Anyone who has done account reconciliation knows the scene: at year-end, the auditors come in, sample a few hundred transactions, check them, and sign off. The remaining hundreds of thousands go unseen. This is called sampling, an old method that financial audits have relied on for decades.
I’m interested in Space and Time because it focuses on exactly this old problem. Banks often keep highly granular order books in off-chain SQL databases, then periodically update summary results to a shared ledger. No one can use cryptography to prove whether the numbers on the ledger are correct, because the step before summarization happens inside a black box. Anyone who really wants to manipulate P&L changes things at that step.
The idea behind Proof of SQL is that every query comes with a proof alongside its results, showing that the SQL query was indeed run on data that had not been altered. The verifier checks only the proof; they don’t need to rerun the query or see the underlying records. The enterprise version can even be deployed on a company’s own premises, so auditors can examine the proof without seeing customer details.
What can this replace? I think, first and foremost, sampling. Going from checking a few hundred transactions to having evidence accompany every result is a tangible upgrade. Second, it can reduce the back-and-forth caused by mutual distrust when institutions reconcile accounts across organizations.
What can’t it replace? Honestly, quite a lot. A proof can only guarantee that “the calculation was correct on this dataset”; it can’t guarantee that the dataset itself is genuine. If a fake invoice gets into the table, it will still be faithfully included in the summary. We used to have a saying: garbage in, garbage out. Whether the source is genuine still depends on controls, signatures, and people. As for how many institutions are willing to move their core accounting onto it, I haven’t seen verifiable data, so I won’t make claims on its behalf.
So this is how I’d characterize it: it doesn’t replace accountants; it turns the most exhausting part of their work—arithmetic verification—into evidence that can be recorded and reviewed.
@SpaceandTime is about evidence, not verdicts. $SXT needs someone willing to pay for that evidence before it can become a real need. #SXT
I’m interested in Space and Time because it focuses on exactly this old problem. Banks often keep highly granular order books in off-chain SQL databases, then periodically update summary results to a shared ledger. No one can use cryptography to prove whether the numbers on the ledger are correct, because the step before summarization happens inside a black box. Anyone who really wants to manipulate P&L changes things at that step.
The idea behind Proof of SQL is that every query comes with a proof alongside its results, showing that the SQL query was indeed run on data that had not been altered. The verifier checks only the proof; they don’t need to rerun the query or see the underlying records. The enterprise version can even be deployed on a company’s own premises, so auditors can examine the proof without seeing customer details.
What can this replace? I think, first and foremost, sampling. Going from checking a few hundred transactions to having evidence accompany every result is a tangible upgrade. Second, it can reduce the back-and-forth caused by mutual distrust when institutions reconcile accounts across organizations.
What can’t it replace? Honestly, quite a lot. A proof can only guarantee that “the calculation was correct on this dataset”; it can’t guarantee that the dataset itself is genuine. If a fake invoice gets into the table, it will still be faithfully included in the summary. We used to have a saying: garbage in, garbage out. Whether the source is genuine still depends on controls, signatures, and people. As for how many institutions are willing to move their core accounting onto it, I haven’t seen verifiable data, so I won’t make claims on its behalf.
So this is how I’d characterize it: it doesn’t replace accountants; it turns the most exhausting part of their work—arithmetic verification—into evidence that can be recorded and reviewed.
@SpaceandTime is about evidence, not verdicts. $SXT needs someone willing to pay for that evidence before it can become a real need. #SXT
