I think he meant ML as the family of meta programming languages?

He did say systems though so you are probably right

Also, the part about "Input some learning data" makes me think to Machine Learning rather than Meta Language.

Btw, Prolog and ML might look like distant cousins: e.g. you use recursion and pattern matching in both. But under the hood there are big differences that arise from Prolog's logic programming paradigm: for instance, you can use append(L1, L2, L3) both to concatenate and split a list, depending on the variables you set:

  append([1,2], [3,4], X)       % ==> X=[1,2,3,4]
  append([1,2], Y, [1,2,3,4])   % ==> Y=[3,4]

They should break it up by time. For example your database only syncs the past month or so as needed and you can choose to request more if necessary. Someone mentioned 150MB right now. That thing is going to get massive eventually.

The quoted 150 MB is for all messages in one's network. On my local node there are 91773 messages, going back roughly a year and a half, taking up 147 MB - of which about 72 MB is the actual messages, and the rest are indexes. gzipped, the 72 MB of messages goes down to 29 MB.

Express and phoenix maybe. Elixir is not a framework. Node is to js perhaps as plug is to elixir. Even that kind of doesn't compare correctly though.

node is to js as beam is to elixir, roughly.

Node is not a framework though

Could it be because older content is not cache prioritized as much as new content? I would imagine anything after 1 day gets accessed much less than new content.