Many young teams, in their excitement to build stuff, end up spending time on things which are not directly progressing their product or service. I have been guilty of this myself.

Often this stems from attempts at first principles thinking (which is great in principle) or Not-Invented-Here syndrome (not so good) coupled with a lack of research on existing solutions.

There has to be a clear advantage - a demonstrable positive ROI - to inventing a better wheel or for coming up with a custom solution. This is especially true of custom micro frameworks for routine things.

In most cases, in the early days of a startup, speed of delivery far outweighs absolute quality of code.

With AI, this is all the more relevant as throwing away code is getting easier by the day. Nay, hour.

Iterate as fast as you can.