Agreed. Everyone seems to be making the same things with AI. I think typically the innovation is not in just the idea, but also in the timing. Harvey and Legora are doing well not because the idea is revolutionary, but because they were early to be kingmade. That doesn't mean they are undisruptable, if we see private market valuations crash and a huge architectural disruption that makes fine tuning extremely cheap and commoditized, then all of their edge is for nought, and the KKRs of the world might as well build custom stuff using an army of consultants hosted on their vanilla cloud ala good old transformation projects. So it's not just the idea, but also the timing that matters IMO, and timing determines execution strategy.
Part of me really wants to agree, part of me just thinks it's too late to go back to the old ways and some slop will always exist. Do you have any good ways to evaluate AI Slop? Made a fun project (could call it a campaign) with a similar PoV, I'm spiritually old, check it out - dontshipslop.com, and let me know if you have any ideas on how to build a slop detector that evaluates things from a deterministic spec to code program synthesis approach...
I think many are trying to attack this, at least a dime a dozen startups like imagine.bo, rocket.new etc which are basically promising to help you build custom stuff that sidesteps your hacking together various subscriptions. The typical model is they're wrappers on various APIs that get you nearly there and then have human support/humans in the loop offshore to get the job done efficiently and respond to urgent queries. It's a tough business, and I honestly think the moat will be in cracking a scalable support layer that isn't AI.
It's tough, I decided to take the plunge (throwaway account) to do it and it was basically serendipity. Knew someone who angels, he was trying to mentor a startup that wasn't doing well, connected me to the bootstrapped founder, and we hit it off and went from there. To be honest, if you've worked in Big Tech, it's non trivial to make the switch to AI native SaaS, particularly if it's a tarpit idea. To give you context, I took a leap of faith on trying to scale a Lovable equivalent in India, and while the thesis was sound, the tech was good and worked great, the non-technical hurdles are quite insurmountable without doing rage bait type growth hacking or simply throwing money at the problem. The alternative if you want to start something in software is to go 'deep and vertical', which in practice ends up looking like AI first software agencies/neofirms. Pricing is also a challenge, and most people are doing variations of the same thing. As someone who took this plunge to work for negligible salary and a bunch of equity and has a few regrets (although I'm trying to pivot the company), the most important thing I'd optimize for is finding a team that's truly talented and driven and doesn't have AI psychosis. The other thing that is probably going to be motivating for you as a technical person, is working on a HARD problem, where the challenge is not just distribution or CAC:LTV shenanigans. There are plenty of startups attacking greenfield spaces like robotics that are capital constrained and need experienced people to solve tough problems. I know of one myself that has a modest seed funding round and is doing well. If you're talented, I'd search for such diamond in the roughs, so first time founding teams with really strong visions/hard problem space and moderate capital where you'd add a lot of value and they'd be more open to giving up a bit more equity.