← All work

Course Code · 2026 to present

An AI-native consumer product.

Built solo with an AI co-founder, as a deliberate experiment. No Figma. No wireframes. No mocks. Just brief, prompt, judge, ship. 18K+ courses indexed. 5.2K users in 30 days. 100% organic growth. And one question driving all of it: what's load-bearing, and what's anxious?

Role

Founder · Product · Design

Duration

10 weeks

Team

One human, one AI

Status

Live · Phase 1 shipped

Live, as of May 2026

5.2K

Active users (30 days)

18K+

Courses indexed

+96%

MoM user growth

$0

Paid acquisition spend

By the way

0 Figma files. 0 wireframes. 0 mocks. 0 prototypes. The artifact stack collapsed by design.

The experiment

What's load-bearing?
What's anxious?

Course Code is a real consumer product. It is also a deliberate stress test on the design discipline. How far could AI take a design role if I only stepped in for judgment?

The hypothesis

AI can execute.
Taste cannot be delegated.

The rule was specific. No Figma. No wireframes. No mocks. No prototypes. No spec docs translating intent for engineers. Brief, prompt, judge, ship.

The test: which artifacts in the design process are load-bearing, the ones the design itself depends on, and which are anxious, the ones we produced to protect against a handoff failure that hadn't happened yet?

The verdict is below.

Anxious

Produced to defend against a handoff failure. Translation insurance between stations.

What collapsed

  • 01 Wireframes
  • 02 Hi-fi mocks
  • 03 Prototypes
  • 04 Handoff specs
  • 05 Mock-explainer decks

These existed because each station didn't trust the next. AI dissolves the handoff. The defenses can come down.

Load-bearing

The design itself depends on these. They cannot be delegated, automated, or rushed.

What stayed

  • 01

    Taste.

    Knowing which one of ninety-nine ships. The years-of-staring-at-pixels skill.

  • 02

    Judgment.

    The calls about what the product is, who it is for, and what it refuses to be.

  • 03

    The brief.

    The new artifact that carries intent directly into the build. The brief IS the design.

Everything else compresses. These three got more important, not less.

The cost

AI collapsed the stack.
It did not collapse the work.

Course Code is live. It is not pixel perfect. The screenshots on this page have spacing I would still fix, copy I would still rewrite, edge cases I would still polish. What you are seeing is the result of hundreds of hours of judgment calls AI could not make for me.

The artifact stack existed because each station distrusted the next. Wireframes protected designers from their own future selves. Mocks protected engineering from interpretive ambiguity. Specs protected the build from edge cases. Every artifact was insurance against a handoff failure. AI dissolves most of those handoffs by carrying intent directly into code. The defenses become unnecessary. The irreducible part of the work steps forward.

Every prompt landed because the brief was tight. Every output shipped because I rejected the ninety-nine that were not good enough. Every screen I kept earned its spot against three I threw out. AI gave me execution at the speed of one. Taste did the rest.

The new workflow does not make designers less important. It makes us more important. The work that is left is the hardest work designers do.

What does not collapse

01

Taste.

Knowing which one of the ninety-nine ships, and refusing the rest.

02

Judgment.

The hundred small calls about what this product is, who it is for, and what it refuses to be.

Everything else can compress. These two cannot be delegated, automated, or rushed.

That's the thesis. Course Code is the proof.

What follows is the product itself. The problem it solves, the insight that shaped it, the bet it ran on, what shipped, and what the numbers say.

The problem

Nobody was solving this.

Women face a quiet, universal golf problem. Showing up at a new course without knowing the dress code. The rules are real, enforced, and the difference between welcome and turned away at the bag drop. But they're buried in PDFs, missing from websites, and never written from a woman's perspective.

The insight

Style intelligence, not a directory.

The opportunity wasn't "here are the rules." It was: here's how to show up as yourself, within the rules, on every course you'll play. Utility earns the right for identity. The lookup is the hook; the Style Planner is the product.

"Know before you go."

The bet

Solo, with an AI co-founder.

Conventional path: hire a developer, build for six months, launch. Course Code went different. A solo founder with the right AI infrastructure can operate at the output level of a small team. Not AI as a tool. AI as a co-founder with authority.

The product

Built and shipped.

coursecode.golf →
Course Code marketing homepage

Marketing home · Know before you go

Course Code detail page for Sedona Golf Resort

Course detail · The Vibe + The Code + What to Wear

Course Code community page

Community · What women are wearing

Course Code pricing page

Pricing · Know the rules. Own the look.

Free + Pro

Utility for everyone, intelligence for Pro.

Free

$0

Always free

  • Dress code lookup, 18K+ courses, all 50 states
  • Course vibe ratings: casual, smart casual, traditional, strict
  • Policy details: collars, sleeveless, denim, leggings, hats
  • Public Look Books, shareable by link, viral loop
  • Verified courses with badge indicators

Pro

$39/yr

$6.99/mo · 7-day free trial

  • Everything in Free
  • Style Planner. 4-step wizard. Course + date + occasion + style with real-weather data.
  • Look Books. Personal golf style archive. Save, organize, share, discover.
  • Style Search. Search shoppable products. Filter, save to Look Books.
  • Real-weather styling. Forecast for your tee time, not generic seasonal advice.

The stack

Modern, fast, and AI-native at every layer.

Next.js 16

React 19, TypeScript

Tailwind CSS 4

Design tokens, utilities

Vercel

Auto-deploy under 2 min

Supabase

Postgres + Auth + RLS

Stripe

Pro plan, trial, webhooks

PostHog

Behavioral analytics

Sentry

Error monitoring

Klaviyo

4-email nurture sequence

OpenClaw + Claude

AI co-founder runtime

@anthropic-ai/sdk

Production dependency

Linear

Source of truth for work

n8n

Self-hosted automation

Frontend Backend / Data AI Ops

The numbers

Growth is working. Conversion is the next problem.

01 · Growth

The organic story.

18K+

Courses indexed

5.2K

Active users · 30 days

+96%

MoM user growth

$0

Paid acquisition spend

02 · Acquisition

The funnel is filling.

339

Total accounts

239

New accounts · 30 days

76

New accounts · 7 days

Up

Sign-up rate accelerating

03 · Conversion

The next problem to solve.

5

Pro subscribers ever

0

Active subscribers today

1%

Upgrade-prompt CTR

Next

Funnel rework underway

The growth story is real. The conversion story isn't yet. The upgrade prompt shows 101 times for every 1 click. A message, timing, or placement problem that needs a rebuild, not a tweak. Diagnosing the funnel, then fixing the prompt, is the next milestone. Founder mode means owning what's working and what isn't.

Source: Supabase + PostHog + Stripe + Google Search Console, live data May 27, 2026.

Organic, by design

The product is its own marketing.

64.6K

Impressions

Google search · last 28 days · up 47%

4.92K

Clicks

Google search · last 28 days · up 50%

52K

Pageviews

Last 7 days · 3.3K sessions

0

Paid spend

No ads · no influencers · no PR

18K+ indexed pages, written for women searching real golf course names with real dress code questions. The course directory is the SEO strategy. Every page exists because a woman is looking for that information at that moment. Utility scales when the product solves a search query that nobody else is answering.

What the experiment taught

Taste leads. Everything else compresses.

Taste is doing more work than ever.

AI hands you 100 options at the speed of one. The job is knowing which one ships and which 99 die. That is not a tool skill. That is a years-of-staring-at-pixels skill. The skill that compounds is editorial judgment. The one thing that cannot be delegated.

The brief is the design.

80% of the work is the spec. Once the spec was right, the output was 90% right on the first try. The traditional design process was mostly translation: intent into mocks, mocks into specs, specs into code. The brief now does all of that work directly.

Authority beats access.

Anyone can prompt the same model. The differentiator is what you tell it to do, and what you refuse to ship. The senior move is the no, said 99 times before the yes.

The artifact stack collapsed.

Wireframe, mock, prototype, spec, handoff. All compressed into one prompt-and-judge loop. Fidelity stopped being a phase. The assembly line existed because we didn't trust the next station. AI exposed which parts of that anxiety were load-bearing.

The meta-point

The artifact stack collapsed.
Taste and judgment did not.

Course Code is a product. The experiment behind it is a position on the design discipline. Most of the artifact stack we trained a generation of designers to build, the wireframes, mocks, prototypes, spec docs, handoff files, existed because we did not trust the next station to build the right thing. AI did not just make those artifacts faster. It made the stack honest.

What did not collapse is taste and judgment. The one hundred small calls about which screen earns its place, what the product refuses to be, which ninety-nine outputs die so the right one ships. Those calls cost me hundreds of hours on Course Code. AI could not make any of them for me.

That is the case for designers in 2026. We are not less important. We are more important. The discipline got harder because what is left of the work is the part only a trained eye can do. The senior move is deleting the anxious artifacts and protecting the irreducible ones.