OpenAI dumped some motherload of math discoveries and apparently all the math people are flabbergasted. But what I'm wondering is how likely is it that there is some AI facilitated math discovery in the next few years that makes it so a GPU cluster can crack bitcoin keys in something like a month or week? This would be pretty awful.
I get that this is a sort of annoying hypothetical. It could be asked for all cryptography currently considered secure. And so maybe the answer is: no way to know. All the same, do you think about this in the little hours of the night?
* checks crystal ball
No way to know. But what is your plan for the occasion that it does happen?
that is probably the question I should have asked.
If it is the case that the cracking relies on a known public key, I would build a business selling private transaction broadcasting.
Then every visible or reused address is doomed, i.e. at least half of the CEX cold storage, every p2pk utxo, and I'd estimate about 90% of stackers' wallets.
So... maybe you can do something today.
That business already exists and afaik Peter just got crowned king of it.
I get reused addresses being doomed, but a pay to script hash or p2wsh? wouldn't these be safe until in the mempool?
he might need competition?
if they're reused yes. if they're not reused, then you need to set something up in advance, otherwise it'll depend on ecdsa and is essentially the same as a single use p2pkh/p2wpkh. Personally I don't think I'll be rich enough to transact within the barrage of wallets getting swept tho.
Also already there. So as appealing as it sounds to get rich off of other people's misery by rentseeking, I wonder, if you're not established by now, how are you going to be able to compete yourself? And if you already have this product, how are you not going to get disrupted by, say, foundry offering it cheaper and with a delivery guarantee?
Clearly I need to think more deeply about this.
What is your plan?
(in the event that something like this happens)
The moment I'd see something that looks really bad, I'll reduce liquidity on LN to a bare minimum. If there is something feasible in a hybrid p2wsh, I'd dump it to there. Otherwise, I'd just let it sit in p2wpkh and see what's what. I haven't reused addresses since... 2013.
Not super concerned about old cryptographic algorithms like ECDSA, but I am optimistic that if machine learning can be used to battle test cryptography, then it will accelerate how quickly we can be confident in newer algorithms based on lattices or isogenies.
I don’t I sleep like a baby
Thanks for sharing