it's complicated.
no, seriously. it's a hugely complex undertaking.
if you can utilize an existing one, you get a lot of things for free. input handling, graphics pipelines, whatever else. everything that you get with An Engine, you likely have to go back and do yourself.
similarly, you lose out on everyone else. there are probably thousands of people out there who know whatever mainline bespoke engines; even for the smaller ones, there's likely still some good resources and guides, and a community you can lean on.
when you're going at it alone, you're going at it alone. you are responsible for everything: handling assets, inputs, rendering, porting (if you care). sounds. physics. events. networking. all of it is completely and totally yours.
it's an immense task, and everything relies on one another. you have to make everything interconnect and interface with one another. and again, it's all on you. nobody else has any experience with your engine.
it's the difference between doing some major renovations on a prefab house versus building one entirely new.
as usual, the real answer is that you should skim advice and figure out what works best for you. there are some situations where using one solution will work, and some situations where diy might work. everyone's situations and capabilities are different. there's no "always do this" and "never do this".
but the complexity shouldn't be overstated. making a game already isn't easy.
(at the same time, you could consider this a good reason to limit scope, too, especially if you're new to developing games. you will absolutely learn things that make you wish you did everything differently. it's easier to do things differently when they're small things.)
