How Effective is AI at Building Full-Stack Applications?

AI shines in greenfield projects, especially when you have a well-structured codebase. Tools like Cursor are pretty good since they index files, making it easier to guide the AI. Just remember, you still need to review all of its outputs.

Sure, a tiny full-stack app like a to-do list? AI can manage that pretty well. However, if you're aiming for something truly useful, you'll need to use techniques like subagents and testing pipelines to get it right. The technology is improving, so these processes should become smoother over time.

For smaller tasks, AI is great! You can describe what you want, and it will create a function based on your description. But building an entire project is where it struggles a lot—it can be done, but it requires more effort and know-how to manage.

Some claim that at Anthropic, team members haven't had to write code manually for quite a while! However, I'd be cautious about relying on AI entirely since it often provides support links that are incorrect.

AI can handle generating code to some extent, but its effectiveness really depends on the complexity of the app. If you're looking to make something straightforward, it can work well, but for anything more advanced, you'll need to guide it and clean up after its output.