A SWARM OF GPUs.
ONE UNSOLVED KEY.

In January 2015, a Bitcoin holder known as saatoshi_rising funded 160 addresses in one transaction, each sitting inside a private-key range that doubles in size from the last — address #1 covers a 2-bit keyspace, #2 covers 4 bits, and so on up to #160. It's a real-money, public demonstration of how fast a keyspace grows out of reach: each step is roughly twice as hard as the one before it. Most of the lower puzzles have been claimed over the years; everything above roughly #130 is still considered out of range for a single machine.

Puzzle #140's address — 1QKBaU6WAeycb3DbKbLBkX7vJiaS8r42Xo — already has its public key exposed on-chain, which is what makes an attack possible at all. Instead of blind guessing, solvers run Pollard's Kangaroo algorithm, a collision-finding method for the elliptic-curve discrete-log problem: two "herds" walk across the range until one lands on a distinguished point the other already visited, and that collision reveals the key. GPUs parallelize that walk across millions of points a second, and pooled efforts — the model CollisionProtocol is built on — split the range across many workers and share whatever any of them finds.