The evolution of P/F vs T/K raises important questions

[npcompl33t]

One simple fact may explain most of this. Plain p and f are almost never followed by e: in the whole manuscript I count 5 cases of pe/fe against about 5,000 of ke/te. It holds in Herbal A, Herbal B, Bio and Stars, in head lines and elsewhere. And it isn’t a head-line effect: k/t on head lines take e just as often as k/t anywhere else.
@npcompl33t’s gradient is to a large extent the rise of e after gallows: k/t+e goes from about 16% of k/t in Herbal A to 57% in Pharma. A glyph that can’t take e will look more and more “behind” as e spreads, so P drifts away from K/T along the gradient without needing any archaic phase. It also rules out the simplest version of the “fancy k/t” idea: a random swap would leave about a third of puffs followed by e. What does seem to hold is that where k/t take e, p/f take ch: e plus ch comes to about the same share after both (52% vs 60%), which fits Mauro’s small ke–pch distance.
My own generator did the random swap, and that’s how I noticed: it wrote 24% pe/fe on head lines, where the real text has 0.4%. Fixing it now.

This does not explain this, unfortunately. Because P words do not change in other ways that k and t do. For instance, kch and tch drop in popularity, while pch does not. 

basically K/T change while P seems to not change as much, or change in different ways, even with no e involved

the rise of e after gallows is also only one of about 13 different peices of the gradient. I finished mapping it out extensively in terms of glyph components, and it they are each active in different phases. 

   

These 13 different component groups combine into about 15 word families. 

   

Full gradient components and mappings can be found You are not allowed to view links. Register or Login to view..  And repo is You are not allowed to view links. Register or Login to view..