In my last startup, I built a parental filter which worked over VPN, could decrypt internet traffic (itself not a trivial feat), and then used small LLM's to remove portions of web pages that violated filters, without blocking the page itself. Highly technical, and I was always pushing for lower latency or better fault tolerance.

Unfortunately nobody cared, including the original customers I was talking to, and so I spent my days stapling flyers to polls and talking to parents trying to learn what their actual problems were. Every small startup I've been a part of looked like this. Companies live and die over whether they can find market fit, not whether the product could be built.