Fair, you gated "not extreme" and I argued against an extreme die anyway. I was going off a similar discussion I read a long time ago, which is where the extreme framing came from, not from what you actually wrote.
Still, I don't think "nobody knows the bias" is the part keeping you safe. "Not extreme" is. Obscurity buys less than it looks like, and the search isn't targeted at you personally, it runs against every funded address at once.
Which leaves only: how extreme is it? That's my actual objection to handmade dice. Not that they're definitely bad, but that you have no way to check before the key exists.
I don't want to state numbers I can't derive myself, so, AI says:
"Not knowing the bias isn't a barrier, it's a cost. An attacker searches over the space of plausible d6 distributions rather than guessing one, and the penalty for that is roughly (k-1)/2 * log2(n) bits, about 16 at 100 rolls, plus a handful for unknown roll count and mapping. Call it 30 bits of obscurity. Each candidate is one pubkey derivation checked against a filter of funded addresses, so no dictionary or dedicated tool is needed and the work amortizes across everyone who used the same method."
dtonon / daniele
npub100…8vwqk
2026-08-07 14:34:52 UTC
in reply to nevent1q…rq4g
Author Public Key
npub10000003zmk89narqpczy4ff6rnuht2wu05na7kpnh3mak7z2tqzsv8vwqkPublished at
2026-08-07 14:34:52 UTCEvent JSON
{
"id": "9d96fd1475e6dcb3e066f68c86dea0a16bd8d4e6d18e53f490aa715810a827eb",
"pubkey": "7bdef7be22dd8e59f4600e044aa53a1cf975a9dc7d27df5833bc77db784a5805",
"created_at": 1786113292,
"kind": 1,
"tags": [
[
"e",
"a81a5a9aed20b6f026554872b98217461d275db6277b03d250da75ec5261e60c",
"wss://nostr.mom/",
"root",
"bb2224743f53e1e334d8c2c4fdb77c6a158b9f0f6885b4d6c715287429dfd6bd"
],
[
"e",
"0000433eb9ee0f023f4b75d8d87cfdda819174ef377876a7ffdc0a06a4734982",
"wss://relay.primal.net/",
"reply",
"5ea4648045bb1ff222655ddd36e6dceddc43590c26090c486bef38ef450da5bd"
],
[
"p",
"7bdef7be22dd8e59f4600e044aa53a1cf975a9dc7d27df5833bc77db784a5805"
],
[
"p",
"5ea4648045bb1ff222655ddd36e6dceddc43590c26090c486bef38ef450da5bd"
]
],
"content": "Fair, you gated \"not extreme\" and I argued against an extreme die anyway. I was going off a similar discussion I read a long time ago, which is where the extreme framing came from, not from what you actually wrote.\n\nStill, I don't think \"nobody knows the bias\" is the part keeping you safe. \"Not extreme\" is. Obscurity buys less than it looks like, and the search isn't targeted at you personally, it runs against every funded address at once.\n\nWhich leaves only: how extreme is it? That's my actual objection to handmade dice. Not that they're definitely bad, but that you have no way to check before the key exists.\n\nI don't want to state numbers I can't derive myself, so, AI says:\n\n\"Not knowing the bias isn't a barrier, it's a cost. An attacker searches over the space of plausible d6 distributions rather than guessing one, and the penalty for that is roughly (k-1)/2 * log2(n) bits, about 16 at 100 rolls, plus a handful for unknown roll count and mapping. Call it 30 bits of obscurity. Each candidate is one pubkey derivation checked against a filter of funded addresses, so no dictionary or dedicated tool is needed and the work amortizes across everyone who used the same method.\"",
"sig": "a5a918798670f02afe200d024b2a3277cf2c1e35bc20154c8ab7f5158229b7a3b474a2bbb7499ca65c250617040dbd3e2e67d9ca0464b7776c3bc30e3ffcbc87"
}