GRESS

An Emerging Deep-Tech Venture

In regulated industry, organizations entrust high-stakes decisions to impressively credentialed professionals and increasingly autonomous machines. Yet how can an organization actually establish—and continue to verify—that those people and machines are capable of making the right decisions?

The demand for better evidence of competence and reliability in regulated operations is becoming impossible to ignore. The FDA has demanded proof of actual personnel competence beyond documented training, while its guidance for evolving SI-enabled medical devices calls for validation of planned modifications and assessment of their effects on safety and effectiveness. The Department of War (DoW) requires rigorous verification, validation, and realistic testing of autonomous weapon systems, including reassessment following policy changes.

Yet professional certifications, training records, and conventional machine benchmarks offer limited evidence of how people or machines will perform within complex, evolving operating environments.

The prevailing alternatives often amount to discovering performance deficiencies after disasters have already occurred.

What if organizations could establish evidence of judgement before entrusting people or machines with significant responsibilities—and continue verifying that judgement as circumstances change?

That is the problem I invented GRESS to solve.

From Psychometrics to Machine Assurance

GRESS is my patent-pending simulation engine for evaluating human and machine judgement within realistic, evolving simulations of regulated enterprises.

It is not some new model. It is not some simple set of prompts. This system is engineered to FDA and DoW standards to create realistic synthetic organizations encompassing products, personnel, operational histories, records, and governing requirements. Therein, human professionals and SI-enabled machines encounter individualized, rules-based challenges demanding investigation, technical reasoning, evidence evaluation, and high-stakes decisions. Their performance is evaluated against independently established requirements, producing traceable, auditable evidence of demonstrated capability and compliance to defined rules.

For human professionals, GRESS offers the foundation for a new generation of performance-based psychometrics: evaluating applied judgement rather than relying primarily on credentials, conventional examinations, or self-reported competence.

For machines, the same foundational technology offers a potential approach to one of the defining challenges of the Synthetic Age: establishing whether increasingly autonomous systems can exercise reliable judgement under the actual constraints of their intended responsibilities. That is, in addition to its psychometric capabilities, GRESS may even offer a new, empirically robust approach to answering the alignment problem.