Built with AI.
The operating model behind it.
A non-technical founder shipped a real, paying SaaS in 28 days by directing an AI coding agent, for about $27 in AI cost. This is the playbook: 12 chapters on the operating model that made it work, every claim backed by receipts.
Written for founders and operators putting AI to work, and for the investors deciding who can actually do it.
The chapters.
Every chapter is readable in full, no email required. Grouped into three sets that build on each other.
Foundations
The frame, the principles, and the field notes the whole thing rests on.
The Honest Frame
Why the believable version of an AI build beats the inflated one, and the one-paragraph thesis the whole playbook hangs on.
Read chapter →The Field Guide: 14 Principles
The fourteen named principles that run the whole operating model, each paired with the RadiusOS evidence behind it.
Read chapter →Field Notes: Thinking with Claude
Eight field notes on the chat and Projects layer: instructions over styles, clean your knowledge, one project per project.
Read chapter →The operating model
The repeatable loop: the brief, the build cycle, governance, and verification.
The Project Brief
CLAUDE.md is the moat. The anatomy of the single highest-leverage artifact in the repo, and how to write your own v0.
Read chapter →The Build Loop and Parallelism
Ship small and reversible, three to ten PRs a day. The loop that makes it work, and the real tradeoffs of worktrees.
Read chapter →Governance and Approval
As AI writes more, control lives in what you let through. Building a graduated approval ladder, cheapest gate first.
Read chapter →Build-Pass Is Not Ship-Pass
A green build is necessary, not sufficient. The prod crash that proved it, and how to verify by actually using the thing.
Read chapter →Sustaining the build
What keeps it honest and alive: cost, marketing, distribution, taste, and the record.
Cost Discipline
Haiku-first, cache every system prompt, batch the non-urgent, cap free-tier spend. The difference between $0.30 and $3.00 a user.
Read chapter →Keeping Marketing Honest
Marketing claims must be true today. The principles, the audit grep, and the build gates that fail the build to enforce them.
Read chapter →Distribution, the Real Moat
When anyone can build, being found is the edge. Capturing personas in a file and shipping the SEO flywheel from day one.
Read chapter →Taste: What the AI Cannot Do
The AI builds anything you describe. It cannot tell you which one is worth building, or when to stop. That judgment is the job.
Read chapter →How It Was Built, Day by Day
The chronological backbone: from create-next-app to live with paying users, the inflection points and the honest scars.
Read chapter →Chad Newell
A CEO, COO, and General Manager with 30 years building, scaling, monetizing, and exiting marketplace, SaaS, media, and creator-economy businesses. A three-time founder with two exits: he grew Media Bakery to $32M in annual revenue before its acquisition, and founded Snapwire, a creator marketplace he scaled past one million creators across 50-plus countries, raised institutional venture capital for, and sold to StudioNow.
In 2026 he built RadiusOS, an AI-native SaaS, as a solo founder using Claude Code, taking it from concept to paying customers in 28 days and engineering its AI infrastructure cost from a projected $4,300-plus per month down to under $100. This playbook is the operating model behind that build. He works with venture and private-equity portfolios as an operating partner, installing this model in their companies.
Chad Newell · chadrnewell@gmail.com · linkedin.com/in/chadrnewell · Santa Barbara, California