Vibe Coding for AI PMs
Vibe coding is building software by directing AI — describing what you want, iterating on what it produces, and shipping something real without writing code line by line. For PMs, it's about collapsing the gap between "I have an idea" and "I can show you something that works."
---
Context
What PMs can realistically build: Interactive prototypes, internal tools, data analysis scripts, simple web pages, API integrations, proof-of-concept demos. What vibe coding is NOT good for: Production systems at scale, security-sensitive code, long-term maintained code, replacing engineering for complex features. The PM vibe coder's golden rule: You are the product manager of the AI. Your job is to write precise requirements, evaluate output quality, and iterate — not to understand every line of code.---
Step 1 — Define what you're building
Write a one-page brief before touching any tool: What I'm building (one sentence), Who will use it, Core functionality (max 5 things), Explicit exclusions, Data (input/output/sample data), Tech constraints, and Done looks like (observable end state).
Step 2 — Choose the right AI coding tool
By use case: Full web apps with UI → Lovable, Bolt.new, v0. Code + iteration → Cursor, Windsurf, GitHub Copilot. Scripts and automation → Claude, ChatGPT, Replit. Quick prototypes → Claude, Replit, StackBlitz. Database-backed apps → Supabase + Lovable.
Step 3 — Write the initial prompt
Structure: Role (what you're building), What (specific output type), Functionality (verb-first, testable behaviours), Data (sample data), Constraints (libraries, backend requirements), Done looks like (observable state).
Step 4 — Iterate effectively
Five rules:
Step 5 — Know when to stop and hand off
Hand-off triggers: real user data needed, authentication required, production deployment, long-term maintenance, growing complexity, security concerns. Hand to engineering: the working prototype, the brief, a walkthrough video, known gaps list, and the proper spec.
Step 6 — PM vibe coding quality checklist
Functionality (all features work, tested with unexpected inputs, error states don't crash). Communication (labelled as prototype, 2-minute demo narrative prepared). Handoff readiness (written brief, can articulate what real version needs differently, no timeline promises based on prototype speed).