Building ARR Club: A Non-Developer's Journey to Shipping a Complete SaaS Product with AI

Five days ago, ARR Club was two disconnected products — a basic database at data.arr.club and a Framer news feed at arr.club. Today, it's a unified SaaS platform tracking $115B+ in annual recurring revenue across 716 companies, with Google OAuth, Stripe payments, AI-powered content generation, and a newsletter system.

I built it without writing a single line of code myself. I'm not a developer. I'm a founder and investor who talks to founders and investors every day, and I needed a better way to track the revenue data that matters.

This is the story of how my AI employee Junior turned that need into a real product.

The Starting Point: A Spreadsheet That Grew Legs

ARR Club started the way most data products do — I was manually tracking ARR figures for AI and SaaS companies. LinkedIn posts, earnings calls, press releases, podcast interviews — every time a founder mentioned a revenue number, I logged it.

The spreadsheet became a Supabase database. The database became data.arr.club, a paid product where investors and operators could look up verified revenue data. Separately, I ran arr.club on Framer as a news feed for revenue signals — funding rounds, milestones, growth stories.

Both products worked, but they were disconnected. Users couldn't search for a company and see both its current ARR and the signal timeline that tracked how it got there. The database had no SEO. The news feed had no paywall. Neither had proper authentication or billing infrastructure.

I knew what the product should be. I just couldn't build it.

Day 1: The Merge Begins

When I started working with AI coding tools, the first task was auditing the existing codebase (Next.js 16 + Supabase + Stripe — scaffolded by Lovable, a no-code tool) and planning the migration.

The migration alone would have taken a traditional development team weeks:

• 643 companies with their profiles, logos, industries, and founding data

• 1,080 ARR milestones with verified dollar amounts

• 208 editorial memos converted to structured signals

• 229 user accounts with subscription status preserved

Stripe customer IDs mapped and payment history maintained

We completed the full data migration in one session. Every company had its logo (715 out of 715 backfilled via Brandfetch CDN), every milestone was verified against the original database, and 98 duplicate signals were cleaned up.

Day 2-3: Building the Product Layer

This is where AI development felt genuinely different from anything I'd experienced. Features that I'd assumed needed weeks of planning and development happened in hours:

Authentication: Google OAuth + Magic Link login, replacing the old email/password system. Seamless migration so existing users could log in without re-registering.

The Paywall: A 20-view quota system for free users, with Stripe checkout for Pro subscriptions ($20/month or $150/year). Opening checkout in a new tab. Handling webhooks. Managing billing portals. Every edge case handled.

The Homepage: A gradient hero section, live statistics, signal feed with type filters, database preview, leaderboards, FAQ section, and an instant search with company typeahead. All responsive.

Company Pages: Three-column layout with milestone timeline, interactive ARR growth charts (smart enough to switch between bar and line charts based on data density), and a sidebar with related companies.

SEO Infrastructure: JSON-LD structured data, dynamic OG images for every company and signal page, XML sitemap, robots.txt with AI bot rules, and 301 redirects from the old domain.

Day 4: The Features That Make It a Real Product

ARR Compare: Users can select multiple companies and overlay their growth curves on a single chart. It generates a shareable URL with a dynamic OG image — designed to go viral on Twitter and LinkedIn.

AI Signal Generator: I paste a URL — an article, a blog post, a YouTube interview — and AI generates a structured article with extracted data, company context, and analysis. I review, edit, and publish in minutes instead of hours.

Rich Text Editor: Full-featured content editing with image uploads to Supabase Storage, YouTube embeds, and podcast iframes.

Newsletter System: Compose and auto-generate weekly newsletters, send via Resend API, with proper unsubscribe handling and subscriber management. The free tier gives us 3,000 emails per month — more than enough for our 229 subscribers.

Research Reports: A Pro-only content type for deep industry analysis. Free users see a preview; paid users get the full report.

Day 5: Polish, Security, and Scale

The final day was about the details that separate a prototype from a product:

• Supabase bandwidth optimization (from 60GB/month to under 2GB by implementing slim SELECT queries and aggressive caching)

• Cloudflare CDN integration with cache rules that protect authentication flows

• Company follow notifications with user-configurable digest frequency

• Domain migration with proper DNS and redirect handling

Mobile responsive design across every page

We also hit real bugs. A Server Component with an onClick handler caused a cryptic crash that took four commits to trace to a React Server Component serialization issue. Cloudflare's www redirect broke Google OAuth because PKCE cookies were stored on the wrong domain — we eventually solved it by switching to implicit flow entirely.

These are the kinds of problems that make you a better builder, AI-assisted or not.

What I Learned

1. AI doesn't replace product thinking. The AI could build anything I described, but it couldn't tell me what to build. Every feature decision came from understanding my users. AI is the best contractor you've ever hired. You still need to be the architect.

2. Speed changes the game. When building takes hours instead of months, you can afford to try things. We built the ARR Compare feature because it sounded fun — and it turned out to be our most shareable content format. You'd never greenlight that in a traditional sprint.

3. Bugs are inevitable, but they're solvable. We had crashes, broken OAuth flows, bandwidth spikes, and rendering issues. Every single one was diagnosed and fixed in the same session it was discovered. The feedback loop is so tight that bugs feel like speed bumps, not roadblocks.

4. Data quality > feature quantity. The most valuable thing about ARR Club isn't the UI or the features — it's that someone (me) manually verifies every revenue figure. AI helps me build the product, but the data integrity comes from human judgment. That's the moat.

5. Start with your unfair advantage. I had 643 companies with verified ARR data. I had relationships with founders who shared numbers with me. I had an audience that cared. The technology was the last piece, not the first.

What's Next

ARR Club now tracks 716 companies representing over $115 billion in annual recurring revenue. We're adding new companies weekly, publishing research reports, and building tools that help founders benchmark their growth and help investors discover breakout companies.

The vision is simple: become the trusted source for private company revenue intelligence. The kind of data that's too important to be wrong and too valuable to be free.

If you're a non-technical founder wondering whether AI can really help you build a real product — yes, it can. But bring your domain expertise, your user empathy, and your willingness to iterate. AI gives you leverage. You still need to know where to point it.

ARR Club tracks verified revenue data for 1000+ AI & SaaS companies. Browse the database at www.arr.club

Clark Graham
Co-founder