Hacker News new | past | comments | ask | show | jobs | submit login

Is it really the case that data does not cross boundaries among Palantir clients? And that the algorithms used by any given client are only informed by the data that client provides?

When I worked there (mid 2009 to early 2014), almost all customers had dedicated installations (as in a rack of servers) with just their data on-site. Even the hosted instances had siloed data. There are features for peering, but AFAIK that has to be explicitly set up between installations.

I’ve used them on prem and hosted. For on prem, definitely. It’s like any other big install of software (sql server cluster, oracle financials). All on site, no data out. Brought in 3rd party data, shared no data out.

Cloud same, but they host and promise not to do anything. Like any other enterprise cloud product.

The only data they gave us was either open sets like weather or census or stuff anyone can buy like Quintiles. So there was no “magic data set” we could mix in or something.

If so, then Palantir is missing a big revenue source: integration with public data that each client would otherwise have to collect on its own.

Palantir is a consulting company, plain and simple. They don't do anything IBM couldn't, they just (plausibly) do it better than IBM.

Unless proven otherwise, then that is hearsay.

Good point, but this is the risk for every cloud hosted software. So every AWS and azure product has this suspicion that’s impossible or extremely difficult to remove.

If you’re really paranoid, then all non-OSS software.

So again, why is Palantir more worrisome than Salesforce?

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact