Now, on to the actual mathematics!
"Understand thus..." says the first line of the treatise, giving the definitions of the earliest level; of course, without a formalized tetrational notation, he had to improvise. In modern notation, we can summarize: hn=n^^n (now we would call the same function "megafuga").
Now we have hhn=h(hn), hhhn=h(hhn), and so on; culminating in the general case mhn=hm(n).
[I use the superscript here in the modern sense; a different meaning occurs in the treatise itself - though that doesn't really create a conflict, since it fortunately doesn't include this case!
Specifically, it does not include cases like hxn, with only one h to the left of the superscript and one number to the right - which, for obvious reasons, includes all the cases where I'll be using it in the modern meaning. Which is kind of lucky.]
Note that hn=n^^n=n^^^2, hhn=(n^^n)^^(n^^n)>n^^(n^^n)=n^^^3, and similarly 3hn>n^^^4, ..., mhn>n^^^(m+1) (which in turn is more than n^^^m). So we're already in pentational territory.
And the next step brings us to the limits of pentation: let hmhn=hhmn; this is more than n^^^(m^^m).
In turn, we can have hhmhn=hhhmn, and so on, the number to the left of the last hn being the number of iterations for the h function.
This means that hkhmhn=hhhkmn; we can easily iterate this further, to outrageous examples like h7h2h15h4h5h8h9=hhhhhhh72154589 (for all I know, this number, being the largest listed specifically in the treatise - further examples always involve a n somewhere - might well have been the largest number ever specifically referred to in all of mathematical history up to that year).
Now we are certainly passing into hexational territory - already nhnhn>n^^^(nhn)>n^^^(n^^^n)=n^^^^3. Similarly, nhnhnhn>n^^^(nhnhn)>n^^^^4, and so on.
At this point, Kharms focuses on a specific case - that where all the coefficients m in hmhm...hmhn are identical; it is for this case that he used the notation mhkn (where the sequence "hm" repeats k-1 times). In particular, for m=n, we have hnhn...hn=nhkn (where the sequence "hn" repeats k times).
It is this case - and specifically its version where k=n - that Kharms uses as the next function, nhnn=tn.
We can easily see that nhkn=hnhk-1nn (this combines the original and modern superscript notations); then nhkn>n^^^(nhk-1n). Since nh2n=hnhn=hhnn>n^^^(hn)>n^^^n=n^^^^2, by induction we have nhkn>n^^^^k, and tn>n^^^^n (though not by much, googologically speaking - it is probably less than n^^^^(n+1)).
Kharms continues, defining the first-level circle notation, which is already as powerful as the full might of (non-iterated) Knuth's up-arrows; but for that, see part 2 (if I ever decide to write it).