Ideas on a proving network

[Anon]

@spalladino

Why can’t we just post proofs to a worker queue with a bounty?

1. A sequencer who must produce a block within a certain window (as defined by the SSP) will not rely on one prover to do this, as they risk revenue. A rational sequencer will either ask many provers to compute the block, leading to redundant compute and extra costs for users, or run their own compute.
2. A rational sequencer will only pay the fastest prover, which means the second fastest prover does not get paid. Over time this may lead to centralisation with one giant prover. This creates bribe attacks, present, unless the protocol enforces all provers get paid.

I think we can (PBPS) (BPA)

1. Sequencers will select bonded provers, where the bond > stake + MEV.
2. With a fixed MAX_PROVER_TIME, the cheapest prover is preferred. If network latency overhead is proportionally small, decentralized networks connecting the cheapest compute will dominate.