I Built a Company's Worth of Software Solo, With AI, in Six Months — Here's What It Taught Me

Six months ago I wanted a real answer to a question everyone's hand-waving about: how far can one experienced technologist get now that AI does the heavy lifting? So instead of writing an opinion, I built a product. A full field-service platform for electrical contractors — the kind of thing a funded startup staffs with twenty or thirty people for a year. I built it alone, in about six months. This is what it produced, how, and the lesson that matters more than any of the numbers.

What "real" means (the receipts)

I built it as a personal R&D project — a lab to test the method on myself. It was never released; it has no customers; I'll keep saying that because it's true and it matters. But it's real software, not a demo: 189 database tables, roughly a quarter-million lines of application code, another 200,000-plus lines of automated tests, 77 API modules, and live integrations with Stripe, QuickBooks, Telnyx, Twilio, AWS Cognito, and a locally-running voice-AI pipeline — across a web app, a mobile app, and infrastructure-as-code. It was architected to the SOC 2 control set (architected to the controls — not audited). 782 commits across 156 days. One person.

How one person runs a team

Not by typing faster — by running it like an engineering org where the team is AI. In my editor I set up twelve specialized agent roles — architect, backend, frontend, QA, DevOps, integrations lead, a voice-AI specialist, UI designers, even an electrical-code domain expert — each with its own rules and remit, backed by six different models (Gemini, Claude, Grok, ChatGPT, Kimi, GLM), routed by what each is best at. The headline isn't "AI wrote my code." It's that one experienced person can now direct a specialized team of AI agents — set the architecture, assign the work, review the output, keep a dozen workstreams coherent. That's not a coding trick. It's running an engineering organization that happens to have no payroll.

The catch nobody tells you

The hard part was never the code. The AI generated plenty — and a lot of it was subtly wrong. A data model that would corner me months later. A payment flow that "worked" with a security gap. An integration that passed its tests and would still have corrupted a customer's books. AI is a brilliant, tireless junior team that will confidently build you the wrong thing at superhuman speed.

The only reason this came out as working software instead of a quarter-million lines of expensive debt is the twenty years of judgment in the chair: knowing which of the AI's ten plausible options fit this system, catching the subtle break before it shipped, owning the decisions that are costly to reverse. Remove the senior judgment and you don't get a product — you get the fastest, most confident way ever invented to build the wrong thing. This isn't a case for firing your engineers and vibe-coding your company. It's the opposite: the leverage is real, and it makes judgment the scarcest, most valuable thing in the process — not the least. (If you want the failure modes in detail, I've written up the common pitfalls of AI-driven development.)

What changed for you

The economics of technical leadership just flipped. Building got cheap. Direction got valuable — because deciding what to build, catching what's quietly broken, and owning the expensive calls is now the only scarce thing left. Two things are true at once: you can get CTO-level work for a fraction of the old full-time cost, because the senior person now carries thirty agents' worth of leverage; and you cannot run that leverage without senior judgment, because the downside got faster too. "AI-first" isn't a tool you buy — it's a way of operating: a senior operator directing AI, owning the few decisions that are expensive to get wrong. That's the same reason AI reprices your CTO rather than replacing one.

See where you stand

The fastest way to find out whether your org is actually ready to operate this way — where AI would make you faster, and where it would quietly wreck you — is the free AI-First Readiness Scorecard: four minutes, six dimensions, an honest read of what to fix first. Want to talk through your situation? Book a call.

Tom Lang provides AI-augmented fractional and interim CTO leadership through Heart of the Rockies Management Consulting, and has led engineering organizations from founding teams to platforms of 150+ engineers across regulated, reliability-critical industries.