When tackling complex scenarios that involve adding a lot of new files to a project, I start by writing pseudocode to sketch out how I imagine the top-level API would look. Then I discuss with LLM potential design structure options to refine the approach. Once I'm happy with the proposed architecture, I have a clear action plan for which files to create and what logic belongs in each-like Models, Actions, Services, Scopes, and so on.
This is just beautiful. So many nuggets and light bulb moments for me. From your experience, ChatGPT and ClauddeAI which performs consistently better in terms of proper syntax and less refactoring of generated codes?
At this moment Claude Sonnet 3.5 (free) is significantly better than ChatGPT 4o (free). But I've just watched a video with o1-Pro version and it blows away all other models in terms of how big context it may accept and how complex project it can generate. So, the battlefield is constantly changing and evolving, I will keep watching this space, doing my own research/experiments and will shoot more videos.
When tackling complex scenarios that involve adding a lot of new files to a project, I start by writing pseudocode to sketch out how I imagine the top-level API would look. Then I discuss with LLM potential design structure options to refine the approach. Once I'm happy with the proposed architecture, I have a clear action plan for which files to create and what logic belongs in each-like Models, Actions, Services, Scopes, and so on.
This is just beautiful. So many nuggets and light bulb moments for me.
From your experience, ChatGPT and ClauddeAI which performs consistently better in terms of proper syntax and less refactoring of generated codes?
At this moment Claude Sonnet 3.5 (free) is significantly better than ChatGPT 4o (free). But I've just watched a video with o1-Pro version and it blows away all other models in terms of how big context it may accept and how complex project it can generate. So, the battlefield is constantly changing and evolving, I will keep watching this space, doing my own research/experiments and will shoot more videos.
@LaravelDaily thank you for this info. I will happily be along for the ride and learn with you. 😊
I feel like using chatGPT is cheating, but on the other side i feel like its a tool i can use to speed up my development.
What is your take on this?
I think it's a tool. Cheating is if you don't understand what ChatGPT does for you and you're unable to maintain that code later.