Case study

One operator, shipping like a team.

A full iOS product, built and kept shipping by one person. Walk the path with me — five decisions did the heavy lifting, and the leverage did the rest.

~3 wks
to first ship
6
releases → v1.0.5
400+
tests green
~$3.4K
total compute
1
operator

The path

Five decisions. Everything else followed.

A solo build dies by a thousand stalls. These are the bets that made sure no single problem could ever halt the whole thing — and that the same velocity held from the first release to the sixth.

I
Scope as a product decision

The refusal.

Before a line of code: what not to build. No public feed, no leaderboard, no gamification. A surface small enough to finish, and a thesis sharp enough to defend when the shiny ideas came knocking.

II
One swappable seam

The architecture bet.

App and backend meet at a single interface — a real implementation and a fake one behind the same calls. Neither half ever waited on the other, and every feature since walked through the same door. One decision that paid off a hundred times.

III
Decisions before code · dead reckoning

The known position.

The thrash happened on paper — brainstorm, spec, plan, then ship. And the codebase compiled itself into a ground-truth map the agent re-read every session, so context was never re-derived and decisions were never re-argued. The model came from robotics: hand the agent a known position and a heading, and let it dead-reckon the rest. See how it works →

IV
Ship vertical, polish in passes

The cadence.

A real v1.0 first; then dedicated polish — a full re-theme, a home redesign, a Lock-Screen Live Activity. Six releases in. “Done and shippable” beat “perfect and pending” every single time.

V
AI as leverage, not autocomplete

The multiplier.

A senior-pair agent — handed the map and a tight heading, run on a verifiable pipeline that tested both halves before every commit. ~4.6B tokens, ~98% of them cache reads, for ~$3.4K of total compute. A team’s throughput at the cost of a laptop.

What it proves

An operator who executes — and repeats.

Product judgment

Knows what to build, and what to refuse. Scope as a feature.

Systems that compound

One or two load-bearing bets that pay off across every feature after.

Velocity with a floor

Fast, but tested and reversible — no debt that sinks the next sprint.

Capital efficiency

A shipping team’s output for the price of pocket change.

“The goal was never to build impressively — it was to make the right calls quickly, kill the blockers, and ship. Then do it again tomorrow.”

The system that made it repeatable — the compiled “known position” the whole workflow dead-reckons from — in full technical detail.

Dead reckoning: the system, in detail →   Get in touch