Conclusion

What we built, what we learned, and what comes next

The finished application

• A walk through the complete app: data loading, browsing, searching, analysis, visualization, and tests
• What 100 prompts produced

What LLMs are good at

• Scaffolding, boilerplate, and explaining unfamiliar APIs
• Translating error messages into plain language
• Concrete examples from the tutorial

What LLMs are not good at

• Hallucinating library APIs and function signatures
• Introducing subtle security flaws (missing CSRF tokens, SQL injection)
• Understanding your specific requirements without very precise prompts
• Concrete examples from the tutorial

Concerns about LLM tools

• Environmental cost: energy and water consumption of training and inference
• Deskilling: what happens when you can generate code you cannot read
• Labor displacement: honest acknowledgment without pretending there are easy answers

Where to go from here

• Deployment: running the app on a server others can reach
• Database migration: changing the schema without losing data
• Making the app collaborative: multiple users, roles, and audit logs
• Reading list