Why is primitive substitution encryption not used anywhere? Unobvious flaws?

Question

Computer networks

What encryption am I talking about:
1(shared secret) + 2(recovery information) = 3(data).
3 And 1 - on client A.
2 - in the cloud, or transmitted in any way.
1 - on client B. (Of course, I include the need for a shared secret as an obvious drawback)
"1" is not a small 4096-bit "key", but an excessively large data array, say, 100Mb.
Thus, it seems to me (naive?) It is possible to perfectly implement a completely secure chat (after sharing a shared secret), synchronization between different own devices, and in general a huge number of cases. Why isn't anyone doing this? Or what am I misunderstanding?
What do I see as an advantage, etc.:
Of course, there are AES and other seemingly stable algorithms, with hardware acceleration and all that. But I don’t understand how any algorithm can withstand a mitm (men-in-the-middle) attack in general, because one way or another I transfer both the key and the data through the provider, SORM, and it’s still unknown what. Yes, and in all these algorithms there are probably some bookmarks, probably known not only to "who is supposed to." And how to crack such an algorithm (of course, if it is naive to assume that there are no backdoors on both sides) - I can’t imagine.

Answer 1

Congratulations! You have discovered the cipher pad.
The obvious disadvantage is that from the point of view of security in this cipher (Vernam) the key cannot be reused, which means that for correspondence you will have to have a copy of the key of everyone with whom you correspond, and even develop a way to secretly update the key when it ends.

Answer 2

Initially, the parties must exchange the key formula in advance.
1. The easiest way is to generate the key of the "infinite" notepad based on the date of the sent message using the formula.
2. The formula must generate EVERY TIME! a unique "key" sequence for the same message (through a random sequence of a piece of information in a formula).
3. The "key" can also be partially used, allowing the method of "selection" from several possible combinations, to restore its true value at the other end. (partial coding with incomplete "key")
PS: By default, for formulas, you can use authorization confirmation dates (so as not to "overload" the user's brain).