perpetual student / trial by failure
a boomer who didn't
get the memo about
slowing down.
Learning AI by building things with it. Some of it works. Most of it doesn't, at first. All of it is educational, usually right after it's humiliating. The plan is to keep shipping things until either the approach works or I run out of ideas. So far, neither has happened. This site is the paper trail.
When not feeling personally victimized by how fast AI is moving, I recharge with video games, board games, and golf. All three are genuine passions. All three also make it abundantly clear that enthusiasm and skill are not the same thing. I am very enthusiastic.
∞
failures so far
1
things shipped
0
fucks given about
looking dumb
looking dumb
ai & tech journey
The timeline, unfiltered.
1990
gaming
Nintendo acquired. Everything changes.
One console. One kid. A gaming addiction that never really went away, and an inability to finish games that persists to this day. Parents thought it was a phase. Parents were wrong.
1992
build
First PC at home. A 286MHz. It was terrible.
Ran 2D games that barely qualified as games. Didn't matter. The moment there was a computer in the house, something clicked. This is what actually spawned the love of computers and the tech around them — not wisdom, just a kid who wanted to know how the box worked.
1996
build
First app: D&D character generator in Visual Basic
Built something that actually worked. The curiosity switch flipped to permanently on and hasn't moved since. The RPG hobby eventually evolved into board gaming, because apparently the addiction needed more cardboard.
1997
gaming
200-page Duke Nukem 3D level-building guide
Wrote it. All of it. Posted to a BBS and watched tens of thousands of people download it. First taste of building something other people actually used, and the last time I was humble about it.
1998
build
Building custom PCs for profit
Over 2,000 custom computers built in six years. Not a hobby, a hustle. Learned more pulling cables and swapping CPUs than any classroom would have taught, and nobody gave me homework after.
2010
homelab
Cut the cord. Became a full-time pirate.
Cable out. Self-hosted everything in. Homelabbing starts with minimal knowledge and maximum confidence, a dangerous combination that somehow worked. The internet called it piracy. I called it infrastructure. Metallica never found out.
2014
homelab
Substantial Plex server build
First serious homelab project. Stopped consuming hardware and started building with it. The rabbit hole deepens considerably. Turns out "I'll just set up Plex" is the gateway drug of self-hosting.
2023
AI
Pokes at AI. Decides it's probably nothing.
GPT-4 exists. Plays with it for twenty minutes. Mostly asks it dumb questions to see if it breaks. It doesn't break. Decides it's a neat trick anyway, puts it down, and goes back to real life. Definitely a fad. Idiot.
2024
AI
Watches everyone else figure it out on YouTube.
Spends an embarrassing amount of time watching other people use AI on YouTube while not actually using AI. Knows it's real now. Has no idea where to start. Bookmarks seventeen "beginner guides" and reads zero of them.
2025
AI
Finally jumps in. Immediately feels behind.
Damnit. Discovers Claude and doesn't put it down. Becomes a near-daily habit inside a week. Spends months catching up on things everyone else figured out in 2023, and has the audacity to call it a learning journey. Feels equal parts inspired and stupid on a rotating basis. The feeling of being two years late to the party spawns this site, because apparently the correct response to embarrassment is to document it publicly.
2026 Q1
build
Vibe coding begins
Discovers Claude Code. Discovers GitHub shortly after, mostly by finding other people's repos and shamelessly borrowing from them like a raccoon in a dumpster. First real coding project without a CS degree. The AI does the heavy lifting; I do the steering and the swearing. Turns out "I have no idea what I'm doing" is a valid development methodology. Also turns out vibe coding colonizes every free hour available, including the ones previously allocated to sleep. This is not a complaint.
2026 Q2
build
First app ships
Something real, in production, built by a non-engineer. Spins up a VPS, fumbles through Linux commands with the confidence of someone who definitely knows what they're doing and the search history of someone who absolutely does not. Has Claude handle git commits and pulls because reading the git docs was not on the agenda. Leans heavily on Claude to not burn the server down. Gets it done anyway. The code is definitely not pretty. The server is running though. Security is probably fine. Probably.
now
AI
Still learning. Still building.
The goal isn't to become a developer. The goal is to build things that are useful, fun, and engaging. I don't want AI as just another tool in the stack. I want it as a real force multiplier. The failures are still free.
things I've built
The catalog. Small but growing.
KDM Agent
A personal take on a Kingdom Death: Monster settlement manager, with heavy inspiration from the app Scribe for KD:M. Built for the way I actually play, not the way the rulebook assumes.
site: kdm.filthyhobo.tech
| Frontend | React 18, TypeScript, Vite |
| Backend | Node.js, Express, TypeScript |
| Runtime | tsx |
| Storage | JSON files |
| Proxy | Caddy + Let's Encrypt |
| Deploy | Linux VPS, systemd |
| Packages | npm |