How clear can you be with your requirements?

When I’m trying to decide which tech stack to go with, what the architecture will be, what are the performance implications of one vs the other, the best way to get answers is to build a prototype.

Having to think about the design and explain in a very detailed way how to implement helps me to discover the unknowns unknowns.

These experiments used to take from a few days to a few weeks.

How long do you need to implement a todo app in a language or framework you don’t know?

It takes a while…

But with the AI stack I’m using now (a combination of Cline for VSCode, Cursor and ChatGPT o1), I’m able to answer those questions in days, not weeks.

It’s so satisfying to watch your prototype assemble in front of your eyes. There are caveats of course. It’s not perfect. It gets things wrong.

But what you learn is to be very explicit and precise with your language to instruct the AI and get good results.

That skill is useful also with humans.

But more than that, the process of explaining the problem, my desired outcome and iterating quickly on a solution has improved the speed of my learning.

Fancy prototypes are just a cool side-effect.

Yours,

Taj

P.S. I’ll make a video showing off my process, I got a few questions already to demo the setup. Let me know if that’s something you’d be into.