OP is making reference to the "second layer" of the "charter web", which is a bogus "secret" part of "the deep web".
This dis-information about "the deep web" comes from a 4chan infographic that get's posted there every now and again when technologies such as TOR and I2P get "discovered" again by new users who have questions about these less-than-conventional networking technologies.
The infographic claims that the first three "layers" of the deep web consist of anything from normal content, to unindexed web pages, and other hard to find pages (but omits actual "deep" content like dynamically generated pages)...but still all common HTTP traffic.
The "charter web" is sub divided into two "layers"...TOR hidden services (which do often contain "dark" material, such as underaged pr0n and drug smuggling contacts) and the second layer which consists of "deeper, darker content that requires a closed shell system hardware modification to access"...
90% of the infographic OP is referring to is disinformation, though, spread by either troll who want to make him look stupid, or by criminals who use TOR and need to generate extra network traffic as to make statistical analysis of that network more difficult by law enforcement types. Going even further, a firm grasp of the TCP/IP model will tell you that, yes, network protocols are "layered", but not in the sense that this infographic states...in reality, a TOR packet is layered like so:
(Application layer data; i.e. HTTP, SSH, FTP, etc.)
(TOR onion routing data...read more about it from the TOR project website)
(TCP headers)
(IP headers)
(Netork-access layer frames; i.e. ethernet, wi-fi specific encoding, etc.)