Divine Intervention: How One Founder Cracked a Two-Company Monopoly with AI

For as long as anyone can remember, an online pharmacy could only get accredited by one of two companies — and the backlog showed it. Aaron VanStone is building the third, ScriptSafe, and he coded the entire platform himself, with no software background, using Claude Code.

"The classic thing they teach you in every coding class is that the more preparation you do beforehand, the better the end output's gonna be. It's no different with these models."

About This Episode

Aaron VanStone has spent more than 20 years in payments. In 2008 he founded Processing Brokerage, a pay-for-performance firm that audits merchants' credit-card processing fees across the U.S., Canada, and the U.K. A project with a compounding-pharmacy trade group pulled him into the pharmacy world — where he ran headfirst into a bottleneck: to accept online payments or run Google ads, a pharmacy needs third-party accreditation, and "since the beginning of time" there have been only two accreditors. The backlog was enormous.

His answer is ScriptSafe, an AI-powered accreditation platform that reviews a pharmacy's licenses and then monitors its website every night to confirm it isn't selling anything illegal. What makes the story remarkable is that Aaron built the technology himself — starting at the end of January with zero coding experience, using Claude Code (with Codex as a second set of eyes). As he puts it, the timing felt like "divine intervention": the tools arrived exactly when he needed them, and what he built "wasn't possible a year ago."

This is a tactical conversation. Aaron and Travis get into the workflow that actually works: writing a detailed planning document before building, running two models against each other so they have to agree, building a brand system for consistent design, and the cheap starter stack — Vercel, GitHub, and Supabase — that lets anyone ship. They also dig into the lesson that cost Aaron a month: there was no observability layer, so the AI was guessing at bugs. Once he added one, nightly pharmacy scans dropped from four hours to under 90 seconds.

It's also a candid take on the risks — vendor lock-in, the coming wave of "AI-slop software," and why vibe-coding your own CRM might be shortsighted if you ever need to hire or train for it. And underneath it all runs an encouraging message: there has never been a better moment to learn this, especially if you've just been laid off and finally have the runway to dig in.

About Aaron VanStone

Aaron VanStone leads three companies at the intersection of payments, fintech, and healthcare compliance, drawing on more than 20 years in the payments ecosystem. He founded Processing Brokerage in 2008, a pay-for-performance consulting firm that audits credit card processing fees for merchants across the U.S., Canada, and the U.K. Most recently, he launched ScriptSafe, an AI-powered pharmacy accreditation platform designed to eliminate regulatory bottlenecks for online pharmacies. He's built ScriptSafe's technology stack using AI coding tools, a workflow he discusses in depth on topics ranging from Claude Code to brand systems in web development.

Key Takeaways

• Prep beats prompts — Aaron spends 90 minutes to two hours building a working planning document (with the model interviewing him) before writing any code. The richer the plan, the better the build.
• Run a second model as a reviewer — He has Codex continuously review Claude/Fable's output; when the two models agree, the code is far more trustworthy.
• Build a brand system first — Define your colors, logo, fonts, and overall look up front, then tell the builder to follow those rules so everything stays consistent.
• You can't fix what you can't see — Adding an observability layer turned guess-and-check debugging into real analysis and took nightly pharmacy scans from 4 hours to under 90 seconds.
• The starter stack is cheap and learnable — Vercel (hosting) + GitHub (code) + Supabase (database), roughly under $30/month, is enough to "do anything."

Chapter Markers

• 0:00 — Cold open
• 0:29 — Meet Aaron & ScriptSafe
• 1:56 — From payments consulting to a new problem
• 3:42 — Divine intervention: finding AI at the right moment
• 6:07 — Fable 5 vs. Opus — you feel the upgrade
• 11:29 — The "building a house" analogy for model progress
• 12:59 — Prep beats prompts: plan before you build
• 15:18 — Running Claude and Codex against each other
• 18:35 — Vendor lock-in and the coming software shakeout
• 20:56 — The "vibe-code your CRM" debate
• 26:47 — The observability fix: 4 hours to 90 seconds
• 29:23 — Starting from zero coding experience
• 31:23 — "If you don't know where to start, ask it" + the hotel scanner
• 34:03 — Why now is the time to learn — even if you've been laid off
• 39:16 — The starter stack: Vercel, GitHub, Supabase
• 40:52 — Design: Manus, Claude Design & brand systems
• 42:04 — Two big tips to finish

Resources & Links Mentioned

Tools & platforms:

• Claude Code — https://claude.ai 
• Claude Design — https://claude.ai 
• Claude (Max plan) — https://claude.ai 
• Codex (OpenAI Codex CLI) — https://openai.com/codex/ 
• Vercel — https://vercel.com 
• GitHub — https://github.com 
• Supabase — https://supabase.com 
• Manus — https://www.manus.space 
• Opus Clip — (mentioned by Travis) 
• ChatGPT — (referenced)

Books / podcasts referenced: None.

Connect with Aaron VanStone

• Website: https://scriptsafe.com
• Social / newsletter: LinkedIn — https://www.linkedin.com/in/aaronvanstone-b86622b8 · ScriptSafe on LinkedIn — https://www.linkedin.com/company/scriptsafe/about/

Connect with Travis

• Rainier RevOps: https://rainierdigital.com
• LinkedIn: https://www.linkedin.com/in/travisscott24
• Instagram: https://www.instagram.com/travis.revops
• YouTube: https://www.youtube.com/@RainierRevOps

If This Episode Resonated

If your company is buying AI tools but hasn't done the work underneath - clean CRM data, documented processes, a single source of truth - that gap is the whole reason the tools feel disappointing. The AI Readiness Scorecard diagnoses it in about 10 minutes.

Get Your AI Readiness Scorecard

Recorded in Descript Rooms. Edited in Descript. Hosted on Transistor.

AI-Ready or Not is a production of Rainier RevOps.