Let's Review the AO Whitepaper's Characterization of ICP and AO's security model

I think you are misguided here. Number of replication is NOT the whole story.

Suppose 10/13 is a good threshold, and there are indeed 13 unique and independent CUs. Now 10 of them have computed output message B from input message A. Can you now trust message B and take your own actions based on B (e.g. buy or sell on another market)?

No, you cannot because it is unsafe. You don’t know whether A was computed correctly in the first place. These 10 CUs are not responsible for verifying A’s correctness either, and their stake won’t be slashed even if A turns out to be wrong, because they computed from A to B correctly.

The only sensible decision in a opportunistic setting is to wait until message B finalizes, which usually would then imply A has also finalized. I’ll just quote myself again:

3 Likes