Four questions

A system running on several sources does not usually fail loudly. It keeps producing an answer in the same format, at the same rate, with the same apparent authority — after the thing that made the answer trustworthy has gone.

These four parts each take one way that happens. They are short, they are independent of each other, and each one ends with what the approach does not establish, because that is the part most worth knowing before a conversation rather than after one.

Each part corresponds to one of four US provisional patent applications.

The series

Part 1 · Which source do you believe?

When your instruments disagree and one of them is confidently wrong. A failed sensor usually looks failed. An interfered-with sensor can look healthy. Comparing sources against each other is a consistency check that does not depend on the suspect source diagnosing itself.

Application 64/144,562 · most relevant to maritime & defense

Part 2 · How many sources did you actually have?

Systems degrade quietly, and that is the part that hurts. Sources drop away one at a time and the output looks unchanged. Tying the system's state to a declared count of sources genuinely taking part — and recording the roster — makes the degradation a declared fact rather than an inference.

Application 64/149,078 · most relevant to critical infrastructure

Part 3 · Prove what the system knew

Six months after the incident, what can you actually show? The hard part of a review is rarely the physics — it is reconstructing what the machine knew at the moment it decided, from logs written to help a developer rather than answer an investigator.

Application 64/159,476 · most relevant to aviation & autonomous systems

Part 4 · When the sources are judgements, not numbers

Several models disagree about what they are looking at. Which one do you act on? Shared failure modes can make agreement less informative than it appears. Measuring how far a source sits from the others, and separating reaching a conclusion from being permitted to act on it.

Application 64/161,448 · most relevant to cybersecurity & AI agents

What these are, and are not

Each part describes a capability at the level of what it does, not how it is built. The mechanisms are implemented in a reference implementation and exercised in simulation.

Not yet proven in the field. No sea trial, no hardware-in-the-loop, no live interference, no installed base. Where a part describes a workflow that is proposed rather than built, it says so in that part.

Not granted patents. Four US provisional patent applications have been filed. A provisional application is not examined, and the scope of any future claims is unresolved.

A first conversation

Bring a nonconfidential example — a case where your sources disagreed, or where establishing what the system knew turned out to be the hard part. We will work through whether this architecture would earn its place on your platform and what a joint evaluation would need to establish. Recorded data, if it ever helps, comes later and under terms you set.

A first call is a technical discussion, not a pitch, and we will say when the answer is that this would not have helped.

info@rvappliedsystems.com · RV Applied Systems, Inc., Las Vegas, Nevada