I'm Doug Kilgore, a software architect with more than 30 years of building systems. I show owner-operators and solo founders how to build AI systems that work, and keep working.
Most people take the first, default answer AI gives them, the same one thousands of others get. Three things follow.
You ask, it guesses, you correct it, it guesses again. The AI fills every gap you left with an assumption, and the afternoon disappears.
When nobody puts their own thinking in, everything comes out the same. Generic output nobody trusts, and your customers notice.
AI patches whatever is in front of it. Without a design, every fix adds bloat, more clutter piled on clutter, until the whole thing stops making sense.
The same order a good architect has always worked in. AI just makes the last step fast.
Before anything gets built, the AI asks what a senior engineer would. Your knowledge goes in first.
A short brief: who it's for, what done means, what must never happen. Structure before code.
Now AI builds against the brief. When something breaks, fix the design, not just the symptom.
Everything I use to get AI to build what I actually mean, in a form you can use tonight.
I've spent 30 years building software that has to work, and I've built ten businesses of my own along the way. One channel I started in 2014 has been watched for 8.4 million hours and still earns on autopilot. So I know what running a small business feels like: too much work, no tech team, and no time for tools that don't deliver.
Early in my career, hard problems meant longer hours, and I burned out more than once. AI changed that, but it brought a new trap: it's easy to lose a whole day re-prompting, chasing output that's almost right. The fix is the same one that works for any system: decide what to build and how it fits together before anything gets built. People who give AI real context and a careful plan get great results. The difference is never the tool. It's what you put in.
If you can describe the problem, you can build the system.
The pre-build interview, the question sets, and a starter CLAUDE.md. Use it on a real project tonight.
Get the kitA YouTube channel with essays on where AI is going, real builds shown end to end with what they cost, and updates when something important changes.
Kit subscribers hear firstBring a real project. I help you plan it, design it, and get it built properly. You build; I guide.
See mentoringA membership for people building their own systems: the full method library, an AI tutor trained on my builds, and live help when you're stuck.
In the works, not open yetNo. The interview works in any AI chat tool and needs nothing but your knowledge of the problem. Some builds use Claude Code, and I show every step.
Mostly Claude and Claude Code. The method works just as well with ChatGPT or any AI chat tool. I show what each tool actually costs, and I'd rather you use fewer tools well than collect subscriptions.
Yes. You'll also get an occasional email about building with AI. Unsubscribe anytime.
I'm not currently taking on client projects. I teach you to build it, and keep it running, yourself. If you'd like hands-on guidance while you build, see 1:1 mentoring.
It's how reliable software has always been built, applied to AI. Before anything gets built, you decide who it's for, what done means, and what must never happen. Then you build in the safeguards engineers use for systems that can't afford to fail: checks that actually catch problems, one source of truth for every fact, and fixing the design instead of patching symptoms. The result is a system you can still trust next month, not just today.
Owner-operators who want AI doing real work in their business. Solo founders building their first tool. And developers who want AI to help them ship software that really works, with real results instead of rework and endless re-prompting.
