How I Turned One $15 Python File Into a $0.41/Day AI Sales System — And Why I’m Open Sourcing the Full Stack | by Alan Scott Encinas | May, 2026 | MediumSitemapOpen in appSign up<br>Sign in

Medium Logo

Get app<br>Write

Search

Sign up<br>Sign in

Press enter or click to view image in full size

How I Turned One $15 Python File Into a $0.41/Day AI Sales System — And Why I’m Open Sourcing the Full Stack

Alan Scott Encinas

12 min read·<br>Just now

Listen

Share

By Alan Scott Encinas<br>This is Part 3 of a series documenting how I built Albert, a custom multi-agent AI sales system, from scratch. Read Part 1 here and Part 2 here.<br>Where It Started<br>On March 20th, 2026, I published a Medium article and an open source GitHub repository called ai-sales-agent, and at the time I had no idea it was the beginning of something much larger.<br>One Python file, two API keys, and a plain English system prompt describing my business — that was the entire thing. That first version scored 1,051 contacts in my CRM in a single run for $15.07, surfaced $82,998 in pipeline I did not know I had, and read actual email conversations rather than just metadata to tell me exactly who to call first, in what order, and what to say when they picked up. Within days of following its recommendations I had generated $18,000 in quotes from leads that had been sitting invisible in my CRM for months.<br>Most people would have stopped there and called it done, and honestly that version alone would have been worth the time it took to build. But I could see the next problem clearly and I knew I could solve it, so I kept going — not because I had a product roadmap or a funding round or a team pushing me forward, but because that is what builders do when they see the next door.<br>But seeing the next problem and actually building the solution are two very different things, and the distance between them is where most solo builders quietly disappear. It is because the project grows from a marble into a galaxy, and somewhere in that growth the doubt moves in and the momentum dies. I had been the problem I kept running into, so I built the solution.<br>The Spiral Problem<br>Every builder knows this feeling even if they have never named it.<br>You start with something small and clear, a marble you can hold in your hand and see every angle of. The problem is specific, the solution is visible, the path is obvious, and you are genuinely excited to get started. Then it grows — more context, more edge cases, more possibilities — and new ideas keep surfacing mid-session while the original scope keeps expanding and the architecture keeps getting more complex. The marble becomes a galaxy, and suddenly you cannot hold the full shape of it anymore. The momentum slows, the self-doubt moves in, and you start asking whether any of it is worth it, whether you are actually building something real or just spinning in place.<br>The work is real and the doubt is not, but in the moment you cannot tell the difference, and that is exactly where most projects die.<br>I hit this wall repeatedly across multiple projects — not just Albert, but LunarSite, mobile apps, and everything else I was building alongside the day job. Every time I got deep enough into something to see its real complexity, the spiral started, and every time, the tools I was using either ignored the problem completely or made it worse by offering generic encouragement that had no relationship to what I had actually built. What I needed was not validation. What I needed was facts, and since nothing gave me that, I built something that would.<br>Marble: The Psychological Architecture<br>I built Marble because I was the problem I kept running into, and that honesty about my own patterns turned out to be the most important thing I could bring to the project.<br>Marble is a Claude Code skill I created and open sourced for builders who spiral, and it acts as a conductor for development sessions by keeping you in the current phase, capturing ideas without breaking flow, committing with discipline, and countering self-doubt with actual project metrics instead of empty encouragement. The first day I used it something was immediately different — I did not drift, not once.<br>Every session opens with three lines: where you are, what shipped last, and what is next. No new work starts until that is acknowledged, and while that sounds almost too simple to matter, it changes everything because the reason most projects stall is not lack of ability but loss of orientation. You forget what you finished, you cannot see what is next, and the gap between them fills with doubt until the project quietly dies.<br>When a new idea surfaces mid-session, which happens constantly when you are building something real, Marble does not let you chase it and open a second front. It acknowledges the idea in one line, logs it to a backlog file, and the session continues on the current phase so the idea does not disappear and neither does the current work. The galaxy waits while you finish the...