AI and Me

To understand my approach to AI, you need to know a little about my history.

My first home office in 1973 while still in college. Owning an IBM Selectric typewriter was a big deal.

In 1979, four years out of college, I'd turned a freelance photography and writing income into a recognizable business — corporate presentation slide shows. I wrote the scripts, took the pictures, assembled the audio track, and synchronized the slide projectors with a tape deck. Ten years later it was the largest corporate presentations company in the region, with thirty employees.

But in 1979 it was just me, and I had a problem. When a client changed something on page two of a script, I had to retype ten to twenty pages. I needed a better solution. There was a new technology called word processing. Xerox was a player, and a Colorado startup of former IBM people had just launched a word processing company called NBI — which stood, honestly, for Nothing But Initials.

I bought an Olivetti-branded version: dual 8-inch floppy disks, a Xerox daisy wheel printer running thirty characters a second, and an acoustic hood to muffle the noise because it was that loud. A week of classes was required just to learn how to use it.

The cost: $18,000. My new Volvo had cost $6,000, and the Olivetti payment ran nearly as much as my mortgage. In today's dollars, that's $78,000. The IBM PC was still two years away. The Mac was five.

That machine made a monumental difference. It saved me hours of retyping, and more importantly, it let me produce proposals that stopped people — right-justified text, comb-bound, one font, and still impressive enough to close business.

As the company grew, I made sure every employee had a computer. I'd learn the new tools at night and teach the staff during the day. That relentless pursuit of a competitive edge paid off when I moved us from corporate presentations into the early dial-up days of the internet. We became a public company, built the first location mapping on the internet — 275,000 Visa ATM locations — and grew to two floors in downtown Denver with over a hundred employees.

Fifty years of firsthand, write-the-check experience with technology. I've watched it arrive and evolve in waves, but nothing has come on as fast or as profoundly as AI.

I remember when I stopped looking up phone numbers in the white pages and switched to AltaVista. And today I've moved from keyword searches to natural language prompts. So when I first introduced myself to ChatGPT in a Walmart parking lot, waiting for my wife, I knew exactly where I was standing. I've never looked back.

Arthur C. Clarke's third law has stayed with me since I first read it: Any sufficiently advanced technology is indistinguishable from magic. I'm a hard case when it comes to technology, and that one still holds.

The more I've written, the more I've found ways to integrate AI into the workflow. My current platform is Claude. Learning to use it well is like teaching yourself piano — anybody can press a key and make a note, but creating something worth hearing takes something more. Uniquely and humanly more.

That required changing how I think about the tool. I can't tell Claude "read this and tell me what you think." I have to be specific — "evaluate this against these four criteria" — the same way you'd brief a smart colleague rather than ask a vague question and hope for the best.

I asked Claude to evaluate all five books of the GW Canyon series — 435,000 words. We worked through the criteria together first, back and forth, the way you'd think out loud with a co-worker to sharpen the brief. Then Claude went away for fifteen minutes and read, deeply read, all five books. Ten pages of notes came back. Because the front-end criteria were strong, the response was too.

If literary agents or publishers claim they're not using AI to evaluate manuscripts — I'm looking at you, slush pile — they're either lying or falling behind.

What AI engines like Claude offer is what I'd call contained intelligence. Remarkably capable within the context and goals you define. But Claude's analysis, however useful, cannot carry the same understanding of five novels that I carry. It doesn't have my frame of reference — the fifty years, the specific memory of why a scene was written, the instinct for what a character would actually do. That's not a limitation to fix. It's just the nature of the tool.

For some things AI writes well. With hundreds of thousands of words as sample, it can approximate voice and style. But I still see the seams. No AI engine can invent a new chapter or generate the architecture of a whole book. Maybe someday. Not now.

What it has done is make my writing better by example. I can feel a new economy in my prose — learning, through AI commentary on my own work, to trust my readers more. The way a great literature professor teaches by holding the standard and letting you find your way to it.

Nobody can predict what comes next, and I've run out of patience for critics who pronounce with very little actual knowledge. What I know is that I've found a tool that helps me be more of me.