Blood on the Carpet: The First Car Makers Weren’t Car Makers
A 25-year operator who has never written a line of code, building a real web app with AI. This will not be pretty. There will be blood. On. The. Carpet.
I’m building HostHub, a community app for short-term rental hosts. From scratch. With Claude. In public. New here? [Start where it began.]
Two weeks since the last update. My head was elsewhere. Mainly on two of my kids: my oldest moved from a CPA firm to an energy project company; my middle son got his first job out of school, at an environmental management company.
My own career started when discs were floppy, books were bought in shops, and phones flipped. I couldn’t have drawn a map of where it would go. I can’t draw theirs either.
I wonder how much of what I’m able to do right now — a non-dev operator building a full-stack app with a fistful of AI tools — is going to reshape their paths.
Where we left off: a domain that resolves, a pipeline that ships my code on its own, a database with nothing in it. All wired, nothing built. Time to write the first real code — me at the keyboard with Claude, pasting its work into Claude Code.
Here’s what I did this session:
• Gave HostHub a place to remember its members (built members table from schema)
• Wired the database into the app (added Supabase client)
• Designed how people get in the door (auth flow — Google + magic link)
• Set rules for what I sign off on (added decision-governance tiers)
When the members table went live, the database’s whole reply was: “Success. No rows returned.” Real work. Nothing to show for it yet, though. I’m getting frustrated at this dev — and he is I.
Two moments from this session stuck with me.
We had to fix what I approve vs. what just happens. Designing the login meant twelve decisions in one sitting, Claude with an opinion on each, every one needing my sign-off. A firehose. I caught myself nodding along — and nodding along is worthless. An approval I don’t understand isn’t an approval. So I built in governance logic: what gets explained to me and approved by me, and what just happens without me. Small calls, Claude makes and moves on. Big ones, it stops, asks, and explains until I actually understand what I’m signing. Whether I drew the line in the right place, nobody knows yet.
My learning path isn’t simple → complex. It’s project start → finish. Terminals, version control, database security — whatever the work coughs up next. A computer science student gets fundamentals, then increasingly complex concepts. I get mine when Claude reminds me for the fifth time not to paste strings containing keys in chat. The product is zooming along and I’m running behind it, grabbing what I can.
As I look at what’s ahead — wiring up the login pages, pouring my old spreadsheet of contractors into the database, getting to the first real test where one host adds a contractor and another host sees it — I find myself thinking less about the next step than about the shape of the whole thing.
What does it really mean for a GM, a product manager, and an engineer to refactor themselves into one brain?
A few days ago, I took my youngest kid to the LeMay, the car museum here in Tacoma. One placard: the first car makers weren’t car makers. They were engine manufacturers, and they were carriage designers. Because engines existed before cars, and so did carriages.
As the auto industry scaled, technology improved, and the first car companies grew, those point solutions merged into “auto manufacturers.” The old world of carriage designers and engine manufacturers disappeared.
Maybe that’s the world my kids will inherit. No visionaries, PMs, devs. Just smart people with a new set of skills who, together with some descendant of Claude, create wonderful things.
Up next: the first real test — one host adds a contractor record, another host sees it.
[Subscribe / Check out earlier editions here.]
