Derek Thompson

I count you in that category.

And so I wanna give you a platform to imagine you're speaking to an audience of not just journalists, but an audience of Americans, people around the world who don't believe that these tools can really do anything useful because at the end of the day, they're just fancy little pieces of autocomplete.

Why are they wrong?

And what have you built that you think has real value?

I don't want to belabor the point because this is not an AI podcast and we do plenty of AI podcasts here, but I do want to recircle what I think is a really, really important point for AI skeptics out there who believe that this technology simply doesn't do what some people promise it does.

In my experience, there are a lot of people that have essentially used a model that is two to three years old.

They used the first edition of ChatGPT, which is 3.5.

We're now up to like 5.3 or something.

That technology was interesting, but incredibly flawed.

It did hallucinate.

It did fail to offer links.

It did fail to serve as essentially a good research partner.

It was not ready for primetime.

But to reject that technology...

and then say that all future AI is essentially equivalent to me is like essentially trying out a cell phone in 1989 and then deciding that the smartphone of 2026 doesn't work.

Like these are essentially entirely different technologies.

And I would just encourage people who are skeptics, pay the $20.

I think Jim is right.

Pay the $20 and see what you can build.

Last question, because let me just bring AI close to the topic that we've been discussing today.