Skip to content
Subscribe
The storyBuilding since 2015Five chapters

From $200of fabric.

To a company run by agents.

I co-founded a vertically integrated manufacturer and apparel brand. Now I'm building the agents that run it, and writing down everything I learn along the way.

Start with chapter one See the ledger
Illustration of a sunlit microfactory with UV printers and printed fabricIllustration
23,722
code changes in 2026,
most of them built
with AI agents beside me.
Chapter one · 2015

The fabric

Illustration of folded printed fabric, scissors and a home sewing machine on a small tableIllustration
$200

That's what we started with. Fabric, an apartment, and an Etsy shop.

My co-founder and I hand-sewed the first pieces ourselves. I learned the business by doing every job in it: making, shipping, answering customers, and figuring out what to make next.

Off the clockThese days I ride the coast of North County San Diego with my family.
Chapter two · 2017 onward

The factory

We didn't hand the making to someone else. We built a microfactory in San Diego.

Design, UV printing, cutting, sewing, and shipping, all in one building. Every operating decision was ours to make, which made it a good place to test better systems.

One roof, the whole chain

Vertically integrated
01

Design

Prints and patterns made in-house

02

Print

UV printed on fabric in-house

03

Cut

Cut to size from digital patterns

04

Sew

Sewn to order

05

Ship

Shipped from the same building

06

Sell

Sold direct, online

  • Illustration of a UV printer head printing a stained-glass pattern onto white stretch fabricIllustration
    Print
  • Illustration of garment pattern pieces cut from printed fabric on a digital cutting tableIllustration
    Cut
  • Illustration of an industrial sewing needle joining butterfly-print fabric to black meshIllustration
    Sew
  • Illustration of finished festival garments on a steel rack, ready to shipIllustration
    Ship
+38%

Revenue per employee

2024 · my company
42%

Less manual operations time

2024 · my company

Cash tied up in inventory

Results from my own company, from its own records. The inventory drop came from better forecasting and operating decisions.

Illustration of a network of agent clusters routing work along hairline paths to a few red decision pointsIllustration
The turn

Then the craft became code.

The same instinct that built the factory now builds software. The printer still runs. Now agents run the work around it.

Chapter three · 2026

The machines

I've been building software for years. This year the work flipped: now I build every day with AI agents beside me.

These numbers come straight from my own work, as of the latest ledger update.

23,722

Code changes in 2026

3,636

Updates shipped live

700+

AI agents and automations

Code shipped per month

02,0004,0006,0008,000Mar: 1,419 code changes1,419MarApr: 1,386 code changes1,386AprMay: 1,427 code changes1,427MayJun: 1,383 code changes1,383JunJul: 2,655 code changes2,655JulAug: 3,292 code changes3,292AugSep: 4,238 code changes4,238SepOct: 7,355 code changes7,355Oct0k2k4k6k8kMar: 1,419 code changes1.4kMarApr: 1,386 code changes1.4kAprMay: 1,427 code changes1.4kMayJun: 1,383 code changes1.4kJunJul: 2,655 code changes2.7kJulAug: 3,292 code changes3.3kAugSep: 4,238 code changes4.2kSepOct: 7,355 code changes7.4kOct

Code changes per full month in 2026, from March, across every project I work on, each counted once. How I count.

When AI took over

  1. Feb

    The flip

    I moved my day-to-day building onto AI agents. The monthly count starts here.

  2. Mar

    The first full month

    1,419 code changes, the baseline everything after is measured against.

  3. Sep 25

    This story

    The pictures on this page are illustrations, made with GPT Image 2.5.

  4. Oct

    7,355 changes in a month

    The busiest month yet, about 5 times March.

Chapter four · How trust is earned

The ladder

More autonomy means tighter guardrails, not looser ones.

  1. 1

    Agents propose. I approve.

    Every system starts with a person making the call.

  2. 2

    Every decision gets recorded.

    What I approved, changed, or turned down, and why.

  3. 3

    Results get measured.

    Each kind of task has to prove it gets it right, again and again.

  4. 4

    Autonomy widens.

    First it suggests, then it drafts, then it acts with an undo button. One step at a time.

One company, three systems.

Each system runs its own part of the business. They share one rule: the decisions that matter come to me.

Demand

Finding customers

Marketing, ads, email, search, content

Supply

Making the product

Products, inventory, orders, production

Back office

Running the company

Money, people, tax, compliance, paperwork

I make
the callOn what matters

Departments of agents.

7 departments, each with a job, a team of named agents, and everything it has learned so far. One agent in each checks what's missing and asks me before the department takes on more.

Illustration of a quiet microfactory at sunrise with printers, sewing stations and dashboardsIllustration
Chapter five · What's next

The company that runs itself.

Agents find the work, do it, check it, and learn from every decision I make. The ones that prove themselves get more room. I'm building it in the open.

Every system I ship knows when to stop and ask.
Read the build note

Build notes

All build notes →
Pace

A full quarter of building with agents

What 1,900 code changes in three months actually looked like.

ReadQuality

An AI that catches its own mistakes

Building the check into the system instead of hoping for good output.

ReadJudgment

Where my systems stop and ask me

The decisions I keep for myself, by design.

ReadFactory

Running a factory floor from a chat bot

Production updates where the team already talks.

ReadGrounding

Keeping AI honest about what we sell

Locking answers to the real catalog and the real prices.

ReadDay to day

A day working with AI agents

How the work actually gets split between me and the agents.

Read