What does 30 years of building the web teach you about the AI moment?
I've worked through every platform shift since dial-up bulletin boards. The pattern underneath this one is the same — and the people who miss it always miss it the same way.
The main lesson is this: every platform shift kills the interface and keeps the need, and the people who get hurt are the ones who confused the two. I've been building online for over thirty years — since before graphical interfaces were standard — and I've now watched the same movie four times: BBS to web, web to search, desktop to mobile, and now everything to AI. The details change. The pattern doesn't.
Where I started
Before the web, there were bulletin board systems — computers with banks of phone lines that people dialed into directly. In the early nineties I ran my own board, Entrepid, on four to six lines out of North Carolina, and I spent years beta-testing Wildcat!, one of the major BBS packages, working booths alongside its makers at ONE BBSCON, the industry's convention. This was a full online world — mail, forums, file libraries, live chat — years before most people had heard the word "internet."
It was also a real business platform. Around 1992 the Texas Department of Agriculture hired me to build a grain-lead database — farmers dialing in for current grain pricing. They sent a representative to my home for a week while I built and tested it. Government data services over a phone line, three years before the browser wars.
And then it all died. When Windows 95 landed, the writing was on the wall for DOS-based boards, and I shut mine down. Here's the part that matters: the need — people connecting, trading, reading, finding each other online — didn't shrink when the BBS died. It exploded. Only the interface died. The sysops who understood that became webmasters and ISP operators. The ones who defended the interface went down with it.
I'd seen how brutal standards fights could be even before that. In the early nineties I was close to TELEGRAPHIX, whose RIPscrip graphics protocol was a genuine contender for how online graphics would work — a standard that, in my corner of the industry, looked like it might beat what became the web's approach. It didn't. Betting on standards taught me early that being right about the need doesn't mean your horse wins the interface war.
Riding the shift instead
So in 1995 I went where the need was going: I founded PRIMELINE Internet Services, the first ISP in Western North Carolina, and ran it until we sold to a private equity firm in 1999. After the sale I built ISPs for underserved rural communities from Indiana to South Carolina — towns the big providers wouldn't touch.
In 2000 my wife and I started WNC Times, a community news site for Western North Carolina. She ran the paper — editor, editorial judgment, twenty-six years of it now. I ran the technology under it, while operating Farrington Development as a boutique agency: web development, SEO, and CRM work, client after client, for over two decades — and hundreds of projects of my own along the way, some shipped, some shelved, some only now coming back to life. Through that stretch I worked through every disruption the industry produced: the browser wars, the SEO arms races, social platforms eating distribution, mobile eating everything.
In 2024, two things happened at once: we moved WNC Times off Joomla onto Next.js, and AI entered my daily workflow. Within two years, the paper my wife had run for a quarter century became the template for NewsroomAIOS — a platform that runs a network of 17 AI-powered local papers, operated by me, alone.
What the pattern says about now
The interface is dying again. This time it's the search results page, and honestly, parts of the traditional website visit itself. AI assistants are becoming how people get answers. If your business is the interface — clicks, pageviews, rankings — you're a BBS in 1995.
The need is exploding again. People need trustworthy answers, local information, and working software more than ever. The question is only who the new layer retrieves and relies on. That's a position you can build toward deliberately — it's most of what I write about here.
The leverage goes to small operators first. Every shift I've lived through handed its biggest early advantage to individuals who went all-in while incumbents held committee meetings. A BBS was one person with phone lines. An early ISP was a few people with racks. Today one person with AI orchestration can run what used to take a staff. That window is open right now, and it will close the way it always does — when the practices become standard and the advantage gets priced in.
And the operator still matters. Thirty years of tools, and not one of them supplied the judgment. The line at the top of my site — always a human orchestration layer — is the through-line of the whole career: the tools keep changing, and someone still has to know what's worth building and answer for what ships.
I couldn't have built what I run now without AI, and I don't pretend otherwise. But I've watched four waves of people mistake a new interface for a new need, in both directions. The need is the same one I was serving over a phone line in 1992. Only the interface changed. It always does.