Phase Playbook: Validate a Problem

Phase 1 — Validate a Problem Worth Building

🎯 Mission

In Vibe AI Engineering, your goal is not to randomly generate code with AI, but to execute the Vibe Software Development Life Cycle (Vibe SDLC).

Phase 1 is about Problem Discovery & Definition. You will find one narrow, painful, real-world problem for a specific human before touching code or selecting AI tools. For our canonical path, we use Career Copilot: a product that transforms job descriptions and resumes into actionable, truthful application plans. You can choose your own problem if it satisfies the same level of operational specificity.

📦 What You Will Deliver (Phase 1 Outcomes)

By the end of this phase, you will have created four concrete artifacts in your repository under docs/:

1. docs/problem-brief.md: A 1-page document detailing the user, pain point, evidence, and success metrics.
2. docs/target-persona.md: A deep profile of your target customer (B2B or B2C) including daily tools, frustrations, and purchasing triggers.
3. docs/competitive-matrix.md: An audit of 5 existing solutions/alternatives and your product's Niche Delta (Unfair Advantage).
4. docs/prd.md: A Cursor-ready Product Requirement Document (PRD) specifying tech stack decisions, boundaries, and acceptance criteria.

🛠️ The Vibe SDLC Phase 1 Step-by-Step Workflow

Step 1: Identify Pain  ──>  Step 2: Business Model  ──>  Step 3: Market Audit  ──>  Step 4: Tech Choice  ──>  Step 5: Cursor PRD
(Hair-on-fire problem)     (B2B / B2C / SaaS model)      (Exa / Perplexity scan)     (Frontend/Backend/Cloud)   (AI Contract)

1. Frame the Problem Statement: Fill in the formula: “A [specific target user] struggles to [costly manual task] when [context/trigger], wasting [X hours/dollars].”
2. Determine the Business Model: Decide whether your solution is B2B (Business-to-Business), B2C (Business-to-Consumer), or Micro-SaaS, and how value is captured.
3. Gather Hard Evidence: Collect at least 3 real evidence points (forum complaints, workflow audits, support tickets). Do not ask ChatGPT if a problem is real—AI will always say yes.
4. Find Your Niche Delta: Identify 3–5 existing competitors. If competitors exist, do not panic—it proves market demand! Determine why users will switch to your tool (e.g., 10x faster workflow, specialized AI agent, 80% lower cost).
5. Select Your Stack & Draft the PRD: Select your Frontend, Backend, Database, and Cloud stack using the decision matrix. Format your PRD using the standardized template so Cursor / Windsurf can build the app without hallucinating.

🛠️ Phase 1 Tool Map: Which AI to Open for Each Task

Students often open ChatGPT for everything. Here is the right tool for each Phase 1 task:

📋 Phase 1 Checklist: Definition of Done

• Specific User: The problem names a distinct persona (e.g., "Freelance Graphic Designers"), not "everyone."
• Hair-on-Fire Pain: The task is painful, repeated, and currently solved through expensive or slow manual workarounds.
• Business Model Defined: Clearly categorized as B2B, B2C, or Micro-SaaS with a pricing hypothesis (in ₹ INR or $ USD).
• Competitor Positioning: You can state in 1 sentence why your product wins over existing alternatives.
• Tech Stack Locked: Frontend, Backend, Database, and Hosting are selected with documented trade-offs.
• Cursor-Ready PRD Written: docs/prd.md is committed to Git with clear Pydantic schemas and system boundaries.

❓ Quick Self-Check

Question: "I want to build an AI app that generates motivational quotes for college students. Is this a good Phase 1 problem?"

Answer: No. This is a feature looking for a problem. College students do not have a "hair-on-fire" operational need for quote generators, nor is there a clear willingness to pay. A valid problem would be: "College seniors spend 15+ hours manually reformatting resumes and writing custom cover letters for every application on LinkedIn."