Ciro Santilli @cirosantilli 37

We define a "Procedural AI training game" as an AI training game in which parts of the game are made with procedural generation.

In more advanced cases, the generation itself can be done with AI. This is a possible Path to AGI which reduces the need for human intervention in meticulously crafting the AI game: AI training AI.

Now that's some basal shit! It's basically a fucking blob!!! Except that it is flat. No nervous system. Not even tissues. It is basically a multicellular

For the love of God, on Ubuntu install from the official AppImage downloaded from electrum.org/#download, not this random outdated Snap snapcraft.io/electrum:

Quick overview at stackoverflow.com/questions/1780599/what-is-the-meaning-of-posix/31865755#31865755

Exmples under c/posix:

The Reuters article directly reported only two domains in writing:

But by looking at the URLs of the screenshots they provided from other websites we can easily uncover all others that had screenshots, except for the Johnny Carson one, which is just generically named. E.g. the image for the Chinese one is www.reuters.com/investigates/special-report/assets/usa-spies-iran/screencap-activegaminginfo.com.jpg?v=192516290922 which leads us to domain activegaminginfo.com.

Oleg Shakirov later discovered that the Carson one had its domain written right on the screenshot, as part of a watermark present on the original website itself. Therefore the URLs of all the websites were in one way or another essentially given on the article.

The full list of domains from screenshots is:

From The Reuters websites and others we've found, we can establish see some clear stylistic trends across the websites which would allow us to find other likely candidates upon inspection:The most notable dissonance from the rest of the web is that there are no commercial looking website of companies, presumably because it was felt that it would be possible to verify the existence of such companies.

Most domains are the only domain for its IP, i.e. the websites are mostly private hosted. However we have later found many exceptions to this general indicator, so it should not be used as a strong exclusion rule.

It would be fun to actually reverse search into one of their stock image provider's original images. Ones we've found:

One promising way to find more of those would be with IP searches, since it was stated in the Reuters article that the CIA made the terrible mistake of using several contiguous IP blocks for those website. What a phenomenal OPSEC failure!!!

The easiest way would be if Wayback Machine itself had an IP search function, but we couldn't find one: Search Wayback Machine by IP.

viewdns.info was the first easily accessible website that Ciro Santilli could find that contained such information.

Our current results indicate that the typical IP range is about 30 IPs wide.

E.g. searching: viewdns.info/iphistory and considering only hits from 2011 or earlier we obtain:

Neither of these seem to be in the same ranges, the only common nearby hit amongst these ranges is the exact 68.178.232.100, and doing reverse IP search at viewdns.info/reverseip/?host=68.178.232.100&t=1 states that it has 2.5 million hostnames associated to it, so it must be some kind of Shared web hosting service, see also: superuser.com/questions/577070/is-it-possible-for-many-domain-names-to-share-one-ip-address, which makes search hard.

Ciro then tried some of the other IPs, and soon hit gold.

Initially, Ciro started by doing manual queries to viewdns.info/reversip until his IP was blocked. Then he created an account and used his 250 free queries with the following helper script: cia-2010-covert-communication-websites/viewdns-info.sh. The output of that script can be seen at: github.com/cirosantilli/media/blob/master/cia-2010-covert-communication-websites/viewdns-info.sh.

Ciro then found 2013 DNS Census which contained data highly disjoint form the viewdns-info one!

Summaries of the IP range exploration done so far follows, combined data from all databases above.

This is a dark art, and many of the sources are shady as fuck! We often have no idea of their methodology. Also no source is fully complete. We just piece up as best we can.

In order to explore IPs in known IP ranges, what we need are good DNS databases.

In this section we document the outcomes of more detailed inspection of both the communication mechanisms (JavaScript, JAR, swf) and HTML that might help to better fingerprint the websites.