Hello there!

My recent (and quite successful) draft for a conlang was based on one
feature: a rather weird way verbs are conjugated from their roots. It kind
of resembles the Semitic roots, but instead of a consonantal frame for
vowels to be inserted, the root is a legitimate word lacking two
consonants/vowels (one being the evidentiality, the other—tense). There are
some rules regarding when to put what, but otherwise it's quite
For example, t·t·a means "to love". Because I've decided to have a very
simple CV phonology, both of the slots must be occupied by a vowel. If I
want to say "I love you", I'd need the observative evidential and the
factual 'tense', both of which happen to be -a-. Putting that into a

[I observe that, in reality,] I love you.
(-lao signifies that "I" is the subject and "you" is the direct object)

Obviously, this system allows for many different meaning nuances without
making the word too long: tetaàlao would make it "I think I love you";
tiatoaàlao—"It's generally accepted that I'll love you" (as a speculation),
totaoàlao—"I've heard that I was loving you"... and so on.
A better example would be ·e·u, 'to talk'. Since both a consonant and a
vowel can be inserted into the slots, the OBS-FACT meaning can be expressed
as je'ù, jeaù, aeaù or ae'ù, the first one being the most preferable. Thus,
"we are talking" would be

[I observe that, in reality,] We talk to each other.
(-sã (here -sàa because of the accenting rules) means "I to each other")

What interesting root mechanisms do you have? Has any of you had a similar
idea before? If so, how has it been implemented? I'm very interested in
such unconventional solutions.

~ uakci