Yes, that's much better...but still really strong. Resolving 2 of them against a deck with few basic lands could very nearly mill them out. It's scary how much better each successive copy of that is.
Basically, the test of how good this kind of card can be is to see what it does to a deck that is fifteen 4-ofs. thorme's card is still really good against such a deck, so it's not really "highlander hate".
15 4-ofs seems an okay measure, but the metagame might also contain decks with 15 identically named basics, in which case the card's value rapidly declines. Doing some quick rough estimates of probabilities, given a deck of 15 4-ofs, this card would mill 7 or less cards about half the time, and 10 or less 90% of the time. It could only ever mill 16 at most (you're bound to hit a duplicate). I tried a few ways to model other types of deck construction, but didn't get far in the few minutes I spent. I did find that adding a dozen 1-ofs in place of 3 4-ofs doesn't really seem to significantly affect the expectations. The concern would seem to be about the extremes-- playing against a true highlander deck, or something like Stephen Houdlette's 2nd-place Tezzeret list I got from the SCG database just now, containing 2 4-ofs, 5 3-ofs, 4 2-ofs, and no other duplicates.
I must admit to kind of liking thorme's card if for no other reason than it's profound metagame warping effect. It's not really much different than Archive Trap or Glimpse the Unthinkable most of the time. But if you're playing a deck that is weak to it, you lose if it resolves. The only parallel I can think of is the efficacy of Leyline of the Void against graveyard strategies. I'm nor sure how many cards like that Magic as a whole should aim for. How tedious would it be to have sideboards consist of 4x Leyline, 4x This, and then other such cards that just annihilate a specific kind of strategy. The strategy this seems to work best against is either reliance on restricted cards or over-dependence on tutoring and toolbox answers. The former only applies to Vintage, and I'm not sure the latter is all that desirable. Toolbox archetypes are a positive aspect of the game, I think. They encourage finding uses for underplayed cards, decision making during game time, and other of the more "interactive" aspects of the game.
If I'm feeling math-y later I might try to compute the cumulative probability of milling X cards for a deck with numbers like the Tezzeret deck. I don't expect it to have more than a 5% chance of getting more than 15 cards.
