> It's just not as good

Again, the evidence (as limited as it is) suggests otherwise. You are more likely to succeed if you're going with dynamic language and not doing "proper engineering". This has been widely accepted before type-checker era, and I see no reason why it would be different now. Utilize type checker when it's free, but don't waste time on type puzzles.

"Proper engineering" doesn't get you to product-market fit faster. All it does is tickle your ego.