AI for dummies
I am not calling you a dummy. But I know you are time-starved so I won’t bury the lede: I may have made $30,000,000 in 30 minutes using AI.
Do I have your attention now?
Well, you need to read this entire story — to the end — so I can prove it. Can you spare five minutes?
I’m serious. Just five minutes.
But don’t take my word for it. Nelson Brown, former AME/News at The Virginian-Pilot, reported to me that, in fact, it took him only five minutes to read this story to the end.
However, there is a passage, clearly marked, that might seem like rough sledding. But you don’t need to read that section. It is merely there as source material for a demonstration you can conduct yourself.
“If your mother says she loves you, check it out,” right?
My fellow journalists: You are the foundation of our democracy and I am merely an ink-stained wretch just like you. But you must snap out of the mental model have about AI.
Because it is the wrong model and it is making you ignore the biggest media story since the launch of the web.
Here is the problem in one sentence:
When I talk about AI I am talking about a working partner inside a governed workflow. When you hear AI you are picturing Google, but faster, or a deep-fake generator.
Those are not the same thing.
You hear “AI wrote an article.” I said nothing about AI writing an article.
You hear “AI is going to replace reporters.” I said nothing about replacing reporters.
You hear “AI is a content toy.” I am talking about AI as a structured, policy-aware production system that can take what is in your head and turn it into a finished, filed, compliant work product in minutes.
You are hearing content. I am describing process.
Let me show you, in detail, exactly what happened to me in less than an hour this morning.
Time: Breakfast at OceanView Diner, Norfolk, VA Thursday, October 30, 2025
Device: iPhone, Notes app
Hands: One on a fork, one on my phone.
Goal: Make AI remember only what I tell it to remember, so it doesn’t slow to a crawl because it is overwhelmed by all my blather.
Here is the actual idea I typed:
[! — warning: rough sledding ahead — ]
AI cannot remember every single word a user says because it is too expensive in CPU cycles and it slows the model which is bad for user experience and bad for cost — so it is a problem on two levels.
So for PersistentMemory™ to work at scale we have to [redacted].
We already do this in GRO (Gated Resource Optimization).
That works for provider-created forms. It does not work for freeform user input. When the user just talks there are no gates. So how does AI know what to prioritize?
If the user’s cache — where it stores all the stuff that creates “memory” — gets too big and the response time slows, [redacted]. That is an expansion of GRO™ and an expansion of Persistent Memory™, so it deserves its own patent.
[! — end rough sledding — ]
That was the memo I wrote over scrambled eggs and orange juice.
PersistentMemory™ is my invention, but it’s only one piece. I designed it to solve the core problem LLMs still haven’t solved.
Granted that original memo, and my summary directly above, are computer science. But imagine ANY task described with this level of specificity. Pretend you are a school teacher:
“I teach U.S. history to 10th graders. Give me a lesson plan that explains the history of Capitalism, from its very beginning (when was that?) then what happened when the Industrial Revolution (when did that happen? What was the thing or things that made it happen) was applied brutally on people then it was reformed (when did that happen? who made that happen ?) then include the quote from Winston Churchill about democracy (I need the exact quote) and explain who Winston Churchill was and cite the reformers (Teddy Roosevelt? I dunno) who helped eliminate child labor and made Capitalism safer for human. I need enough material to keep 10th graders interested for 30 minutes. Make it relevant to them.”
Now, my fellow journalists — copy and paste this paragraph, above, into OpenAI’s ChatGPT 4 or ChatGPT 5 to see what happens. You will see a live demonstration of what an AI can do.
Go to https://chatgpt.com/
Just copy-and-paste the teacher’s paragraph, above, into the text-entry field that says “Ask Anything” then hit “return” or tap the white button in the lower right corner of the text-entry field.
See?
Now, if you really want to s**t your pants, take the text I wrote at breakfast and ask AI to write a “specification” (use that exact word, because it is a legal term of art) to file a Provisional Utility Patent with USPTO (United States Patent and Trademark Office.) Hit “Return,” look at the result, then go change your underwear.
Now, back to my story:
I pasted my memo into AI.
I said, in essence: Turn this into a patent specification, Keep the structure, Use USPTO language, Create an abstract, I will upload it.
In less than 10 seconds AI generated the specification. Not a blog post. Not a headline. A specification in actual legalese I could file.
Five minutes later I had uploaded:
- specification
- abstract
- drawing (a blank document, but required by USPTO)
- and sb16 to the USPTO
From breakfast idea to provisional patent on file with the United States government, in under an hour, for the cost of my monthly ChatGPT 5 subscription, which is $20.
If I had hired a patent attorney the bill would have been around $20,000 and the turnaround would have been weeks.
You can verify every bit of what I just said. And you should before you report it:
- USPTO uses sb16
- Provisional Utility Patent is an actual thing
- Specifications and abstracts and drawings are actual required documents
- Uploading via Patent Center is an actual workflow
You can do it in an hour if the documents are ready, and if you know all the workarounds for all the bugs in the onboarding process at USPTO — which are an enormous stumbling block — and that’s why most inventors hire attorneys to file patents.
So now you have two stories:
1. My first-hand account
2. The two demos you can run yourself
Now the BIG story:
Google, Microsoft, Apple, Meta and everybody else are about to ship personal AI to billions of people.
Let’s be super conservative. For this model, let’s just talk about 1 billion users. I know that Meta (Facebook + Instagram + WhatsApp) has 3.4 billion users. But for the purposes of this exercise, let’s stick with 1 billion users on a single AI.
Google or Apple or Microsoft decides that having a method to let the user tell AI what to “remember” is worth one penny per user per year in saved computing power, faster response, fewer hallucinations and higher trust. One penny, per year, per user.
$0.01 x 1,000,000,000 = $10,000,000.
So what AI just did — for me — is worth at least $10M to Google and $10M to Microsoft and $10M to Apple. And I only spent $20.
I paid $20 for ChatGPT 5 and it generated a $30,000,000 idea that could be protected by the Patent and Trademark Office and 35 U.S.C. § 101, 35 U.S.C. § 102, 35 U.S.C. § 103, 35 U.S.C. § 111, 35 U.S.C. § 113, 35 U.S.C. § 115, 35 U.S.C. § 131, 35 U.S.C. § 151, 35 U.S.C. § 154, 35 U.S.C. § 261, 35 U.S.C. § 271, 35 U.S.C. § 281, 35 U.S.C. § 282, 35 U.S.C. § 283, 35 U.S.C. § 284, 35 U.S.C. § 285, 35 U.S.C. § 287, 35 U.S.C. § 292 of the United States of America.
Am I gonna see a dime? I dunno. No one is taking my calls.
My name is Alan Jacobson. I'm a web developer, UI designer and AI systems architect. I have 13 patents pending before the United States Patent and Trademark Office—each designed to prevent the kinds of tragedy you can read about here.
I want to license my AI systems architecture to the major LLM platforms—ChatGPT, Gemini, Claude, Llama, Co‑Pilot, Apple Intelligence—at companies like Apple, Microsoft, Google, Amazon and Facebook.
Collectively, those companies are worth $15.3 trillion. That’s trillion, with a T—twice the annual budget of the government of the United States. What I’m talking about is a rounding error to them.
With those funds, I intend to stand up 1,400 local news operations across the United States to restore public safety and trust. You can reach me here.