A public ledger usually makes verification easy by exposing the facts needed to check a transaction. Anyone can inspect where value came from and where it went. That transparency is useful for consensus, but it also creates a durable record of financial relationships.
The interesting question behind Zcash is whether a network can verify the rules without seeing all of the underlying information. I saved its code because that separation feels like a more precise use of cryptography than the vague promise of anonymous money.
A zero-knowledge proof lets someone demonstrate that a statement is true without disclosing the private values used to establish it. In a payment system, the relevant statements include authorization to spend and preservation of value. The network needs assurance that the transaction obeys those rules. It does not necessarily need the plain contents of every note involved.
For an intuition, imagine a sealed accounting exercise. The verifier must know that the inputs belong to the spender, that they haven't already been spent, and that the amounts balance. Opening every envelope would answer those questions, but it would reveal much more. A cryptographic proof is a different route to confidence in the checks. The sealed-envelope picture is only an analogy; the security comes from a precisely defined protocol and its assumptions.
The Zerocash research is useful background for that design. Commitments represent hidden information, while separate mechanisms allow the network to reject repeated spending without publishing the original secret. Hiding transaction content doesn't mean abandoning an auditable set of validity rules.
The boundary of the guarantee matters. Zcash supports transparent as well as shielded activity. Moving value between those regions can reveal information that a fully shielded transfer keeps hidden. Timing and distinctive amounts may give observers clues even when they cannot directly read the shielded transaction's participants.
A simple hypothetical illustrates the problem. One visible transfer enters a protected pool, and a matching amount leaves shortly afterward. The ledger may not contain an explicit link between them. An observer can still form a hypothesis. The strength of that hypothesis depends on other activity and on what information the observer has outside the ledger.
Network metadata is another layer. Encrypting or hiding transaction fields doesn't automatically conceal which machine first transmitted a message, what a service knows about its customer, or what a recipient later chooses to disclose. A privacy claim needs a subject: hidden from whom, at which layer, and under what assumptions?
I find the technology most compelling when described at that level. Proving validity with less disclosure is already a substantial change. Understanding exactly where that disclosure stops is part of understanding the feature itself.