My response was "somewhere in the middle", because it seems like a good part of what Mike Mearls is musing about already exists in the game--just written in a different way and/or in a different spot--and the other stuff seems fairly situational (likely making it harder to remember, especially if you get a lot of "talents"). I do like the idea of making it more explicit as to what abilities can do; 4th Edition kind of glosses over them and sticks to direct benefits, such as telling you what defense or skill(s) it affects as opposed to giving a broader explanation.
As it stands you can play Dungeons & Dragons without skills at all and just default to ability scores (for example instead of Athletics, just use Strength), and characters can already at least try any skill application except for detect magic (Arcana) or reduce falling damage (Acrobatics). Aside from that being trained in a skill just drastically increases the odds that you will succeed. With the proposed system you still get small bonuses, but you also get some sort of passive benefit.
I like the concept that Mearls is going for. Groups that want a skill-less game can do so, while groups that want skills can choose their complexity (skills as bonuses, skills with benefits, etc). Even better groups can dial the complexity up and down as they want without having to switch between editions. The problem is that groups can already do something very similar with relative ease by removing skills from the game (modifying DCs, of course), sticking with skills, or using skill powers.
My final verdict would depend on whatever edition of the game this ruleset is actually intended for; in 4th Edition I would not like it because players already have to deal with a lot on their plate, and adding skill talents to feats and powers would just bog down things even further. I recall something similar near the end of 3rd Edition, though I never used it and I do not think it was particularly popular. In a theoretical 5th Edition though, who knows?