Storing together is storing cheaper.

Memory stopped being cheap: AI is soaking up RAM and NAND production, and by the chipmakers' own warnings the shortage runs through 2027 and beyond. Scattering data and reassembling it with indexes and caches gets expensive. xyzDB does the opposite — it stores together what belongs together — and does more with less: one read instead of many, tiny containers, fast even on mechanical disks. No tricks: f32, lossless.