A Message to the Reader

There are very few times in life when someone can say they did something genuinely “important” and have it be more than self-delusion. I implore you to closely examine what is stated on this page. A decision made some 30+ years ago has put me in a position where my life’s work—and I as a person—have been effectively erased by the relentless marketing hype of the billionaires who run the United States.i

Thirty years ago, I created a product called Floater, sold by Floater Corporation. It was a personal finance program with a unique core: it was designed specifically for working-class people whose expenses exceeded their income. It analyzed payment due dates and used cash-flow management techniques—similar to those used by banks—to juggle payments to benefit the user. This was powered by what I called the Floater Artificial Intelligence engine. I used the term “AI” because it was popular in science fiction, but I never claimed the application did any original thinking. The engine learned the preferences of the user and, when decisions were required, relied on what the user would prefer. It was much closer to cybernetics than to what is now called AI.

Soon after receiving funding, my path crossed with the father of all tech billionaires: William Gates. It was Bill Gates who decided there should really only be one software company—Microsoft. Most smaller software companies were crushed through Microsoft’s unfair competitive practices. If that hadn’t happened, the Floater AI engine would be 30 years further along today, and the current AI hype cycle might not even exist. But it did happen, and Floater AI vanished.

Approximately 13 years ago, I began working on a project to assemble my entire life’s work into a single platform called Reflex. Reflex contains an evolved version of the Floater AI engine. Reflex AI does not resemble the AI being pushed today. The only things they have in common are the letters “AI” and the fact that neither Reflex nor mainstream “populace AI” is actually artificial intelligence. Both are software applications taking vastly different approaches.

Reflex was built using a variety of techniques developed by original software engineers from 1982 to the present. Mainstream populace AI sits on top of a stack of old libraries combined with complex statistical mathematics. Neither model is actually thinking. Reflex AI utilizes an invention that teaches a computer how to understand human behavior. It does not rely on scraped data, does not simulate human thought, and therefore does not suffer from hallucinations. Populace AI, on the other hand, is built on scraped human data stored across the internet—data that was never intended for AI and must be heavily manipulated to fit accessible formats.

The creators of populace AI like to liken their neural networks to the human brain. (I cover this topic in depth on my Substack, linked below.) A neural network may resemble a brain in the mind of a self-proclaimed “father of AI,” but I have never heard that claim backed by anyone working in medicine or neurology. I believe I am uniquely qualified to speak on this.

I began my working life as a marine biologist. At 18, I was chosen over hundreds of qualified graduate students to serve as the research assistant to the world’s leading shark researcher, Dr. Samuel Gruber. I was selected for a reason no one could quantify at the time, but today I’d put it simply: I was a hacker (I figure sh*t out). Today, however, just as the word “AI” has been corrupted, so has “hacker.” If I need a more formal term, I’ll call myself what I am: a Polymath Hacker.

My research involved shark vision and neurobiology. I worked directly with central nervous systems and biological brains. From hands-on experience (on the page detailing that job, you can see a photo of me physically holding a shark), I can tell you unequivocally: my sharks were smarter than populace AI. A neural network is not a biological brain.

There is one more factor that proves comparisons between the human brain and a neural network are invalid. If you ask a top psychiatrist or neuroscientist, “How does human thinking actually work?”—if they are secure in their status, they will answer: “We don’t know.” So how is it that the proponents of populace AI claim to have this knowledge? My substack article

When I moved back to Massachusetts, there wasn’t much demand for an expert on shark vision. With the arrival of early personal computers, I joined the ranks of all the misfits who had dropped out of other professions or didn’t fit in anywhere else, and became a software hacker.

Today, the word “hacker” implies a criminal. But the word “hack” is much older, originally meaning to chop or cut roughly. By the 19th century, it meant making a rough attempt at something. At MIT, a “hack” came to mean a clever, playful, or unconventional solution—something built quickly that made a system do something its designers never anticipated. The earliest documented technical use appears in MIT’s Tech Model Railroad Club minutes from April 5, 1955. Unless adult men playing with model trains is a crime, there was zero connection between hacking and criminality. When I use the word hacker throughout this site, I am referring to those early programmers who built the foundation of all modern technology.

I urge you to consider the following: This site was born when I got my first modem. I decided to put a record of my career online and keep it updated until I retired. The first version lived on an early Bulletin Board System (BBS), followed by FidoNet, then Lotus Notes, Lotus Domino, and finally the web.

Put aside the notion that a person becomes irrelevant past a certain point in life. Imagine one of the original hackers who succeeded in commercializing an invention without venture capital or corporate hand-outs—driven purely by perseverance to make Floater a reality. Venture capitalists used to talk about an entrepreneur needing “skin in the game.” Imagine putting up your home and all your possessions, only to have it wiped out by a megalomaniacal billionaire. Imagine watching that billionaire structure a world where only other billionaires could succeed and raw intelligence was rendered irrelevant. Is it possible that person would spend the rest of his life proving he was right all along?

Now look at the experiences outlined on this website. Do I have the credibility to claim that the technology I built over my career is priceless? Priceless in that even current populace AI is unable to calculate the number of industries where it applies.

The previous paragraphs have been about credibility. I pointed out that claims by populace AI companies about their software actually “thinking” are false. I would be remiss if I did not challenge my own statements about what I will later describe as GRAYBELT AI. I state that much of the platform depends on technology that teaches a machine how to “understand” a person. I’ve already stated that a machine can’t actually understand anything—so how is this possible? The answer, again, is perseverance. Or in this case, maybe an obsession.

The Origin of GRAYBELT Architecture

My earliest invention was called Telejob. It was a dial-up application where job seekers could post their resumes to be matched against companies seeking employees. That is where I first faced the core problem: how do you determine if a person is truly qualified for a particular job? Beyond raw keyword matching, there was no real solution. From 1982 until somewhere around 2020, that problem stayed in the back of my mind.

After 38 years, I had an epiphany. For as long as people have paid other people to perform a job, an employee has never been evaluated by an organization as a biological person. Instead, an employee is viewed as the sum of their skills. Job postings focus strictly on a candidate’s experience and skill set—there is, appropriately, nothing personal about it. A job candidate is represented by a collection of skills: not just the specific skill names, but the actual amount of time spent applying them.

Now we have a text representation of a skill paired with a numeric value. By describing a person to a computer in this way, the machine can “understand” that person in a structured, actionable way. If the tasks that need completion are described using that exact same model, a computer can calculate the probability of success based on the people assigned to the work.

This is a simplified explanation of how this component of GRAYBELT AI works. It isn’t magic or hypothetical—it returns to the cybernetic nature of Floater AI. But because a computer cannot think, it cannot inherently know that one listed skill is simply a synonym for another unless every skill has a fixed name. Solving that specific problem involves using the same underlying mechanics found in populace AI: embedded vector space coordinates that provide the illusion of understanding.

Imagine how many real-world scenarios this invention applies to. (FYI: they are all listed in my patent application.)

Historical Reality vs. Silicon Valley Hype

They say that history is written by the winners, but if you want to know the truth, ask one of the losers. There is nothing but truth on this site—truth recorded at the exact moment it occurred. There is no rewriting of history here.

This is my real-life story. And it’s not going to have a happy ending. My life’s work resulted in a technology that is difficult to separate from the hype surrounding modern artificial intelligence. The American people voted for a country controlled by billionaires. The billionaires claim artificial intelligence will replace everyone, that they have invested successfully, and that we should all accept our fate. I don’t know what it takes to make people believe in truth over money, but everything on this site is a truth I lived through.

There is something time-critical you need to understand while reading through this site: it isn’t only about me.

When I heard Jensen Huang’s opening speech at the October 2025 NVIDIA conference essentially claim that NVIDIA invented everything, it was upsetting. It was upsetting because I knew many of the hundreds of pioneers who contributed to the foundational libraries NVIDIA is sitting upon ten layers up. I am the last one I know of who has stayed in the industry; others have retired, and many have passed away. What I am trying to introduce with my company, GRAYBELT Innovations, is not just my own work—it is what I learned working alongside those original pioneers. And when I say original, I mean original: writing in assembly language with zero documentation.

Decisions made back then shaped every software development from 1980 until today. What might be the foundational innovation in software history is now just used as a simple introductory exercise: the “Hello, World!” program.

In early personal computers, there was no single instruction for putting text on a screen. Displaying even a single letter required a complex sequence of machine instructions. A programmer had to calculate the exact video memory location, load the character code into a processor register (like the accumulator), and execute the instructions to push that value into memory. That entire process had to be repeated for every single letter in the output. It made no sense for every developer to constantly rewrite the exact same low-level code, so we created reusable routines. As methods for sharing code evolved, these routines were compiled into libraries, allowing others to display “Hello, World!” without reinventing the machine-code wheel.

Libraries were built on top of libraries. When a coder spent days or weeks solving a brutal problem, they didn’t keep it to themselves—they contributed it to repositories for other programmers to use. Over time, many platforms emerged, with Stack Overflow becoming the most famous. These were contributions we made freely. In those days, working seven days a week was standard, and we weren’t paid for those extra hours. For anyone wondering where all those uncounted hours went: you are using them every time you think AI is helping you write code.

I considered taking down this website because I believe we’ve reached a point where very few people actually understand the full stack underlying modern AI. People believe tech billionaires who maybe wrote a “Hello, World!” program early on, but that’s about it. I would challenge any current tech executive to explain how their underlying stack actually functions.

Which brings us to why this site is still live. Look at GRAYBELT and consider this: this technology is grounded in reality.

I demonstrated this technology to people at the NVIDIA conference and was told it would be of great interest in two years, once people get tired of what they currently call AI. But I am working two jobs while managing serious health issues. If I don’t pass this on, all of this technology goes with me.

Even more upsetting than being told to come back in two years was the fact that these employees were genuinely surprised anyone my age could still be functioning at this level. I am not exaggerating—I was directly asked how I was capable of doing this work. And, of course, I couldn’t get anywhere near NVIDIA’s top executives.

A few weeks prior, I attended an IBM AI conference. I had hoped to speak with an executive and propose how my technology fits into IBM’s existing business. It wasn’t about having an inflated sense of self-importance, but I encountered the exact same reaction as I did at NVIDIA: “How does somebody your age still function?”

Why does this matter? Because this technology clearly demonstrates how it actually benefits humanity. It is not “artificial intelligence.” The term itself is an oxymoron; there is no such thing as artificial intelligence—not yet, anyway. The technology I pioneered began in 1982 with a simple question: How do you teach a computer about human behavior? That question stayed with me throughout my entire career.

The technologies developed by GRAYBELT require human input at critical stages. It does not make critical decisions without human direction, and it does not suffer from hallucinations. It directly contradicts everything today’s AI promoters claim: it doesn’t require massive, energy-draining data centers, nor does it demand endless token purchases. While certain components of the system—which I call cogs—utilize frontier AI to process information, those tokens are spent only once. Think of it like charging a battery once and storing that energy indefinitely.

My goal is to launch projects in developing nations where current automation is displacing workforce jobs, as well as initiatives in the U.S. that genuinely serve people. Frankly, I’ve reached a point where I don’t care what the establishment thinks. The venture capitalists I spent my career doubting would be a welcome sight compared to the baseless claims of today’s tech figureheads.

Is the American Dream dead? It is for anyone without connections in the top 1%. But the inventions featured on this site are real, backed by media coverage from the eras they were built. GRAYBELT wasn’t built by just slapping old technologies together; it was built by combining the fundamental techniques used to create them.

Why does that matter? Consider one invention from my patent application: a communications system designed to route data between servers without any server knowing the IP address of the others. How? By adapting techniques from early Bulletin Board Systems (BBS)—real dial-up era methodology. Why take that approach? Because it renders the system inherently unhackable.

If that sounds like a bold claim, ask how I verified it: I spent three years in Russia working alongside top security researchers and hackers. They admitted they couldn’t find a way around it. People love to talk about “eating your own dog food” in tech. Did any of today’s tech giants risk their lives testing theirs? I did.

I have been dealing with a rare situation, and I would very much like to hear from anyone else who needs to do the same thing.

GRAYBELT Innovations can be viewed as multiple companies. It is not an exaggeration to say that more than 46 years of coding and engineering experience were put into this organization. This was not the intended result of what I started more than 13 years ago. The intended result was an information security product called Reflex.

Untethered by a project manager, solving one problem led to the discovery of another. And by problem, I mean a fundamental hole in technology that was required to create a comprehensive solution. I went from Digital Equipment Corporation to Cullinet Software to Lotus Development Corporation—companies that led the world in innovation. I am programmed not to focus on what currently exists, but to imagine a solution and work backward. I would dare say any experienced software engineer reviewing the Reflex platform would be surprised by all the new technology that has been created.

The situation mentioned above occurred when I had to organize every advancement into a patent application. Reflex breaks down into 15 distinct, unique inventions. I don’t mean changing a few lines of code and claiming it as an invention—I mean totally new concepts.

And these inventions are packaged into what I call COGS. A COG is not like a software library; it is more like an entire system contained within an object. COGS are combined to make new applications—not “apps” in the casual software sense, but entire businesses existing across different market segments. And this is where things get complicated when describing the system to a potential acquirer.

The Acquisition Portfolio: 15 Unique Inventions

The Portfolio Breakdown

• Reflex (CISOware): As stated above, many of these COGS originated in Reflex. Reflex is managed under a company called CISOware and is marketed as an information security and incident response platform. However, it is really a framework at its heart. It has its own version of AI—the evolution of the principles behind the Floater AI engine. My vision of explaining a person to a computer was successfully implemented. The vocabulary used is industry-specific, but by swapping in a different vocabulary, an application for an entirely different industry (where Reflex-like functionality applies) is created.
• GRAYBELT Situations: Built using these same COGS, this application applies to any planned procedure. It does not involve understanding human behavior and is much simpler than Reflex, but it is applicable to almost any organization.
• GRAYBELT HR: Uses a COG called the Universal Translator to understand words used in a domain-specific way. For example, an information security manager might post a job requiring “malware expertise,” while an applicant lists “experience with antivirus.” Current recruitment technology cannot match those two variants of the same meaning. GRAYBELT HR solves this. It is a disruptive technology for human resources, using a complex formula that accounts for skill terminology, years of experience, and the actual percentage of time spent on that skill to calculate a precise numeric rating.
• GRAYBELT GOV: An offshoot of GRAYBELT HR focused on the U.S. government NICE (National Initiative for Cybersecurity Education) framework. It goes far beyond standard compliance mapping. Not only can it translate common terms into NICE terms, but it can also evaluate a specific certification and determine exactly what it represents in terms of hands-on NICE experience.
• GRAYBELT Domains: The platform with likely the biggest potential impact. This is a system for creating domain-specific AI that runs on consumer-grade hardware. This is not a RAG system. It is based on GRAYBELT AI and does not require a Large Language Model to operate. Because GRAYBELT AI does not natively speak in human prose, a small open-source LLM is strictly used when the desired output is a constructed response in English. If raw data is preferred, the system returns it directly for user editing. The system comes complete with hardware specifications.

Global & Sovereign Capability

GRAYBELT Domains can be used by a nation to create sovereign, domain-specific AI and become a major player in the AI category without relying on support from American tech billionaires. I am actively seeking interest from foreign governments.

(Note: GRAYBELT AI is fundamentally different from current frontier AI models. There are no national security-related risks or functions associated with it.)

The rest of this website is all about credibility. I’ve made some big statements above, and you will find the applicable skills and history detailed throughout these pages.

Thank you for your time!