Building a startup isn't a normal work project. Most work projects are not expected to fail, but for a startup, nearly all the time you'll fail.
When I worked on my previous job, as an ML engineer, my main work was in porting pre-existing ML models to work for different datasets & runtime environments. I may have not been able to predict when my projects would be finished, but could easily predict that once they were done they'd work.
That's not the case with a startup. The most intuitive way to start a startup, is to just work on something you think might turn in a profitable business. That might work, but it also might not, and more often than not it won't. ~Most startups fail. ~ There's a much better way to do make build something. Something that increases your chances of success (although of course there will not be a surefire hit), but before we get to what it is we need to go on a bit of a rhetorical detour around a couple observations about building startups that will logically lead to us to our ideal strategy.
Getting in front of customers is hard. They're busy people and they've seen the likes of you before. Some random stranger purporting they've built something that could be a game changer only to be pointing to something that doesn't even work. You'll have a crappy product at the start, and with a crappy product not many people will care, or even give you the time of day for a quick call.
Easiest way to make something that doesn't work is to build it in the dark. You need easy access to feedback so you can iterate and build something people will actually want.
The faster you can get access the less time is spent working on something that won't lead anywhere, so it crucial you get feedback as quickly as possible. In the early days, you probably won't have easy access to a bunch of users, so frankly the only way to do that, is to know what your product needs to do intuitively. You need to be able to dogfood your product and figure out if the product actually works for you.
Build as fast as possible, build to build for yourself, iterate until it's good for you, launch get feedback. Feedback is the most important thing you could ever have in the early days. Without it you just don't have any direction.
In fact, feedback should be the goal. Prior to having a way to consistently get feedback and work towards something that really hits the needs of your customers. You're aimless. Stop aiming for anything else, and just focus on getting a stream of feedback set up.