October 20–21, 2026 | My Journey Into Artificial Intelligence

OMNI BRAIN / BUILDING SOMETHING FROM NOTHING

October 20–21, 2026 | Two Days of Progress

There’s something about creating something from nothing that I absolutely love.

For the past couple of days, I’ve been spending just about every spare minute working away on Omni Brain, my offline artificial intelligence project. And by golly, I’ve been having the time of my life.

It’s funny, really. Here I am, one guy sitting at his computer, working with QB64 Phoenix, trying to build something that resembles an artificial brain. Not just another chatbot that responds to whatever I type into it, but something that can eventually remember, reason, carry on conversations, and perhaps someday become something genuinely remarkable.

Now, I’m not going to sit here and pretend I’ve invented artificial general intelligence. Good grief, I’ve still got a mountain of work ahead of me. But something rather exciting has been happening over these past two days.

Omni is starting to do a few things that, not too long ago, existed only in my plans.

And I think that’s pretty damn wonderful.

Teaching Omni to Connect the Dots

One of the biggest things I’ve been working on is evidence-based reasoning.

The idea is fairly simple. I want Omni to take pieces of information she’s learned and figure out how they relate to one another.

For example, let’s say I tell her that a robot is inside a hangar. Then I tell her that the hangar is inside a facility.

Rather than simply repeating those two statements, I want her to recognize the relationship between them.

DIAGRAM 1 / CONNECTING INFORMATION

ROBOT
↓ inside
HANGAR
↓ inside
FACILITY

Possible conclusion: The robot is inside the facility.

That’s what I’ve been testing with the 3.3D.4 reasoning system.

The interesting part is that Omni can identify these two-link connections and explain how she arrived at a conclusion. She can also recognize that the conclusion depends on the information being correct.

It’s not human reasoning, and I wouldn’t insult anyone’s intelligence by claiming otherwise. But it is an actual reasoning operation taking place inside the program.

For something I’m building in QB64, I’ll happily take that as progress.

What Happens When Omni Learns Something Wrong?

Now, here’s where things started getting particularly interesting.

Suppose I tell Omni that a beacon is online.

A few moments later, I tell her that the same beacon is offline.

What should she believe?

Well, neither statement should automatically win simply because it was entered first or last.

And that’s exactly what I wanted to accomplish with contradiction detection.

DIAGRAM 2 \ CONFLICTING INFORMATION

BEACON STATUS
↓
ONLINE / OFFLINE
↓
TWO CONFLICTING CLAIMS
↓
MORE EVIDENCE NEEDED

During my tests, Omni successfully identified both statements and reported a possible conflict. More importantly, she didn’t simply pick whichever answer sounded better.

She acknowledged that the information disagreed.

This might not sound terribly exciting to someone unfamiliar with programming, but think about how important that is.

An artificial intelligence that confidently gives you the wrong answer isn’t particularly useful. I want Omni to have the ability to say, in effect, “Hang on a minute. Something here doesn’t add up.”

That’s the sort of behaviour I’m trying to develop.

Giving Her a Memory That Actually Lasts

Another major part of my work involved memory persistence.

After all, what’s the point of teaching an artificial brain something if it forgets everything whenever you close the program?

I spent a fair amount of time testing whether Omni could save information, shut down, restart, and recover what she had previously learned.

DIAGRAM 3 / MEMORY PERSISTENCE

LEARN INFORMATION
↓
SAVE TO DISK
↓
CLOSE OMNI BRAIN
↓
RESTART PROGRAM
↓
LOAD SAVED INFORMATION
↓
MEMORY RECOVERED – PASS

And the wonderful thing is that it worked.

The information survived the restart. Even the conflicting beacon statements were still there, and Omni could recognize the disagreement afterward.

That was a particularly satisfying moment.

It’s one thing to have a program demonstrate something while it’s running. It’s another to shut the bloody thing down, reopen it, and discover that the information is still intact.

Of course, there are more difficult storage and recovery tests to tackle in the future. But these particular tests passed in the actual QB64 application, and that’s what matters to me the most.

Two Days of Progress

Here’s where things stand after the October 20-21 development sessions.

DEVELOPMENT STATUS

Feature| Result
Evidence-based conclusions| PASSED
Two-link reasoning| PASSED
Source identification| PASSED
Contradiction detection| PASSED
Explaining uncertainty| PASSED
Knowledge persistence| PASSED
QB64 native startup| PASSED
New 3.3D.5.1 evidence review| TESTING NEXT

These are results from specific tests, not claims that the entire artificial intelligence system is finished.

After completing the 3.3D.4.6 regression tests, I archived the working version and began 3.3D.5.1.

This next part focuses on evidence review.

Essentially, I’m trying to give Omni the ability to examine the information she’s working with and recognize what is missing. If she has one unsupported statement, I want her to know that it needs verification. If she has conflicting statements, I want her to recognize the disagreement and identify what would need checking.

I’ve already managed to get the latest corrected version compiling and running successfully in QB64 Phoenix.

Naturally, I also managed to run headfirst into a compiler error along the way.

And, in another wonderful moment of programming genius, I had an earlier build opening and immediately closing because the required data folder wasn’t where it needed to be.

I had to laugh.

Here I am trying to create an artificial brain, and a missing folder brings the entire operation to a screeching halt.

Such is life.

So, Where Do I Go From Here?

My next job is to complete the remaining tests for 3.3D.5.1.

Once those are finished, I’ll know whether the new evidence-review system is working correctly or whether it needs more attention.

After that, there’s still plenty to do. Better conversations, stronger reasoning, more dependable memory, an improved interface, and eventually a voice that doesn’t sound like some miserable machine reading instructions over a telephone.

And yes, I have rather ambitious plans for what Omni will look like, too.

But none of that matters if the brain underneath doesn’t work.

That’s why I’ve been taking this development process one piece at a time. Build something, test it, find what’s broken, fix it, and move forward.

I want the end result to be something I can actually use, rather than a gigantic mess of code that looks impressive until someone asks it a simple question.

I’m under no illusion about how difficult this project is going to be. There are problems ahead that I haven’t even encountered yet.

Still, I can’t help feeling excited about it.

When I’m at work, I often find myself thinking about what I’m going to tackle next. Sometimes I’d much rather be sitting at my computer trying to solve another programming problem than dealing with the usual nonsense of the working day.

Perhaps that’s what I enjoy most about Omni Brain. It’s something I genuinely look forward to working on.

Outside of my family, very few things have given me this particular kind of excitement in a long time.

There’s something rather special about having an idea in your head and slowly watching it become a functioning piece of software.

Is Omni Brain finished?

Hell no.

Is she intelligent in the way I eventually hope she’ll be?

Not even close.

But can I open the program, teach her information, test relationships between facts, introduce contradictions, and watch her explain what she’s found?

Yes. I can.

And two days ago, I was still working toward some of those very things.

I suppose that’s what keeps me going.

I’m one guy with an old-school programming language, a rather ridiculous ambition, and a determination to see just how far I can take this thing.

And by golly, I’m going to keep at it.

Omni Brain isn’t finished. But she’s coming along, and I’m having a hell of a time building her. Who knows perhaps itll end up like that spooky ABC robot.