[fapjacks]

I know how Palantir works under the hood because I was in the infantry and conducted many missions using information "curated" by various intelligence organizations and Palantir. Very easy to tell we were looking at intelligence products generated by Palantir because the company is (was?) fond of putting its logo all over the stuff it makes. What Palantir does is completely off-the-shelf kind of stuff that Google and Facebook do all day. Except that Palantir gets to feed classified metadata into its software to generate its graphs. It was impressive a decade ago. Today it's nothing that any HNer taking CS classes at university couldn't do over spring break. Don't get me wrong, it's still very useful, crucial stuff for battlefield intelligence. But there is absolutely no magic happening behind the curtain.

Incidentally, I would like to take this opportunity to state in no uncertain terms that this is an extremely dangerous thing for domestic law enforcement to be using against American citizens. Or anyone anywhere that we aren't trying to kill. We should be fighting tooth and nail against companies with access to mountains of metadata like Facebook and Google. I deployed before the widespread use of this kind of analytics, and I deployed after it became ubiquitous. It is the single most powerful force multiplier I can think of, right up there with nuclear weapons. And it will be abused.

[zaroth]

> It is the single most powerful force multiplier I can think of, right up there with nuclear weapons.

Hell of an endorsement.

[fapjacks]

Actually, I appreciate the reply as a prompt to clarify what I mean. I really hope I am able to decouple Palantir from the metadata analytics that I'm talking about in my endorsement of the technology. Speaking in sweeping generalities, the only thing special about Palantir is the fact that it's entrenched, which is what gives it the opportunities other companies don't have. Ignoring for a minute its highly-specialized workforce (needing a TS clearance just to get into the parking lot) and focusing on the process and technology. I would even go so far as to say that I think most other companies in this space would do a better job than Palantir, just because entrenched companies get comfortable and lose their edge.

[WingH]

“Today it's nothing that any HNer taking CS classes at university couldn't do over spring break.”

How can any HNer do this if it relies on classified metadata? And how does Palantir even get this classified metadata? their customers?

[brazzledazzle]

I believe the point is that their tech stack isn’t that unique, it’s access to the data that let’s them apply standard stuff any HNer could come up with.

[wavefunction]

They've been offered a sinecure by politically aligned allies in the Federal government, which makes sense since P Thiel's strategy for innovation boils down to rent-seeking.

[nl]

Palantir gets deployed on classified networks. There’s no central Palantir instance. Those instances get integrated with classified data sources by people with security clearance

And yes, getting access to that data is hard (and should be!) One of the reasons Palantir gets contracts is because of their large cleared workforce.

[AgentME]

Google and Facebook give soldiers intelligence products? Or Palantir operates websites and does analytics of their users? I don't think I understand the comparison.

[zdragnar]

I worked for a competitor to Palantir at one point. Basically, they're using what you can think of as an adaptation of Google's PageRank- instead of ranking how important websites are by evaluating the links between them, Palantir and the like estimate a person's importance / influence by evaluating their connections to other people.

You can do some pretty interesting market analyses for commercial purposes using just data from Twitter, or Facebook if you can get access to it. What Palantir can do with covertly / illicitly gained data is much the same, only with much deeper wells of information to draw from.

[WingH]

What are some examples of interesting market analysis you can do this with this data?

[zdragnar]

That depends entirely on the product and market, though there are a few general uses. Generally speaking, find the people who are very passionate and vocal about topics your product addresses.

Now, you know two things: the types of words they use, and with a bit of digging, the sentiment typically associated with those words. This can influence your marketing copy and promotions.

Repeat with a few demographics to identify associated areas of passion. Now, you have a group of very passionate people, and maybe even a loose network. Dig in and see if you can find more common associations to tighten the network up a bit.

From there, you could run a targeted campaign- say, a contest with freebies awarded to a number of people who retweet a certain message or hash tag.

People who are tangentially interested hear about your product. People who would have heard about it anyway (by virtue of being in your target market) will now be talking about your product, using your specifically chosen wording.

For the cost of a few freebies, you now have reached your target audience- and thousands more, if they have a significant following. Further, you're not spamming airwaves, or cold emailing, or running ads that are blocked or ignored anyway.

Take all of this with a hefty dose of salt. I'm an engineer, not a sales person or marketer. I'm not on Twitter or Facebook or any other network, so I don't really know all the ins and outs of how to use the product itself, only how to build things those people found useful.

EDIT: for what it's worth, you can do all of this by hand. Doing it well, doing it right, is harder. I won't recommend anyone specifically, but if you're interested in marketing this way, I'd suggest finding someone who already knows how to calculate influence very well and using their services. Otherwise, you'll spend a lot of time on a pointless goose chase, because there is a LOT of noise to sift through to find a good signal, so to speak.

[lallysingh]

Influencer determination. Who's got the most viewed/respected opinion in a group?