The fact that we have lambdas/serverless functions and people are still over-engineering k8s clusters for their "startup project" is genuinely hilarious. You can literally validate your idea with some janky Python code and like 20 bucks a month.

The problem is that people don't like hearing their ideas suck. I do this too, to be fair. So, yes, we spend endless hours architecting what we'd desperately hope will be the next Facebook because hearing "you are definitely not the next Facebook" sucks. But alas, that's what doing startups is: mostly building 1000 not-Facebooks.

The lesson here is that the faster you fail, the faster you can succeed.