In 2025, alongside my day job running engineering at OmniMetrix, I founded NexSpark — a human-driven, AI-assisted matchmaking service. A values-based quiz builds a structured profile, an AI pipeline shortlists compatible matches across identity-verified members, a human reviews every pairing, and members get anonymized match reports with identities revealed only on a mutual yes. It launched in Atlanta at nexspark.com, and the same engine now powers NexSpark Friends at nexspark.org, matching people into friend groups of four to six.

I built all of it solo: product, matching pipeline, verification flows, the marketing sites, the content operation. Twenty years ago that sentence would have described a five-to-eight person team and a seed round. It's worth being precise about what changed — and what didn't.

AI collapses the specialist gap, not the judgment gap

A solo founder's real constraint was never hours; it was breadth. Every product needs work in disciplines you're mediocre at, and mediocre work compounds into a mediocre product. That's the gap AI actually closes. With AI assistance, my design output stopped looking like an engineer's design. Legal drafts, SEO strategy, copywriting — each went from "hire someone or do it badly" to "do it competently in an afternoon and know what questions to ask."

What AI didn't change: knowing what to build, what to cut, and what "good" looks like in each of those disciplines. Twenty-five years of shipping gave me a trained eye across enough domains to review the work AI produces. The leverage isn't the generation — it's the review. A founder who can't tell good output from plausible output just produces bad work faster.

AI made me faster in every discipline. It made me better in none of them. The judgment still has to come from somewhere.

Put AI in the product where it's strong, and humans where it's weak

The same principle shaped the product itself. Dating apps optimize for engagement — endless swiping is the business model, not a flaw in it. NexSpark's bet is the opposite: AI does what it's genuinely good at (scoring compatibility across structured profiles, both directions, at scale), and a human does what AI is bad at — the final read on whether two specific people should actually meet. Every pairing gets human review before a report goes out. That's not a limitation we apologize for; it's the product.

Solo doesn't mean no process

The most counterintuitive lesson: the disciplines I'd built for 150-person organizations mattered more alone, not less. Deploy checklists, QA runbooks, editorial calendars, decision documents. With no colleagues to catch your mistakes, process is the only reviewer you have. I write specs for an audience of one because the alternative — holding the whole system in my head — stopped scaling somewhere around the second product.

What this means beyond my two little sites

Whatever the eventual outcome of NexSpark the business, the pattern is transferable, and it's the same one I bring to leading engineering teams: AI doesn't replace the team — it changes the minimum viable team, and it moves the leader's job further toward judgment, taste, and review. The organizations that figure out where humans belong in the loop will beat both the ones that refuse AI and the ones that trust it blindly. I've now run that experiment at both ends of the scale — a 150-person IT organization and a company of one — and the lesson is the same at both.