Imagine you're at a potluck and two dishes remain: a spicy curry and a mild pasta. You both reach for a dish, but you like heat and your friend doesn't. When you grab the curry, that's not bad news — it's confirmation you got the thing that suits you. That's the mechanism here, transplanted into auction theory. Chen, Nguyen, and Shum introduce the 'winner's bliss' — a direct inversion of the winner's curse that has dominated common-value auction theory since Milgrom and Weber (1982). The standard result says: if you won, you probably overestimated the object's value, because you only win when everyone else thought it was worth less. This paper asks: what if bidders aren't interchangeable? What if they have horizontally differentiated preferences — they value the same object differently not because of information asymmetry, but because of genuine taste variation? In their two-bidder parameterization, winning actually conveys good news about the object's match quality to the winner. You won because the object fits YOUR preferences better, not because you were the biggest sucker in the room. The informational content of victory flips sign. This is the core theoretical contribution, and it's a clean, elegant inversion rather than a marginal extension. The downstream implications are where things get interesting for mechanism design. Under horizontal differentiation, information disclosure by the seller actually reduces revenue — the opposite of the standard linkage principle. In classical common-value auctions, the seller benefits from revealing information because it reduces the winner's curse discount. Here, more information erodes the winner's bliss premium that bidders are willing to pay. The seller is better off keeping bidders partially ignorant of match quality. The paper also shows that advantageous selection — not adverse selection — sustains bilateral trade under asymmetric information. This is a direct challenge to the Akerlof lemons logic. When preferences are differentiated, the buyers who self-select into trade are the ones for whom the object is genuinely a good match, not the ones exploiting informational advantages. Trade doesn't collapse; it improves. Methodologically, this is pure theory — a specific two-bidder parameterization with closed-form results. There are no experiments, no empirical tests, no simulations. The validation is mathematical proof within the model. The strength is the clarity of the inversion; the limitation is the restrictive parameterization. Two bidders with a specific preference structure is a long way from a general N-bidder result. The real question this paper opens is whether horizontal differentiation is the norm rather than the exception in real markets. Housing, art, collectibles, spectrum licenses, ad slots — these are all markets where bidders plausibly have differentiated preferences. If winner's bliss is the typical case and winner's curse the special case, a lot of mechanism design advice (revenue-maximizing disclosure rules, reserve price theory) needs revisiting.