I Wrote 28,000 Lines of Code Without Being a Programmer

I'm going to tell you something that would've been impossible five years ago: I've written over 28,000 lines of code across multiple production systems, and I'm not a programmer.

I'm a field sales account executive for Bottleless Nation. I drive around Wisconsin selling bottleless water and ice systems. That's my actual job. The coding happens after hours, on weekends, and during stretches when my brain needs something different than prospecting.

But the code is real. The systems work. And they've fundamentally changed how I operate in the field.

What I Built

JacobOS — a unified field sales dashboard. Territory data, pipeline management, CRM, and an AI assistant called Sebastian — all in one place. This is what I look at every morning to know where I'm going and why.

VISION — an autonomous AI operating system built on what I call the PAUL framework: Plan, Automate, Unify, Learn. It handles service health monitoring for all my systems, workflow automation, data quality pipelines, and route optimization. The long-term vision is a system that improves itself.

Sebastian — an AI assistant powered by Claude that's embedded in JacobOS. It knows my territory. It knows my pipeline. I can ask it questions about my data, get follow-up email drafts, and plan my day — all through a chat interface in my dashboard.

Territory Universe — a territory intelligence engine with scoring across 16 factors, geocoding, deduplication, and data quality checks. It started as a separate system and eventually got consolidated into JacobOS.

All of this runs on a single Linux server. No Kubernetes. No complicated cloud infrastructure. Just systemd services, SQLite databases, and straightforward Python and Node.js.

How AI-Assisted Development Actually Works

People hear "I built it with AI" and assume I just told ChatGPT to write an app. That's not how this works. At all.

AI-assisted development — at least how I do it — is a collaboration. I'm the architect. I know what I want the system to do, how the pieces fit together, and what the constraints are. The AI handles implementation details: writing the functions, debugging the errors, suggesting patterns I wouldn't have known to look for.

I use Claude as my primary development partner. Here's what that looks like in practice:

1. I describe the problem. Not "write me a function." More like "I need a scoring system that ranks territory prospects across 16 factors including ICP fit, recency, geographic clustering, and seasonal patterns."
2. We iterate. The first output is never the final product. I test it, find edge cases, push back on decisions I disagree with, and refine until it actually works with real data.
3. I maintain context. The biggest challenge in AI-assisted development is keeping the AI aware of your full system. I've built persistent context systems so that Claude understands my architecture, my database schema, and my design decisions across sessions.
4. I make the decisions. When there's a tradeoff — performance vs. simplicity, flexibility vs. speed — I decide. The AI doesn't know my priorities. I do.

The result is code that I understand, that I can maintain, and that solves problems specific to my work. It's not a black box. I read every line. I just didn't type every line.

Why a Sales Rep Needs This

Field sales is operationally complex in ways that people in offices don't always appreciate. I cover a large territory. I manage a pipeline of prospects at various stages. I make dozens of decisions every day about where to go, who to call, what to prioritize, and when to cut bait on a dead lead.

The tools available to field reps are designed by people who've never done the job. They optimize for reporting — for managers who want to see dashboards. They don't optimize for the person in the truck who needs to know: where am I going next, and why?

That question — "where do I go?" — is the single biggest failure point in field sales. When you don't have a clear answer, entropy wins. You drive to the easy stops, skip the high-value ones, waste time on leads that were never going to close, and end the day feeling busy but not productive.

My tools answer that question. Every morning, JacobOS tells me where to go based on territory scoring, pipeline status, geographic clustering, and a dozen other factors. That's not a nice-to-have. That's the difference between a good day and a wasted one.

The Background That Made This Possible

I'm not starting from zero. I have a web development background — I studied it at Madison College and built websites at UW-Madison. I made award-winning educational sites that were adopted by schools in Singapore. I understand HTML, CSS, and the basics of how web applications work.

But that was 15 years ago, and the world has changed. What I'm building now — full-stack applications with AI integration, real-time data pipelines, automated workflows — would've been way beyond my skill level without AI as a development partner.

The technical foundation helped me think like a developer. The AI let me build like one.

The Compound Effect

Every tool I build makes the next day in the field slightly better. Every data point I capture makes the scoring slightly more accurate. Every automation I ship gives me back time to spend on the work only a human can do — building relationships and closing deals.

This compounds. After months of building and iterating, the system I'm working with today is dramatically better than what I started with. And it'll be better again next month.

That's the part most people don't think about with AI-assisted development. It's not just about building something once. It's about building something that learns and improves. Every interaction, every decision, every field day adds to the system.

I'm one sales rep. But with the right tools, one rep can operate at the capacity of a team.

That's not futurism. That's what I do every Tuesday.