Development Tips

1. Save money with these tips
   • Use Free or Cost-Effective Tools: Start with free-tier services like Vercel, Firebase, or Supabase for hosting and databases to avoid upfront costs.
   • Caching AI results (when testing): Limit API calls to costly AI models by caching results for testing and debugging purposes.
   • Scope the Project: Reduce unnecessary scope early to avoid resource waste (e.g., limit scraping or feature generalization unless necessary).
   • Monitor API Usage: Be mindful of API request limits and use strategies like batch processing to reduce costs.
   • Modular Development: Keep functionalities modular so you can build only what’s needed without over-investing in features.
   • Know AI’s Limitations: Some tasks are better not using AI. If heuristics might work better for some use-cases, don’t use AI!
2. Save time with these tips:
   1. Product Market Research: Do research on product fit/if it already exists. You don’t want to make something just to realize it already exists!
   2. Tool research: Identify the core functionality required by your app. Make sure whatever full-stack products you choose support these core functionalities.
   • Optimize Workspaces: Stick to one workspace or framework to minimize overhead from switching between incompatible tools.
   • Plan Testing Pipelines: Design testing and evaluation pipelines early to avoid scrambling for validation setups later.
3. General development tips
   1. Version control is your best friend; once you have a version that you’re happy with, use Git to be able to revert back to it!
   2. Decomposing the problem.
   3. Iterate Quickly: Start with small, testable versions of the app and iterate based on user feedback and testing.
   4. Focus on modularity: Build features with modularity to scale them later without extensive rewrites.
   5. Documentation Is Key: Record every useful solution or prompt; it can save time during future debugging.
   6. Preprocess when possible: Figure out whether you can make some internal tools and cache data (instead of doing everything real-time).
   7. Testing and iteration.
   8. Choosing which LLM to use?
   9. Philosophical: how much to rely on LLM-generated code vs. not?