Technical founders - is "bulling" your way through learning right for a startup? [I will not promote]

Sup,

This is a question for technical founders.

--a little backstory--

I am starting a company in AI field that creates something nobody has ever done before. 7 months in.

---

How most software companies are created - you have an improvement idea, then you have a thousand or so problems to solve to make that improvement happen, and for each one that you don't know, you go to Stackoverflow or ChatGPT to look for solutions for that problem.

Which involves next-to-no upfront preparation because for vast majority of traditional software you can solve it on-the-go - "traditional" software is very easy compared to, say, mechanical, pharma or AI engineering.

However, for more advanced disciplines - can you just "Google" it on-the-go?

I'm a solo founder, and 8 months in, creating a foundational model, BECAUSE I did not know things upfront, I've wasted at least 3 months doing something which was mostly technically unviable in the first place. Out of 14000 lines of code that I've done (including tests), I had to scrap 10000 recently. Imagine the scale of it.

Obviously I didn't even know how ML works when I've started.

Major fuck-up.

How do you operate in industries which you've done before? How do you determine that it's time to start creating you big technological leaps instead of continuing to learn?

Cheers.

Edit: No need to push me on business topics. I know how to create value very well. It's only a tech question, and I'm only asking because - well - to deliver my value, I need to do a lot of novel tech.