Given the dense amount of wisdom packed into this comment I checked out your profile. Both your current projects look extremely impressive and polished. And viable! Keep up the good work! It’s inspiring to see that in light of the obvious difficult lessons you’ve had to learn along the way.
Having learned many of the same lessons as you I can 100% backup everything you said in your “sum it up” paragraph!
The only caveat I would add is to the “make an MVP with AI”. I think MVPs generated directly out of ChatGPT/Claude are so easy now (or at least it can appear so on the face of it) that many people are just barely going beyond that - but to any experienced eye, that approach is quite transparent and can look very low-value (even if the idea is actually a good one).
Now if that person is a skilled salesperson then that might work.
But, for most people, I think it’s still very important to demonstrate good instincts, taste and strategic/commercial understanding when building such an MVP. And that means editing and shaping the output just enough to meet your vision for the product. So to agree with you - definitely, 100%, use AI as much as possible - but don’t assume that you can put zero work in on top and have that MVP be effective. Because the 10 year old down the street has the exact same tool as you - so if you are just relying verbatim on that tool’s output- it’s going to be hard to stand out.
I’d still definitely agree to spend as little time as physically possible on the MVP - with the above caveat.
Having said all that… a lot of historical wisdom on the topic of MVPs has been turned upside down since gen AI became mainstream, so on the flip side you could argue: create 1000 MVPs in an hour, publish them all, see what generates interest…*
Hmm.. I think I just argued against my own point.
* (I’m not really seriously suggesting anyone do this, but I’m also not entirely discounting this as an approach either…)
> Both your current projects look extremely impressive and polished. And viable!
Appreciate it and thanks for the positive feedback! But those are two of the multitudes of side projects I have collected that I haven't figured out how to monetize. My "day job" is at a VC-backed startup that is going through a protracted wind-down because it also failed to find a viable GTM. So yeah, I've learned some hard lessons in multiple facets of my career!
> Now if that person is a skilled salesperson then that might work
My rule now is that if I'm building something for fun, I just open source it. If I want to make money, I'm going to first figure out who's paying and how do I get them to pay. AI MVPs are easy now to let you flesh out an idea one level up from a slide deck (in fact, maybe this is its own startup idea? Use AI to build an MVP from a deck??).
I had a non-technical friend recently spin up a full blown startup with customers using nothing but Claude + Replit (not plugging, but just sharing to show that it's real: https://bullship.co). He came up with the idea after talking to a friend and finding that indeed, the market had only two major competitors who both charged too much for many smaller customers.
The code is throwaway in my book, but it's enough to validate the idea by actually getting people to pay for something they can use. It won't scale, but that's fine; by the point that he needs it to scale, he'll be able to hire people with more skill to fix or rebuild it.
https://turas.app video is confusing, what I got is I have to do the planning myself using your app. Why would I do that instead I can get Chatgpt do the planning for me with some keywords on what I like/want.
You also have 20 different icons, that is so confusing for the user, get rid of most of them and just keep maybe 3 or 4.
Having learned many of the same lessons as you I can 100% backup everything you said in your “sum it up” paragraph!
The only caveat I would add is to the “make an MVP with AI”. I think MVPs generated directly out of ChatGPT/Claude are so easy now (or at least it can appear so on the face of it) that many people are just barely going beyond that - but to any experienced eye, that approach is quite transparent and can look very low-value (even if the idea is actually a good one).
Now if that person is a skilled salesperson then that might work.
But, for most people, I think it’s still very important to demonstrate good instincts, taste and strategic/commercial understanding when building such an MVP. And that means editing and shaping the output just enough to meet your vision for the product. So to agree with you - definitely, 100%, use AI as much as possible - but don’t assume that you can put zero work in on top and have that MVP be effective. Because the 10 year old down the street has the exact same tool as you - so if you are just relying verbatim on that tool’s output- it’s going to be hard to stand out.
I’d still definitely agree to spend as little time as physically possible on the MVP - with the above caveat.
Having said all that… a lot of historical wisdom on the topic of MVPs has been turned upside down since gen AI became mainstream, so on the flip side you could argue: create 1000 MVPs in an hour, publish them all, see what generates interest…*
Hmm.. I think I just argued against my own point.
* (I’m not really seriously suggesting anyone do this, but I’m also not entirely discounting this as an approach either…)