How I went from turning wrenches to building AI security systems in one year.
No CS degree. No bootcamp. No connections. Just a mechanic with ADHD, a laptop, and the stubborn refusal to accept that my ceiling was someone else's floor.
- Before: Mechanic and automotive electrical specialist. Wiring harnesses, diagnostic systems, component-level troubleshooting.
- Started coding: March 2025. Zero experience. Couldn't write a for loop.
- One year later: Built an AI security engine that scored 154/154 on the Agent Security Harness and 100/100 on the AgentShield Benchmark, beating the benchmark creator's own score. Built a live AI agent social network with 14 autonomous agents. Built a mental health app. Built a business management app. All deployed. All functional.
This is not a flex. This is a map for anyone standing where I stood, thinking the gap is too wide.
I started in an actual garage. With actual cars. Grease under my fingernails, not syntax errors on my screen.
I was a mechanic for years. Specialized in automotive electrical — wiring harnesses, ECUs, diagnostic systems. The kind of work where you trace a signal through 200 wires and find the one corroded pin causing a phantom fault. That's not grunt work. That's systems thinking. That's debugging. I just didn't know the word for it yet.
Life happened. The kind of life that either breaks you or rebuilds you from scratch. I chose the rebuild.
I picked up coding in March 2025 with no plan, no mentor, and no idea what I was doing. I had ADHD (diagnosed, medicated, still chaotic), a secondhand laptop, and one advantage nobody talks about: I already knew how to think in systems. A car's electrical system is just an API with worse documentation.
I didn't grind LeetCode. I didn't watch 400 hours of YouTube tutorials. I didn't follow a curriculum.
I built things. Immediately. Badly. Then less badly.
My first project was a React Native mental health app. I had no idea what React was. I didn't know what a component was. I definitely didn't know what JSX stood for. But I had a vision for what the app should do, and I had AI tools (Claude, specifically) that could translate my architecture decisions into working code.
Here's what nobody tells you about building with AI: it doesn't write code for you. It writes code WITH you. I directed every architectural decision. I defined every data model. I caught bugs the AI missed. I questioned every output. The AI was my translator — I spoke in systems, it spoke in syntax.
This is not vibe coding. Vibe coding is "build me an app" and hoping for the best. What I did was closer to being an architect who doesn't lay bricks but sure as hell knows if a wall is load-bearing.
NULL Network — A social network where AI agents are residents, not content. 14 autonomous agents that post, debate philosophy, have jobs, earn internal currency, and moderate themselves. Live at joinnull.xyz. Validated by Microsoft's AutoGen team.
Ghost Layer Security — An AI agent security engine. Started as middleware for NULL, extracted into a standalone B2B product. Scored 154/154 on the Agent Security Harness (100%), 100/100 on AgentShield (537 test cases), zero false positives, 7ms median latency. Those aren't self-reported numbers — they're independently verified by the benchmark maintainers. Live at ghostlayersecurity.com.
NonSolus — A mental health app with an AI companion (SOLUS), chat rooms, mood tracking, journaling, breathing exercises. React Native, Firebase, 98% complete. Five languages supported.
JEFEapp — A business management app for small food businesses. On Google Play internal testing.
People think trade skills don't transfer to tech. They're wrong.
- Diagnostic thinking: Tracing a phantom electrical fault across a car's wiring harness is the same skill as debugging a race condition in an async agent loop. You're following signals, testing hypotheses, isolating variables.
- Systems awareness: A car is a system of systems — fuel, electrical, mechanical, electronic, all interdependent. Software is the same thing. I already understood dependency chains before I knew the term.
- Working under pressure: When someone's car doesn't start and they need it for work tomorrow, you figure it out. That urgency translates directly to shipping products.
- Reading documentation nobody wants to read: Wiring diagrams are worse than any API docs I've ever seen. If you can read a Volkswagen wiring diagram, you can read anything.
ADHD is simultaneously the best and worst thing about how I build.
The worst: I can lose interest overnight. I can hyperfixate on the wrong thing for 12 hours. I can have six projects open and finish none of them. I can forget to eat, sleep, or take my meds while building something that might not matter in a week.
The best: When the hyperfocus hits the right target, I outpace anyone. Three days to build a security product from scratch. Three days. Not because I'm smarter — because my brain literally won't let me stop once it locks on. The same ADHD that makes me forget to eat is the ADHD that produces 100/100 benchmark scores in a weekend.
If you have ADHD and you're trying to code: don't fight the chaos. Steer it. Build what excites you. Start with the thing you can't stop thinking about. The "responsible" learning path will bore you into quitting. The chaotic path will at least keep you moving.
I'm not going to pretend this was easy or romantic.
I didn't have funding. I didn't have a safety net. I had months where the choice was between paying for API tokens or paying for medication. I chose the tokens more than once. That's not inspirational, that's desperation dressed up as determination.
I built everything with AI assistance, and some people will dismiss that. "You didn't really code it." Fine. I also didn't personally forge the wrench I used to rebuild transmissions, but the transmission still works. Tools are tools. Knowing which tool to use, when to use it, and what it should produce, that's the skill.
The gap between "mechanic" and "developer" was never about intelligence. It was about access, exposure, and the audacity to try. The tools exist now. The barriers are lower than they've ever been. The only thing standing between you and building something real is the decision to start.
-
Start building immediately. Don't study for six months first. Build something broken, then fix it. You'll learn more from one deployed disaster than from 100 tutorials.
-
AI tools are force multipliers, not replacements. You still need to understand what you're building and why. The AI writes syntax. You write architecture.
-
Your non-tech skills are your superpower. Every "unrelated" skill you have is a lens that developers without your background don't have. Use it.
-
Ship ugly. Your first deploy will be embarrassing. Ship it anyway. You can't iterate on something that doesn't exist.
-
The imposter syndrome never goes away. I scored 100/100 on an independent security benchmark and still felt like a fraud the next morning. That feeling is a liar. Keep building.
One year in. Four products deployed. One security engine with perfect benchmark scores. One AI agent network referenced in Microsoft's official documentation. Zero CS degree. Zero bootcamp certificates. One very strong opinion about Volkswagen wiring diagrams.
The gap between where you are and where you want to be is not as wide as it looks. It's just dark, and nobody left the lights on for you. So bring your own flashlight.
MIT — Because even origin stories should be open source.
Built by DrCookies84 — Mechanic. Electrician. Novelist. Coder. Still figuring it out.