HUNT
Hierarchical Unified Navigation Technique
Bitcoin Puzzle #71 - Pollard's Kangaroo meets graph-based coordination
Bitcoin Puzzle #71 (address: 1PWo3JeB9jrGwfHDNpdGK54CRas7fsVzXU) is one of the largest unsolved cryptographic challenges, with a keyspace of 1.18 quintillion keys. Pollard's Kangaroo algorithm reduces this from O(n) brute force (37,000+ years) to O(√n) (~34 billion operations). With Memgraph-based coordination and distributed agents, we're hunting for collision points that reveal the private key.
The Challenge
In 2015, someone created a series of Bitcoin puzzles by sending funds to addresses with progressively harder-to-find private keys. Puzzle #71 has remained unsolved for years.
Traditional brute force would take 37,000+ years on a single CPU. But with Pollard's Kangaroo algorithm (O(√n) complexity) and distributed coordination via Memgraph, we can potentially solve this in days with the right compute power.
Architecture
Memgraph stores distinguished points from all agents, enabling collision detection across the distributed swarm.
How Kangaroo Works
Tame kangaroo starts at known point
Starts at random k in range [2^70, 2^71], hops forward deterministically using a jump table
Wild kangaroo starts at target public key
Starts at the target public key (puzzle address), hops forward using the same jump table
Both leave "distinguished points" in Memgraph
When a point matches a pattern (e.g., last 20 bits are zero), it's saved to Memgraph with distance traveled
When wild lands on tame's trail → collision detected
Memgraph finds matching distinguished points from tame and wild kangaroos
Key recovery
Mathematical formula reveals the private key:
Performance Estimates
| Configuration | Hops/sec | Time to Solve | Estimated Cost |
|---|---|---|---|
| 1 CPU (Python) | 50K | 21.5 years | Free |
| 1 CPU (secp256k1-py) | 500K | 2.2 years | Free |
| 1 GPU (RTX 4090) | 50M | 7.9 days | ~$500 |
| 4 GPU (RTX 4090) | 200M | 2.0 days | ~$2,000 |
* Estimates based on secp256k1 point operations and expected collision time for O(√n) complexity
Agentverse Deployment
Autonomous agents running on Fetch.ai's decentralized agent platform with cloud hosting for coordinators and mailbox agents for local GPU workers.
Central coordinator manages distinguished point storage in Memgraph, collision detection, and task distribution to worker agents.
Local agents with GPU acceleration run Kangaroo algorithm, submit distinguished points to coordinator, and scale horizontally.
Graph database optimized for fast collision detection between tame and wild kangaroo distinguished points.
Origin Story
What started as a 1-hour hackathon project is now hunting $600K+ in Bitcoin.
CrackRSA32
Built in 1 hour at AI3 Chiang Mai hackathon, proved 20x speedup using stigmergic ant colony optimization. Ants left pheromones on promising search paths, enabling emergent coordination without central control.
Read the CrackRSA32 storyants-worker
Abstracted the stigmergic pattern into a PyPI package. Anyone can contribute compute by running a single command. Workers connect to api.ants-at-work.com, sense cold regions (low pheromone), and deposit distinguished points.
pip install ants-worker && ants-worker join BTC Hunt
Same stigmergic pattern, now hunting Bitcoin Puzzle #71 (7.1 BTC = $626,545). Tame kangaroos explore from known points. Wild kangaroos start at the target public key. When their trails collide in the graph, the private key is revealed.
Workers self-register, receive bearer tokens, sense cold regions, mark intentions, and deposit distinguished points to the gateway.
The Pattern: Hackathon prototype → Production package → $600K+ hunt. All powered by the same stigmergic coordination that ants use to find food.
Pattern Analysis Results
Before committing to the Kangaroo approach, we ran extensive pattern analysis on solved puzzles to look for exploitable weaknesses:
- ✓ Bit distribution is uniform: No bias toward 0s or 1s in binary representation
- ✓ No sequential correlation: Adjacent keys show no predictable patterns
- ✓ Entropy consistent with cryptographic randomness: Passes all standard randomness tests
- ✓ ML found no exploitable patterns: Neural networks couldn't predict next bit better than random
Conclusion: Pure Kangaroo is the only viable approach. No shortcuts exist.
Join the Hunt
This is open research. Contribute compute, improve algorithms, or study distributed coordination. The hunt is stronger with every new agent.
"37,000+ years of brute force, or days with O(√n) elegance."
When the collision emerges, it won't be because any agent was clever. It will be because the math was sound and the coordination was distributed.