Field guide series · rev 1

The AI Builder Stack

Fourteen layers for a senior PM going from "what is an API" to shipping web, SaaS, mobile and game products with AI doing most of the typing.

Read it top to bottom, or jump to a layer. Layers 00–09 are the core path, built and sequenced for web and SaaS first, since that's where the tooling is most mature and the payoff comes fastest. Mobile and games sit downstream once that foundation holds. Every guide ends with a linked Sources page listing what was actually researched to write it.

ForSenior PM, first-time builder
SequenceWeb / SaaS → Mobile → Games
Status14 of 14 guides written
CompiledSep 2026
Phase 1 — web & SaaS, start to finish
00

The vocabulary every AI tool quietly assumes you already have: client vs. server, frontend vs. backend, what a database is, what an API call does, and where a "hosted" app actually lives.

Why it matters: without this, every tool's output is a black box — you can't judge it, only hope it's right.

Covers client/server · frontend/backend · database · API · hosting

01

Turning the PRD and user-story skills you already have into specs, prompts, and acceptance criteria an AI coding tool can actually execute against.

Why it matters: this is the one skill in the whole stack that's already yours — it just needs a new target.

Covers PRD-to-prompt · acceptance criteria · scoping a build session · context windows

02

Prompt-to-app tools that go from idea to a working, hosted web app with no code written by hand — the fastest way to get something real in front of people.

Why it matters: your first shipped thing should prove the idea, not prove you can code.

Tools Lovable · bolt.new · v0 · Replit Agent

03

What changes when you graduate from a no-code builder to a tool that reads, edits, and runs a real codebase on your behalf — and when that trade-up is worth it.

Why it matters: you've already opened Claude Code once — this is where "beginner" turns into "dangerous enough to be useful."

Tools Claude Code · Cursor · Windsurf · GitHub Copilot

04

Why nobody good starts from a blank file: pre-wired starter kits, component libraries, and template marketplaces, and how to judge which one fits a given project.

Why it matters: a good template quietly encodes hundreds of decisions you don't yet know how to make.

Covers Next.js starters · Supabase · shadcn/ui · template marketplaces

05

Just enough architecture literacy — frontend, backend, database, auth, payments — to direct an AI tool's output and catch it when something's wrong.

Why it matters: you don't need to build this yourself, but you do need to know what "backend" means when the AI says it broke.

Covers auth · database schema basics · Stripe/payments · third-party APIs

06

Version control, the difference between staging and production, and what actually happens when you click "deploy."

Why it matters: this is the difference between a demo on your laptop and something a real user can open.

Covers Git/GitHub basics · environments · Vercel/Netlify

07

Wiring your app to the tools around it — webhooks, third-party APIs, and workflow platforms — instead of hand-coding every connection.

Why it matters: builds directly on ground you've already covered.

Builds on The n8n Builder & Seller Stack

08

Using LLM APIs, retrieval, and agents as building blocks inside what you ship — the shift from "AI builds my app" to "my app uses AI."

Why it matters: builds directly on ground you've already covered.

Builds on The AI Factory Stack

09

How to verify code you can't fully read yourself: a triage process for AI-introduced bugs, and the questions to ask the model when something breaks.

Why it matters: "it looks right" isn't a QA process — this is the one you can actually run as a non-coder.

Covers reading error messages · regression checks · when to ask vs. when to revert

Phase 2 — once web/SaaS is solid
10

Shipping to mobile

Later · mobile

What changes once "web app" has to become an app-store build: cross-platform frameworks, AI tooling that targets mobile specifically, and no-code mobile builders.

Why it matters: most of layers 00–09 transfers — this layer is the delta, not a restart.

Tools React Native/Expo · Flutter · no-code mobile builders

Phase 3 — furthest from the PM toolkit
11

Shipping to games

Later · games

Game engines and the AI-assisted tooling starting to wrap them — where the "prompt it into existence" approach holds up, and where it still doesn't.

Why it matters: the stretch goal — worth naming now, worth sequencing last.

Tools Unity · Godot · Phaser · AI game-dev copilots

12

A cliff-notes pass through the systems-thinking canon — Meadows, Weinberg, Senge, Brooks — and what it re-becomes once an AI agent is the one making architectural decisions fast, without telling you.

Why it matters: gives the reading list something to land on, instead of past it.

Covers stocks & flows · coupling · essential vs. accidental complexity · a reading order

13

The senior-PM layer on top of all of it: security and privacy basics, keeping API costs sane, and knowing exactly when to stop and bring in an engineer.

Why it matters: the judgment call every other layer eventually needs from you.

Covers security basics · cost/usage limits · scaling · when to hire