Solo AI Builder: 6 Years, 4 Frameworks, No Partners

From Google Ads scripts to 450 commits a month — and what ten years of going solo actually taught me about building alone in the AI era.

Open pocket watch revealing intricate mechanical movement — representing the engine built over six years of AI development

Two years ago, "AI coding" for me was a Python script pulling keyword data from the Google Ads API.

Last spring, my repos were generating ~450 commits a month.

The compressed timeline:

2020 — scripting Google Ads pulls.
2021 — customizing Shopify for a business I'd invested in, wiring EFT data for investing.
2022 — client budget trackers, custom SEMRush tools, more Ads API.
2023 — the first version of my ad optimization platform. Founded Semalytics that September.
2024–2025 — AI coding matured. GPT-4o, Claude 3.5 Sonnet, Gemini 1.5 Pro's million-token context.\

The distance between "an idea" and "a running system" collapsed.
I shipped faster than I ever had. More ideas made it out of my head.\

December 2025 — demo'd at AI Tinkerers: the Concept Clarity System, Claude Project Chat, and a Claude Project pulling 12+ months of personality, engagement, and framing frameworks into one integrated system. Rooms of people who'd seen a lot of AI tools said they hadn't seen the approach before. That landed differently than I expected.

That's the inflection I'll remember.

But the commit history doesn't tell the whole story.

2026 also marks 10 years since I left a comfortable in-house job to bet on a business partnership. I've been in three of those since. All three ended teaching me the same thing: I trusted wrong. I picked wrong.

So I picked myself.

Solo is a better bet — but it's harder than it looks from the outside. Especially when you're building something new. You lose the sanity-checker. The person who tells you the plane is on fire before you fly it into a mountain.

Right now, I feel like I've built an engine. Compact, powerful, tested.

What I don't yet have is the vehicle around it — the shape of the product that finds its market. That's the work of the next stretch.

The one lesson I'd hand anyone: "just do it" is the whole game when you're solo. Nobody hands you permission. But "just do it" also needs a "keep me honest" — and finding that second voice, when you've deliberately opted out of partnerships, is the puzzle I'm still solving.

If you've built alone and figured out the second-voice problem — I want to know what worked.