THIS BLEW MY MIND

XOR’s “memory” and cancellation properties let a missing-number problem be solved without maps or sets—and the same core trick scales to duplicates and even two missing/duplicated values. The key insight is that XORing a multiset of values causes pairs of identical elements to cancel to zero, so only values with odd occurrence counts survive. That turns many seemingly different interview questions into variations of “XOR everything, then use bit-level structure to isolate what’s left.”

The transcript starts by walking through XOR on bit patterns and then generalizes the behavior to integers: XOR is applied bitwise, so it works on more than booleans. From there, several foundational identities drive the method: x XOR 0 = x, x XOR x = 0, and XOR is commutative (order doesn’t matter). The practical consequence is the “duplicate removal” rule: if a value appears twice, XORing the whole collection eliminates it; if it appears once, it remains in the final XOR result.

That rule immediately solves the classic interview task: given an array containing every integer from 1 to n exactly once except one missing value, find the missing number. The approach XORs all array elements together, then XORs that with the XOR of the full range 1..n. Since every present number appears once in the array and once in the full range, they cancel out, leaving exactly the missing integer. The transcript emphasizes that ordering doesn’t matter—XOR’s commutativity makes the method robust.

The same cancellation logic connects to forward error correction (FEC). By XORing multiple packets into a single parity packet, losing one packet can be recovered by XORing the parity with the remaining packets. The missing-number algorithm mirrors this: XORing a “complete set” with a “received set” cancels everything that matches and leaves the missing piece.

Next comes in-place swapping: using XOR, two variables can be swapped without a temporary variable by applying a sequence of XOR assignments that effectively reconstruct each value from the other. After that, the transcript extends the missing-number idea to duplicates. If one number is duplicated instead of missing, XORing the array with the full range leaves the duplicated value (because it appears twice in the array and once in the range, producing an odd-count residue). For two missing numbers (or two duplicated numbers), XOR alone yields only their combined XOR (U XOR V), so the method adds a partition step: find a distinguishing bit (the least significant set bit of U XOR V), split both the full range and the array into two groups based on that bit, and then solve the “single missing” problem inside each partition to recover U and V separately.

The transcript ends by noting the limits: once more than two values are missing or duplicated, the simple XOR partitioning strategy breaks down because multiple combinations can produce the same XOR pattern, requiring more complex techniques beyond XOR-only reasoning.