Coverage is evidence, not a prediction.

What the homepage measures

The live statistics count database rows by lifecycle state and sum the number of keys represented by completed intervals. “Ranges completed” is a count of verified work units, not a count of keys. “Range completion” is the completed key total divided by the full keyspace of the selected puzzle. Those two values answer different questions and should not be used interchangeably.

Why public ranges are shortened

The activity table intentionally masks most hexadecimal digits. Visitors can confirm that work is distributed across the keyspace and see when a range completed, but the public page does not need to expose every exact boundary. The central database retains the full interval for overlap and completion validation.

Network speed is a recent aggregate

Network speed represents recent progress reports from active assignments. It can fall to zero between heartbeats, during maintenance or when no lease qualifies as active. It is not a benchmark database and it should not be used to compare participants. Local benchmark speed is primarily used to size work for an approximately consistent duration.

Operational observations

• Long-lived clients reduce setup overhead by claiming a new range after completion.
• A safe stop produces cleaner evidence than terminating an engine process manually.
• GPU temperature, power limits and background workloads can materially change sustained speed.
• A short benchmark sample may overestimate performance; validation samples are therefore important.
• Cancelled ranges must be recycled from their beginning to avoid assuming partial coverage is complete.

What the numbers cannot prove

Completing a fraction of a puzzle does not increase the probability that the next range contains the key. Each unsearched key remains one candidate among an enormous set. Coverage demonstrates which work has been performed; it does not provide a schedule or guarantee for discovery.