I Overestimated The Power of AI | Project Destroy Tedium 17

So after letting AI take the reins for a little bit, mainly for the past three days, I have come to realize that the method of working with it and iterating is overall better.

I think I tested out my Destroy TDM program that it helped me create overnight two nights ago and there are already some problems with it.

There are already some things that despite me saying you know test it out and make sure that everything functions properly, It's mainly like logical elements where I'm trying to figure out exactly how to use the program in general because you know, we built it over time.

I built it over time with Cloud Code so everything functions but if you're starting from scratch and you want to build a flow, how do you do that? How do you go about doing that exactly? And I don't think that Cloud Code really thought it through because it feels very confusing when I'm trying to make a new flow.

So I basically have to I guess think of it myself and then figure it out with maybe cloud code.

Maybe I can tell it that it feels unintuitive and you know can you try to structure it in a better way blah blah blah.

Either way one prompt was not enough basically and it probably never will be enough to just do one prompt.

There will have to be a lot of refining but I definitely overestimated the power of AI in this occasion where it's just one and done type of deal.

Like it's not capable of quite doing that yet.

And I don't, like I said, I don't think that it will be because there will always be like personal preferences and quirks that you prefer, that you want to change.

So that's something I'm quickly reminded of because I have been trying this method more where it's like, okay, don't stop, keep doing it until this certain point and just letting the AI go from there.

Even for my text graphic stuff that I've created recently for my main channel video, which is still not uploaded.

Oh my god.

The whole month, basically.

It will be because they are coming along.

But if I were to guess, I've done at least 10 back and forths with the AI and creating these with Codex specifically.

Currently using Sol.

I've gone back and forth at least 10 times and I am juggling a lot of projects, which is probably not a great thing.

I should keep focused on this one project and that'll probably make it go faster.

But either way, it does take a lot of iterations before it gets to a point where I'm like, yes, this is it.

This is what I will actually use in my main channel video.

Again, for these ones where it's my second channel, I'm okay with just testing things out in a more, I guess, open-minded way where it's like, if it doesn't look that great, it's fine.

It's just a test and it's just to see.

For instance, my color grading overall has improved.

but it still needs work and that's mostly me adjusting it and then telling and then changing the preset so then when AI builds the new timeline it uses the new version so it's kind of me improving my own skill like while using these AI tools and then like iterating I guess my own skill as well folding that into the mix So that's what I feel like I'm learning is that despite using these tools a lot, I feel like the thing that's changing the most is actually me.