Dan Nottingham

After a great deal of effort to kind of tune it so that it would actually behave consistently the way we want it to, the analysis started coming back very deep and the depth of the information that it was feeding back was more than just a something quick that you might get.

on a prompted version of ChatGPT, it was much deeper.

It was citing sources like, in one case, it was citing Fox News and MSNBC in the same analysis.

You could see it trying to weed out biases.

You could see it actually coming out with credibility scores.

That made sense.

And the exciting moment to me was that I learned, just as I did with that example with that friend, that we had created something that wasn't any of our own worldviews.

All of us working on it were discovering that

None of our opinions were baked into this at all.

We were all learning from it.

And I remember that the first time it came back with the full analysis, with a credibility score that just felt right, it was spot on.

Now we have something.

It doesn't matter what the rest of the product looks like.

That analysis and that credibility score with its source material, that is the product.

That's the information that we're trying to share.

And that was exciting when it all came together and it worked.

Yeah.

So it's, it's, it's a spectrum, right?

It's from zero to 10, 10 being the most credible thing.

And zero being, I can't find anything credible about this statement at all.