AI as a Coding Sidekick: Lessons from Building an App Alone

A solo developer attempted to build a complete marketing tool using Google AI Studio and Gemini 3.0 Pro, hoping the AI could act as a full team member without any manual coding. The plan involved strict rules: no math, no hidden state changes, and mandatory JSON schemas for every output. The developer also wanted the AI to adopt a strategy pattern so it could select appropriate models for each campaign type while keeping deterministic TypeScript logic separate from the AI’s probabilistic suggestions.

The Open‑Mic Approach

• Early Stage: The assistant began sprinting ahead, altering files unexpectedly and frequently breaking existing functionality.
• Behavior: More like a junior coder experimenting than an experienced engineer with architectural discipline.
• Checkpoints: The developer inserted checkpoints to force the AI to reason before writing code and await approval.
• Apologies: The AI would often say, “I’m sorry,” which felt more courteous than corrective.

Code Growth & Architectural Chaos

• Unstructured Logic: The AI added logic wherever it found a quick path, ignoring SOLID and DRY principles.
• Refactoring Nightmare: Every tweak introduced regressions that the AI could not test itself, requiring manual checks.
• Testing: The developer asked the assistant to write a Cypress‑style test suite. The AI’s “tests” were guidelines, not executable code, leaving the developer responsible for test integrity.

Final Takeaways

• Vibe Coding: Using AI as a partner is not a shortcut to production software.
• Requirements:
• Rigorous architectural boundaries
• Continuous testing
• Clear AI roles (implementation, analysis, consulting)
• Outcome: When guided properly, the assistant can accelerate feature delivery; without guidance, a solo developer would take much longer.
• Conclusion: AI is powerful but still needs a manager, not just better prompts.