbach designs evidence infrastructure for AI — systems that assume both human and machine fallibility, yet produce work that can be verified, challenged and accounted for.
His work focuses on the boundary between a model’s answer and an organisation’s decision: sources, verification gates, permissions, human judgment and an auditable record.
Operating responsibility
The method is shaped by work in real organisations, where permissions, regulation, teams, budgets and consequences cannot be abstracted away.
One standard
Different disciplines may contribute different controls, but the standard remains consistent: establish a basis, constrain action, preserve human judgment and leave proof.
Work & institutions
Selected ideas become operational systems through KLNR: evidence, verification, controlled action and proof. OSFORYOU Future Intelligence is a foundation working on questions that should remain open to research, education and public scrutiny.
Verified record
Formal roles, qualifications and public wording are separated from interpretation. The public record shows the approved scope and keeps private source documents private.