The thing about modern AI development - both developing things with AI and developing AI things - is that you often need to know the right magic incantations. Knowing or not knowing them can mean tens or hundreds of hours wasted.
Recently I spoke about the iterative loop for AI-enabled development, debugging and sorts: [ Ссылка ]
This is meant to be a companion, where we attempt to go from raw materials to UI as quickly as we can, with as little direct coding involved. We'll look at some of the common pitfalls, some useful things it's taken me a long time to learn, and you'll also see me make some mistakes.
Inspired by @AndrejKarpathy 's build from scratch series, which helped me so much when learning things: [ Ссылка ]
I'm using a few transcripts from @LatentSpaceTV for demonstration - definitely check it out! [ Ссылка ]
Also useful things:
1. shadcn/ui: [ Ссылка ]
2. nextJS: [ Ссылка ]
If you'd like to jump around, here are the key sections:
00:00 The problem to solve
01:58 Getting transcripts
02:29 The iterative loop for AI development
03:52 Chat
05:52 Play
08:02 Structure is everything
12:56 Loop - let's code!
18:34 Debugging
20:18 UI/UX
25:25 First version!
31:06 Much better - conclusion & ramble
(Small error - at 20 minutes I say 'not going to work' I mean 'now' :))
Also thanks to Hebe Hilhorst ([ Ссылка ]) who built a far better version of transcript processing than I could!
![](https://i.ytimg.com/vi/8w0hUcQSDy8/maxresdefault.jpg)