I was checking DUSK prover requirements and noticed the missing number was not speed. It was location.

A network can show 100 provers and still have weak privacy availability if most proof capacity sits in one region. Europe, Asia and North America do not matter as labels; what matters is how many proofs each region can actually produce when demand spikes.

That changes how I look at DUSK. Prover count is activity. Regional proof capacity is resilience.

ZK proof generation is single-threaded per worker, so strong single-core performance matters and more workers raise parallel capacity. Fine. But that makes hardware concentration more important, not less.

If 70% of proof throughput comes from 10 machines in one data center, 100 provers is a comforting number with very little operational meaning. A regional outage could remove 30% of proving capacity while barely touching stake.

And latency alone can mislead. A 500 ms proof consumes 5% of a 10-second workflow 2 seconds consumes 20%. I want P95, P99 and proofs per CPU core per minute under load.

DUSK may have fast proving. I’m still watching whether that speed is geographically survivable.

#dusk $DUSK @Dusk