Proto-Verrilin

I've been making some progress on a proto-language for the new world I'm
starting to put together. The world is a distant one which has some
connection to Azir although it exists in a different region. The people
who live there, tentatively called Verrilin, are probably related to the
Zireen and Sangari in some way, or more distantly related to the larger
group that also includes Nithra, Hralta, etc. I haven't found out yet
whether they have the full range of color vision (with 4 primary colors)
or some more limited subset of colors; many details are still vague.

Actually, Proto-Verrilin may be a misleading subject, since this is only
one of many proto-languages for the world, but the word Verrilin comes
form one of the descendant languages of Proto-Verrilin, so I'll continue
calling it that for now.

Proto-Verrilin is an unattested, unwritten language from which many of
the classical languages of a significant part of the world are derived,
although there will be a couple of unrelated classical languages. My
plan is to start with the oldest languages, sketch out some general
ideas, and move forward in time to more and more recent languages. I'd
be interested to know if anyone's had much success with that approach.
Before, whenever I've tried to build a history connecting languages
together, I already had one or more modern languages to start with,
which hasn't worked out very well.

The idea is to make Proto-Verrilin a realistic language from the start,
and to use realistic changes in sound and meaning to build the more
recent languages from it. But it doesn't have to be a very complete
language.

What I've got to start with is a couple of vowel alternations. The basic
vocabulary can be described with 3 vowels (a i u), which have allophones
based on the position of stress and other factors. E.g.

*/panat/ ['panEt]
*/ka-panat/ [kEp@'nat]
*/lakus/ ['lakOs]
*/lakus-at/ [lEku'sat]

However, because of syllable-final sounds that have vanished, there are
situations where the [E] that was originally an allophone of /a/
contrasts with an [a] from another source, and where [O] contrasts with
[u]. The [@] has become indistinguishable from [1], an allophone of /i/.

So in the version of the language as I'm describing it, what I end up
with is:

/panet/ vs. /ke-pinat/
/lakos/ vs. /lekus-at/

Does this seem plausible so far? I can see some interesting
possibilities with this approach.