Matt Bevan

When it was used to manage logistics and supply chains, it looked like Factorio.

When it was used by law enforcement agencies,

Well, an ICE official said it's basically like a Google's map interface where you can look around the United States.

You can zoom in on targets.

You then click on an individual person and it brings up their name, a photo.

We'll get into how it's being used for immigration enforcement in our next episode.

But looking at it dispassionately, it's clearly impressive software.

way better than the PowerPoint nightmares they used to deal with.

But programming these maps, which Palantir calls the ontology, was slow, painstaking work.

It involved writing incredibly complicated software and talking to a lot of different people, so they didn't overemphasise the importance of any one particular data source or ignore another one.

It also didn't entirely solve the marble jar problem.

Because each industry needed information laid out in different ways, the systems didn't talk to each other perfectly.

For instance, Gotham, Palantir's military and law enforcement software, was different from its logistics software, which was called Foundry.

But then something came along that was going to put Palantir's marble sorting technology into hyperdrive.

Let's start with this language AI, ChatGPT.

What is it?

Great question.

In 2022, people were starting to get excited about the potential uses of large language models like ChatGPT.

It can create custom code.

It can create entire books that are written from scratch, that are not plagiarised.