Logan Kilpatrick

Every time you do it.

Again, we're in AI Studio providing this vibe coding experience to people for free, which is really cool.

But at a certain point, you can't always provide something for free or maybe we'll be able to with Flash.

But that is literally the conversation is maybe with these Flash models and maybe as models get better at coding, that code generation piece actually will just be free because it's so economically possible.

I think you see this in a lot of developer use cases, actually, where I think the fun thing about us having this stuff and putting it in an API is that we get this feedback from developers all the time of like, my use case wasn't possible before.

I needed a model that was like $5 per million input tokens.

um and 20 out and now i have that capability with three flash 50 cents in three dollars out and like my use case that was 10x too expensive to put in production is not possible um so you see with like how interesting the like cost intelligence latency uh trade-off is for for flash which is really exciting

Go ahead.

i can stay for i can stay for a few more minutes but i do i have to run to another meeting so let's let's finish vibe coding this yeah let's test this out yeah i only put in so the prompt i put in originally was um like the first half of my message i set the other two halves of here where you can prompt like an actual screenplay enter ascendance until a unique shot you establish scene heading at the top character speaking indicate shot

you should be able to write a whole page as a single prompt and get a whole page with all the shots.

So I wonder what is needed to actually make this happen.

Parse script.

Interesting.

I don't know how easy this is actually going to be.

This might actually take a minute to generate everything.

Yeah.

I'll copy in your other prompt.

Every time a character talks, the model knows what the character looks like, et cetera, et cetera.

I love that.

What did it end up looking like?