I started programming when I was 13 years old.
The software I wrote for Mister Dawson on a TRS-80 Model 1 with phonetic voice synthesizer was cutting edge for its day, professionally produced and used academically if not commercially to help remediate English with students who were struggling.
I remember how hard it was to step away from the machine and let other kids have their chance at what I had created for them. Without ever really realizing that was what I was doing. The dopamine rush of puzzle solving had already swept me up, long before I started regularly drinking every programmers favorite brown liquid, coffee.
Boolean math, English grammar, sentence structure, spelling, phonetic word structure and programming logic with its required day-dream like imagination of execution all got sweetly mixed up in my still rapidly developing teenage brain.
It was just the start of the journey.
I mowed lawns for an entire summer to pay for a TRS-80 Model 3 my freshman year of high school. By the following year both Apple and TRS-80s were appearing in our classrooms. I remember showing another student how to make his first self-replicating program in the boot sector of a floppy. We didn’t have the word “virus” yet.
I designed my first neural net around 1985 while using my TRS-80 to help my dad with his work as a produce manager. I remember it was a deep network, about 12 cells deep with floating point weights and using the Tan function. It worked so well it literally scared the hell out of me. Because I didn’t entirely understand why it worked.
I do now.
Eventually Silicon valley would pull me in. Lockheed, Physics International, the Apple campus, Akashic memories, IBM (of course) and even Intel were abodes I visited or even got to know somewhat. But it was a brief whirl-wind tour of a few years before I got burned out on high wages and alcohol fueled weekends.
I returned home to the Rogue valley humbled and a little … stretched out.
I’m sure it was quite the mirror of the Prodigal son story. I’m also sure my dad would have loved that comparison.
It didn’t take long for me to pay dad back and to buy myself a printer to go with my rapidly aging TRS-80. IBM had just released their standard and the world was on the precipice of change. I was loving it, generating black and white fractals of all sorts with my 9 pin dot-matrix printer. I used that thing up in less than a year covering my walls in beautiful errors and intentional mundanity.
I also started helping people with their computer problems.
That was ‘91. ‘92 saw me laser copying purple ink on yellow paper to give my fractals more pop! I did that at the request of my girlfriend at the time, Alice. What a sweety she was putting up with my nonsense and not only that actually providing interested artistic input.
‘92 also saw the release of winsock and the general availability of affordable phone modems. By 94 I was designing my first web pages, CGI scripts in Perl and offering a variety of programming services in my community.
By 96 I had conceived of the ideas that would grow into an attempt at what back then wasn’t even called a startup. DUCER was the first real predecessor to what is now known as OpenRouter. Doug, Steve and I designed a distributed system for training neural nets using spare CPU cycles. Like SETI@Home or Folding@Home only we spent 6 years getting nowhere with it while selling short-term services just trying to keep our head above water.
That sort of thing kinda stings.
My addiction to problem solving meant I was going to stay in computer services and my boundless itch of curiousity meant even through failure I would keep exploring wild ideas in software.
Time flies, until you aren’t having fun. That’s why I remember clearly having the bitcoin miner on my laptop while my dad was having a CT scan for his pancreatic cancer at Oregon Health & Science University in Portland. My brother called me up that day because his printer driver wasn’t working. Frankly I was grateful for the distraction. My laptop didn’t really have the power to mine by that time, all the easy coins had already been had.
We lost dad in late 2011.
In 2012 AlexNet's success demonstrated the power of GPU acceleration for deep neural networks. And I was back in the saddle again!
By the time Deepdream was released I was fully prepared for that madness! Style transfer popped up on the heals of that and all my love for fractal weridness found a new home.
It was rapid fire after that. 4 more years flew by and I was fine-tuning GPT-2 with the works of Mark Twain, the great philosophers of world history, and all the religious books of the world. That was fun and weird.
In the background the usual grind of printer drivers, website design and repair, standalone apps, hacklabs, the Day of Code at Zeal and etc. kept providing a steady supply of loose pocket chain and dopamine to compliment the summer work I did running the family produce farm.
COVID was horribly destructive here. Our tractor had just broken down and you can’t finance new equipment if you aren’t open to make money. I did get a brief uptick in software work from people now stuck at home trying to make their computer work.
It was around this time I started working with VQGAN+CLIP and eventually Stable Diffusion to create wild and amazing videos. Some of my favorites were made from the textual content of physics and biology journals. (You can check those out here: Guy's Wild Diffusion Animations )
The ability of completion models to generate code came as a lightning bolt in the dark of night. And now we slow down to observe. In the last year there has been an uncertainty that I never experienced nor expected to experience in the software industry.
In 2024 I landed a few small jobs using deep learning techniques but in 2025 there has been a deathly silence from the entire field. I don’t think it is just the availability of LLM coder assistants aka Agents. I think it is a complete reshuffling of fundamental paradigms that has some elements of chance, like a dice game but no one knows the rules.
Every month of every year since 1978 I have spent time reading about, studying, experimenting with and writing software code of some kind or another. I’ll probably do that till they lay me in the grave if not beyond.
I’ve never been more thrilled by or concerned about what software is becoming.