I played Academy (formerly known as "Conference Room") against an AI last night. My suspicion was that the other-player benefit was too strong, and I think my test bore that out. I have a proposed fix, but first let me go into more detail.
If you think about it, the opponent benefit is roughly the equivalent of playing this on your own turn:
+2 Cards
+1 Action
Discard a card.
...because when it's not your turn and someone is triggering this for you, you don't need the cantrip part to replace the card slot and action that it consumes. So you're left with just "+1 Card, discard a card." Anyway, recently Donald X was saying that he tried out a card like this at a cost of $4, and it was slightly too strong for a $4 card. That surprised me, but I could see why: you almost always have, in a hand of five, a green card you don't mind getting rid of, and if you don't, you probably have a Copper you don't mind discarding when it means you can have a Gold or something instead. But, true, it's strictly weaker than Laboratory, so let's call it a $4.5 effect.
Well, Academy gives you a $5 effect and your opponents a $4.5 effect, which is net-positive but doesn't seem like a great use of the opportunity cost to pick one up.
In one of my test games, I set up a 3-player game against two AI players, which were set up to pile-drive the Academy pile, and I played a money game without buying any. Soon, the AIs were each able to play 4-5 Academies per turn. What that meant is I essentially got to play with my 5 best cards every turn, because I got to filter 10 times between turns. I bought 6 Provinces in 12 turns, winning the game easily. Interestingly, one of the AI players got 4 Provinces in 12 turns, which is remarkable on its own -- he was getting benefit from his own Academies, plus the ones bought by the third AI player, who only ended the game with 2 Provinces due to a couple of random bad buys.
Obviously this isn't conclusive in itself: just because pile-driving Academies isn't a winning strategy doesn't mean some number of Academies employed differently wouldn't be better. And I also have to concede that the AIs were not designed to stop playing Academies once they'd reached $8 (although I believe I reached $8 after roughly as many plays as it took them). But it's not a promising sign, and I can't personally envision a situation where I'd want to buy an Academy for myself. As surmised above, receiving the opponent benefit of an Academy felt roughly like getting a free Laboratory. I blew through several green cards, replacing them with Coppers or better, and by the time I had to discard Coppers, I was pretty much already at (or well above) the $8 Province-buy threshold.
I did play another game or two wherein I actually bought Academies. Those were more difficult to distill in words, but they certainly felt wrong.
But I think I know a promising way to fix this:
+2 Cards
+1 Action
Each other player may draw a card. If he does, he discards two cards.
This is a much lighter opponent benefit and, perhaps just as critically, one that doesn't scale as well with multiples, such that a chain will cause your opponents to match the benefit you get yourself.
It resembles the Vault penalty but is still stronger, as the draw happens first.
I also like the symmetry: You go down a card (Academy taking up a slot in your hand) and up two cards. Opponents go up a card and go down two cards.
More importantly, I think it preserves the spirit of the original card people voted for.
But I confess I haven't tested that variation yet. It's just something that came to mind and held up under speculative scrutiny. Thoughts?