Computational power of extended PDAs: two-stack PDA is Turing-equivalent

One stack = exactly the CFLs; TWO independent stacks = a full Turing machine. The number of stacks, not the fancy access rules, is what jumps you past context-free.