Why am I doing this?

Beginnings

It seems fitting to write the introduction to what I hope becomes a collection of observations, questions, and occasional explanations on New Year’s Eve, right around midnight. I briefly considered changing the date once the clock struck and I hadn’t quite finished, but decided against it. This felt like the beginning, even if it wasn’t perfectly aligned.

This writing is meant both for the unlikely visitor who stumbles across it and for my future self. I want a record I can return to, to hold myself accountable and to measure not just progress, but how my thinking changes over time.

What This Is

This site is a public record of learning in progress.

I am using it to document an attempt to understand how modern AI systems work by engaging with them directly: training small models, running constrained experiments, reading widely, and reflecting on what I observe along the way. Some entries will be informal and exploratory. Others will be more structured, including experiment notes, hypotheses, and research summaries.

This is not a tutorial series, a product pitch, or a claim of expertise. It is an effort to learn by doing, to make that process visible, and to preserve both the questions and the mistakes that shape understanding over time.

If there is a unifying theme, it is this: I am less interested in what AI can already do, and more interested in how and why it behaves the way it does, and what that implies for how we work with it in the future.

Who I am

I suppose it is important to give those reading this an idea of who I am. It is easy, and common, for outsized and undeserved influence to be granted to voices on the internet all the wrong reasons. I am a 43 year old father of two with my wife. We live in Maine, have an austalian shepherd and two black cats. I manage a delivery team for a large company, I plow in the winter, and now I have rekindled my love of tech with the help of AI. I was a delinquent in my youth, reading Kevin Mitnick and bruteforce hacking my local ISP only to realize I didn't know what to do once I saw the the terminal prompt. I loved reading exploits from the Cult of the Dead Cow, and spent countless hours on IRC finding warez sites and downloding autcad for no reason other than to have it.

That brought me to college for electrical engineering, and that is where the love stopped. I have never enjoyed working without purpose, theory didn't interest me because I wasn't creative enough to apply it and writing and building xor gates on bread boards felt pointless. I ended up taking 7 years to graduate with several long breaks to start businesses and travel on a shoestring budget and rack up debt. I ended with a degree in history from a small state university, got a job in a call center, worked my way up, and enjoyed stability in career so I could enjoy life.

The last two years I have rekindled the passion from my youth as I found a tool, in AI, that finally suited how I learned. {{ I struggled to learn from textbooks and I hate watching youtube videos. In fact I struggled to focus on anything because I decided at 20 that I didn't want to take medicine for ADHD and after resuming at 40 realized how much I had given up. probably removing but i do think there is aplace for this}}. Until then deep learning had been frustrating. I am almost incapable of learning without understanding why I am doing something. It is a major roadblock and it meant that I would spend hours trying to figure out by looking through indexes and flipping through pages to find anything that explained why I should do something one way and not another. I needed to understand what I was doing before I could even think about taking the next step. In my youth, that made it nearly impossible - reading a 500 page linux admins guide to find a single page that told me how to edit the hosts file only to find that I did something wrong and now had no idea how to fix it meant I reformatted a LOT of hard drives - the internet only made it easier to find the wrong answer.

It was working with AI that made me realize how much I had missed out on. I started coding again, actually writing code at first and AI helped answer questions in real time about syntax so I didn't need to keep moving back and forth. I spent a lot of that time writing bad code, throwing it away, trying again, learning more and repeating. I still work on a few things and I am trying to learn to code over time, but what I realized pretty quickly was the inevitability that AI was really replacing the need to write code, at least to write it from scratch or to spend a lot of time to learn and memorize all the syntax, libraries, and best practices. In all honesty this was a disappointment at first. I had thought I would find a niche and write interesting apps and see if I could build something commercially, but it was clear that I would be competing not just with seasoned engineers, but with AI that would always be better, faster, and would never be tired or distracted.

I began with just chatgpt and copy/pasting code and debugging in that manner. When codex arrived, I was initially blown away, but soon realized that codex was really only as good as the person directing it, that the public impression you could just write a prompt and get a working application was incredibly naive. What did work, was learning about systems and methods, becoming steeped in the architecture of system design, project management, and iteration all of which were what I had spent my career learning - how to direct resources towards a goal. I found that documentation and context, keeping it up to date, writing it and editing it to find what agents would understand and build on, made all the difference. It was fascinating, and it was powerful. I realize there will be major disruptions, the world is going to change, drasticaly, and soon. And like before, I couldn't let it go, I needed to know why.

That brought me here. I wanted to train an AI, but had no idea where to start. To be honest, I probably still don't. That is who I am and that is why I am doing this. I want to learn, first and foremost, but I can't learn without a purpose. I need to be doing something that I find valuable and useful. So I am going to begin this journey by making a bold, and probably wrong, hyptothesis. It is based on my experience, some questionable logic, and a gut feeling. This is all to say: I want to understand how AI works, how to build it, how to use it, and how to work with it to prepare for a future in which it is part of the fabric of our lives. I have no delusions about my part in this, I know I am currently unequipped to make predictions, and that my intuitions are based on philosophical thought experiments and a bit of solipcism, but that's the fun part.

Why I’m Writing

I haven’t written anything like this, in substance or length, for decades. My writing since college has been almost entirely functional, shaped by work and utility. While this project may eventually intersect with professional work, the motivation behind it is personal.

I write because I tend to think out loud. Writing gives me a way to capture ideas before they harden or disappear, and to revisit them later when the initial excitement has faded. I want to test my thinking after time has passed, to see how perspective, mood, and understanding reshape what once felt obvious or compelling.

If nothing else, writing these thoughts down may spare my wife from having to endure long, enthusiastic monologues about topics she neither asked for nor cares about. That alone may justify the effort.

What This Writing Is (and Is Not)

This is not a theory, a research paper, or a claim about how intelligence or consciousness works. It is a record of curiosity from a position of not yet knowing enough to know where an idea fails.

I am writing these notes now because this perspective is temporary. As I learn more, the questions will change, narrow, or disappear altogether. I suspect there is value in preserving the questions as they arise, before they are constrained by formal frameworks or existing answers.

How to Read What Follows

What I write here should be read as exploration, not conclusion. These posts will contain observations, questions, partial ideas, and reflections that may later prove incomplete or wrong. That is intentional.

Some entries will be personal. Some will be technical. Some will simply be attempts to articulate why certain questions feel difficult to let go of. They are not meant to persuade, but to document a process of learning as it unfolds.

Questions That Follow (Unanswered)

What changes when learning is continuous rather than episodic?
Does the order and structure of learning matter beyond efficiency?
What role does context play beyond content?
Is “experience” something that can be approximated operationally, even if not subjectively?
What does it even mean to model awareness without anthropomorphizing it?

These questions are not yet testable. That is part of why they are interesting.

What I Do Not Believe

I do not believe current AI systems are conscious.
I do not believe they are alive or possess subjective experience.
I do not believe consciousness has been achieved or is imminent.
I am not claiming novelty, correctness, or profundity.

What Comes Next

This blog will document questions, reflections, and lessons learned.
A separate hypothesis document will define what I am actually testing.
Experiments will be scoped, incremental, and explicitly limited.
Theory, experimentation, and reflection will remain separate artifacts.

This post is the beginning, not the argument.