Building Full-Stack Web Apps with Claude Code: Stacks, Workflows, and Iteration

Building full-stack, database-backed websites with AI coding assistants like Claude Code is proving to be a viable and efficient approach for many developers. While the capabilities of AI models can evolve and sometimes diminish over time, the core workflow for leveraging these tools remains highly effective for rapid development and proving concepts.

A notable example involved constructing a sophisticated retro-planning application. This project, built with Elixir, Ash Framework, and PostgreSQL, showcased the AI's ability to assist in developing a feature-rich application. The site included real-time collaborative features, custom markdown parsing for card creation (e.g., :-) for "What Went Well" columns), dynamic card movement, action item tracking (! - [ ] with checkboxes), and assignment capabilities (@email) that integrated directly with the database to update dashboards and future retrospectives. This demonstrates the AI's utility beyond simple CRUD operations, extending to complex business logic and real-time interactions.

Effective AI-Driven Development Workflow

A crucial takeaway from successful implementations is the adoption of an iterative workflow. Developers find immense power in a structured plan → review → iterate loop:

1. Plan Mode: Start by dropping the project idea into the AI, allowing it to outline an an initial approach or architectural design.
2. Review and Confirm: Human developers then review the AI's output, confirming what's correct and identifying areas that need adjustment.
3. Refine and Execute: Based on the review, the AI can be prompted to refine its code or generate specific components, which are then executed.
4. Debug and Repeat: The process continues with debugging, further refinement, and iteration until the desired functionality is achieved.

This iterative process transforms the AI into a powerful assistant, accelerating development by automating much of the initial scaffolding and iterative coding, while keeping the human in control of the overall direction and quality.

Common Technology Stacks

The discussion highlighted several popular technology stacks suitable for this AI-assisted development:

• Elixir with Ash Framework and PostgreSQL: Used for robust, real-time applications, facilitating complex feature development.
• React with Supabase/Neon/Postgres: A modern full-stack JavaScript ecosystem, leveraging serverless databases and frontend frameworks, often hosted on platforms like Vercel.

These choices reflect a preference for modern, scalable technologies that can be efficiently integrated with AI-generated code. Platforms like Fly.io and Vercel are frequently used for deployment, facilitating rapid deployment of prototypes and MVPs.

While AI models' performance can fluctuate, the core methodology of using AI for outlining and iterative coding, combined with human oversight, remains a powerful strategy for building complex web applications efficiently.