01-projects/personal-brand/ship30

tagapet drafts

2026-11-08·source: founder closet clean-out photo (iMessage 2026-11-08 11:34 ET)

TagaPet in a plastic bag

Cleaned out a closet today and found my first startup in a plastic bag.

TagaPet, 2013. NFC and QR pet tags: someone finds your dog, scans the collar, and the dog gets home. Lost. Scan. Found.

The idea held up. The inventory didn't. We bought tags wholesale before we had the demand to move them, and that cash sat in bags like this one until the resin turned yellow.

Thirteen years later I build AI agents for a living, and I keep relearning the same lesson: prove someone will pay before you stock the shelf. With today's tools, that test costs a weekend instead of a pallet.

What's the artifact from your first venture you still can't throw away?

[CHECK] scan flow (what happened after a scan), "wholesale inventory drained capital" sourced from SC-020 (07-archive/newsletter-2021/sc-020-analytics-antenna.md); Amsterdam accelerator detail left out on purpose (optional add).

v2 (founder edit + combined para 4, 2026-11-08 17:20 ET)

I did some spring fall cleaning this weekend and found my first startup in a plastic bag.

TagaPet, 2013. NFC and QR pet tags. Someone finds your dog, scans the collar, and you're immediately in touch with a location. Lost. Scan. Found.

The idea held up. The inventory didn't. We bought tags wholesale before we had the demand to move them, and that cash sat in bags like this one until the resin turned yellow.

Thirteen years later I build AI agents for a living, and I watch teams make the TagaPet mistake in a new costume. They want token budgets, cost controls, and every connector approved before they build anything. That's buying the pallet before you've sold a tag. Optimization comes after demand, and iterating has never been cheaper. TagaPet's lesson cost a pallet. Today the same lesson costs a few tokens.

What's something from your first venture or side project you still can't part with?

Voice notes from his edits: he swapped the abstract lesson for his own domain specifics (tokens, cost controls, connectors); a "prove demand before maturity" framing; he rejected the pure then-vs-now line.

v3 (2026-11-08 17:30 ET)

Para 4 → "Thirteen years later I build AI agents for a living, and I watch teams make the TagaPet mistake. They want cost controls and every connector approved before building. Optimization comes after demand, and iterating has never been easier. TagaPet cost us a pallet. Today the same lesson costs a few tokens."

His line edits (candidate WRITING-rdco rules, confirm after posting): no costume/metaphor filler; no near-synonym pairs (token budgets + cost controls); save the payoff word for the last line; a strong image is used once; shorter phrasing beats complete phrasing.

v5 (final candidate, 2026-11-08 18:45 ET)

Para 4: "Thirteen years later I build AI agents for a living. People envision a recursively improving software factory agent swarm working on their project. That's possible, just not first. Get it working, then add cost controls and connectors once it earns them, then build the hill-climbing improvement loop." Para 5: "Optimization comes after demand, and iterating has never been easier. TagaPet cost us a pallet. Today the same lesson costs some tokens."

Lesson he settled on: don't jump to the advanced/scaled setup; crawl-walk-run. He questioned the analogy himself ("Is our TagaPet problem really the same?"). Rule candidate: the analogy must match the mechanism (skipping steps), not just the surface (spending).