NEEDLESPACE

#73

Unsolved7.3 BTC

Still unsolved, and the public key is not known either — so the only approach is generating keys through the interval one at a time and comparing addresses. There are 272 keys in the interval, so sweeping all of it costs that many attempts (on average, half).

Address

12VVRNPi4SJqUTsp6FmqDqY5sGosDtysn4

mempool.space
Key range (hex)

0x1000000000000000000 ~ 0x1ffffffffffffffffff

Interval size

2^72 = 4.7×10^21

Balance

7.30013849 BTC · 6 tx

snapshot from 2026-08-23

Method of attack
Address brute force
work ≈ 4.7×10^21 (2^72)
Browser (JS, 1 thread)the search tool on this site8.3×10^9 years
CPU · keyhunt, 8 threads18,705,345 years
CPU · keyhunt, 16 threads9,352,673 years
GTX 1060748,214 years
RTX 3070187,053 years
RTX 309099,762 years
RTX 409059,857 years
RTX 509037,411 years
RTX 4090 × 105,986 years
RTX 4090 × 100599 years
RTX 4090 × 1,000the scale of a large pool60 years
All approximate. Implementation, clocks and target count move these by multiples, so read the order of magnitude and nothing finer. The calculator lets you put your own rate in.
Finding it is not the same as keeping it

A transaction moving puzzle funds carries the public key and the signature. The moment it reaches the mempool, bots take the same coins with a higher fee. That is exactly how #66 and #69 were lost. How to prevent it

Doing it

How to actually run this puzzle

Recommended: BitCrack (GPU)

With no public key there is nothing to do but generate addresses one at a time and compare. BitCrack does that on the GPU. On NVIDIA that means the CUDA build, cuBitCrack.

0. Build (Linux · CUDA)
git clone https://github.com/brichard19/BitCrack
cd BitCrack
make BUILD_CUDA=1
# On Windows, open BitCrack.sln in Visual Studio and build the cuKeyFinder project
# to get cuBitCrack.exe. macOS is not supported.
Run
./cuBitCrack -d 0 -b 288 -t 256 -p 1024 -c --keyspace 1000000000000000000:1ffffffffffffffffff --continue progress-73.txt -o found-73.txt 12VVRNPi4SJqUTsp6FmqDqY5sGosDtysn4
Watch out for
  • The best values for -b/-t/-p differ per GPU. Run it once at the defaults, look at the rate, then adjust.
  • Always include --continue. It keeps your position across restarts.
  • To split across machines, turn on “split the work” above. --share divides the range without overlap.
  • This shard covers 1000000000000000000 ~ 1ffffffffffffffffff.
Other options

CPU only. Handles both addresses and public keys, and is the easiest to install — good for learning and for checking your setup.

Kangaroonot usable

This puzzle's public key is not exposed. There is no discrete logarithm to solve.

A reality check

On a single RTX 4090 this takes 59,857 years. That is not “eventually”; in most cases it means “no”. Doing it anyway means accumulating GPUs or joining a pool.