Sjors

https://github.com/sjors/
Book: https://www.btcwip.com
Podcast: https://bitcoinexplainedpodcast.com/@nado/episodes

Public Key

npub1v25jqj332n630klwtqeqxgyrjl4alw4phmluhu2x6lcxzwy9m03suqexw3

NIP-05 Address

Profile Code

nprofile1qqsx92fqfgc4faghm0h9svsryzpe067lh2smal7t79rd0urp8zzahccprpmhxue69uhkummnw3ezuumswfhhvmm0wd6zumnvqyt8wumn8ghj7un9d3shjtnswf5k6ctv9ehx2aq0nq8ua

Publishing to

nostr.sprovoost.nl

relay.primal.net

relay.damus.io

nostr.bitcoiner.social

nostr.oxtr.dev

nostr.land

Author Public Key

npub1v25jqj332n630klwtqeqxgyrjl4alw4phmluhu2x6lcxzwy9m03suqexw3

Show more details

Last Notes

2026-08-27 09:27:58 GMT

- reply

Also, how much do you estimate Aaron and I charge for bribes? Do you have evidence other podcasters were offered a bribe?

2026-08-27 09:12:41 GMT

- reply

> I would believe you that Rodolfo didn't buy you out
With what money? You know how clawbacks work in bankruptcy?

2026-08-27 08:10:22 GMT

Going offline vs going --offline is quite different.
#nevent1q…x6zu

2026-08-27 07:48:36 GMT

- reply

> details that hint
We're not a True Crime podcast. Backdoors that look like accidents are technically interesting, but I find it too implausible to even bring up.
Nobody denies that Coinkite (employees, contractors, CTO) wrote the code. The use of pseudonyms didn't cover anything up.
Now, if this is ever proven in court, or crosses some other evidence bar that makes me take it seriously, I'm happy to do another episode.
#nevent1q…3h47

2026-08-27 06:30:41 GMT

- reply

So the key is to run it in --offline mode, not turn it off entirely.
#nevent1q…u3qv

2026-08-26 18:56:39 GMT

How long can you keep a lightning long offline, if there's no HTLC in flight, before someone can screw you over? Asking for a friend who doesn't like binaries.

2026-08-26 12:17:01 GMT

Finally, @npub1art…m0w5 and I got a chance to dive deeper into the ColdCard hack. Also a little meta discussion on sponsorship at the end.
https://bitcoinexplainedpodcast.com/@nado/episodes/bitcoin-explained-101-the-coldcard-hack

2026-08-26 11:58:21 GMT

- reply

No, though Microsoft / Github sometimes do fight / ignore such takedowns. I'm annoyed by Musk trying, not worried about the actual code being in danger.

2026-08-26 07:44:30 GMT

> and the project's repository.
Ah yes, Musks unending love for free speech.
#nevent1q…u0md

2026-08-25 11:59:53 GMT

Are there any serious use cases for multipath descriptors with more than two items (<0;1> for receive and change)?
Can be in future fancy softforks too.

2026-08-24 15:21:28 GMT

- reply

Technically the client needs to pay this (as they also lost the case).

2026-08-24 15:15:08 GMT

Lawyer gets €2300 "fine" for hallucinating jurisprudence, because it wasted court time and billable hours for the opponent. It was not necessary to prove they used AI. Makes sense, fix your skill file to check this.
https://www.at5.nl/artikelen/239607/advocaat-gebruikt-nepcitaat-mogelijk-door-ai-en-moet-ruim-2300-euro-betalen

2026-08-23 17:59:15 GMT

- reply

I was thinking when they run out of money and the bond market collapses :-)

2026-08-23 14:59:00 GMT

- reply

I'm just riding out the subsidy wave for the big US models, it's not going to last forever probably.

2026-08-22 07:48:39 GMT

- reply

It's pretty good, but you still need to range-diff and otherwise inspect the result. It mainly saves me from tedious header and function argument conflicts.

2026-08-21 15:28:20 GMT

- reply

It's already pretty good at rebases by default, but this might help: https://github.com/Sjors/skills/blob/master/rebase/SKILL.md

2026-08-20 19:10:52 GMT

I'm glad they fixed this...
https://blossom.primal.net/3dbdf60407ceef7292a31f13ea4d700fa796c25682c2ff051be0c06f7750fd06.jpg

2026-08-18 10:09:20 GMT

- reply

@npub1zkh…qqcm thanks for the test, it worked.

2026-08-18 08:09:16 GMT

- reply

Thanks! I switched to a minibits mint, maybe that helps. Still waiting for bolt12 zaps, so I can just receive them sovereignly (without installing stuff with access to my CLN node).

2026-08-18 07:57:53 GMT

I know the US has a debt problem, but Congress mining crypto in my browser for several minutes to „verify I'm not a bot“ seems a bit desperate?
https://blossom.primal.net/d18e55e84e6938eb1677ddc71aa412d9b7d324f26c5a828b833e9d3915ecfab8.jpg

2026-08-17 21:15:16 GMT

As people working on imaginary internet money, calling out people trying unrealistic things with said internet money, it's good to do an occasional sanity check.
https://youtu.be/KnDuWfzKx5o

2026-08-17 20:43:31 GMT

- reply

It's a bad proposal that he should have abandoned after people told him a million times. Instead he shifts the blame to Bitcoin Core. I'm not upset about him starting an altcoin. I think it's stupid, people should ignore it and not lose money.

2026-08-17 20:40:13 GMT

- reply

Mmm, might be a profile relay issue. Try now?

2026-08-17 19:11:44 GMT

- reply

I deleted my account there years ago. Most web shops will have an ad hoc option to use creditcard through paypal without having an account. If not, I might remind them that Bitcoin exists. And otherwise unless it's life saving medicine, I'll just move on.

2026-08-17 19:09:36 GMT

- reply

Fascinating. I didn't (have my agent) throughly check every claim and screenshot in the blog post, but it looks fairly accurate to me. Denial in action?

2026-08-17 19:07:20 GMT

- reply

That last AI video is impressive.

2026-08-17 18:50:16 GMT

- reply

Do you know how often Paul has bashed Bitcoin Core for the past decade? Free markets require good information, which includes calling out bullshit forks.

2026-08-17 17:47:02 GMT

Oh by the way, how's the Sztorc-fork going?

2026-08-17 17:39:06 GMT

- reply

If BIP110 proponents weren't harassing the sane and productive people, there wouldn't be any temper tantrums.
https://blossom.primal.net/23977962f12201545019abf079b00896064fa60f254388548cea6146fdd72a63.jpg
#nevent1q…mgqz

2026-08-17 15:55:38 GMT

How does a two year old get a drivers license? She also totally doesn't care that every word she says on camera can further incriminate her boyfriend.
https://youtu.be/zOpndE9NjYo

2026-08-14 18:26:06 GMT

- reply

Speaking of France, this is good though!
https://www.reuters.com/world/frances-top-court-rules-social-media-ban-curtails-freedom-expression-2026-08-14/
#nevent1q…89jn

2026-08-14 15:53:45 GMT

- reply

Only French people

2026-08-14 10:27:29 GMT

- reply

> Among individuals, 26,805 would have a reference tax income of €100,000 or more, 386 would exceed €1 million and 8 would exceed €10 million. According to the cybercriminal, the file would also be offered for sale for several thousand euros.

2026-08-14 10:23:22 GMT

And the incompetent French tax authorities were hacked again. Almost 700k future $5 wrench victims.
https://frenchbreaches.com/alertes/direction-g-n-rale-des-finances-publiques-dgfip-msq5rgy4cql9o0l7cus
cc @npub18ul…dcl9

2026-08-13 16:26:48 GMT

This is the only way forward. Or maybe collab with darknets for drop zones.
https://x.com/Trezor/status/2087885428313543059
https://blossom.primal.net/07e3ef142a345e8ef8228a7828dcb120cdfe5bde84b59a81996bd676ad313b05.jpg

2026-08-13 12:31:06 GMT

- reply

The most fascinating question is always, do cult leaders actually believe the things they say?
#nevent1q…gr3g

2026-08-13 12:15:02 GMT

- reply

Sounds more like a wealth tax to me :-)

2026-08-12 08:53:03 GMT

- reply

The bip-110 community had years to come up with a good design and get technical consensus on it (among themselves). But they were too busy fooling themselves that a majority hash power would go along.
Even now there should be no rush to replace the algorithm. Sure, their 3 block history could get wiped out any time, but that same risk exists with any new PoW. At least Bitcoin miners don't have an economic incentive to wipe out the BIP110 fork blocks. A fully functional low PoW altcoin is a more juicy target.

2026-08-12 08:30:42 GMT

- reply

No idea, but presumably someone will keep mining with the original proof-of-work, so that'll be bip-110 coin classic?
#nevent1q…plyh

2026-08-11 18:47:12 GMT

- reply

This was a great one.
https://fountain.fm/episode/Adb0vXrt93c7dHYTDRu6

2026-08-11 07:55:59 GMT

- reply

And then China would blame the US for a lab leak.

2026-08-11 06:59:29 GMT

Thanks Claude, because the EU doesn't have enough mass surveillance tools already.
https://support.claude.com/en/articles/16266773-how-claude-marks-ai-generated-content

2026-08-11 06:32:37 GMT

- reply

I assume they mean BIP-110 coin?

2026-08-09 19:27:47 GMT

- reply

And thankfully my place was designed by a sane architect, so I can wash outside windows easily.

2026-08-09 19:26:45 GMT

It sounds like more work than cleaning windows myself. Plus, I'd rather raise my heart rate through cardio than by being exposed to this noise.
Now if it was quiet, could climb between windows, and from inside to outside, and could rinse its own cloth...
https://youtu.be/SPV-hstjRGM

2026-08-09 15:29:40 GMT

- reply

It's ze Germans! https://nos.nl/artikel/2626122-zeppelin-vol-meetapparatuur-boven-nederland-onderzoekt-luchtkwaliteit

2026-08-09 07:31:35 GMT

- reply

So you get your lightning payout in 2031 or so?

2026-08-09 07:26:19 GMT

- reply

It can be done with Stratum v2 as well, as long as the pool runs one BIP110 compatible node.

2026-08-08 20:41:06 GMT

Oceans apart
(works as designed, since miners can make their own template)
https://image.nostr.build/e81d9d4403b2cb0118f06bbf285691e97697ed727cd58df60f061be553970aad.jpg
#nevent1q…95m3

2026-08-08 20:25:21 GMT

- reply

Core back on top.
https://image.nostr.build/83998cf183e918d392f4c1a88d4033c6c83c98be5bd2bd4076e07b8b750123d9.jpg

2026-08-08 20:21:39 GMT

First-seen rule will be interesting now...
https://image.nostr.build/7fb75e26b69bc72dea516d7b03072dc34f952f9bdb8e1d6a0044a480147aa2d7.jpg
https://image.nostr.build/8a9ca6171ed1f474c72f1f707de3653d69d2205ec14e8285b897364f1af63cef.jpg
#nevent1q…aufn

2026-08-08 20:19:19 GMT

- reply

Yes, at which point the chains would just converge and we get another split.

2026-08-08 20:06:55 GMT

- reply

To be fair, Bitcoin Core also doesn't hold on to headers that it considers invalid for any reason. With some exceptions.
https://blossom.primal.net/ab4d9461ec810372eeb7f8c4a426eb753a7af87d6dcd40ef04c33a6837db311a.jpg

2026-08-08 20:01:13 GMT

- reply

Annoyingly, Knots rejects non-signaling headers without even adding them to the block index. So they don't show up in getchaintips, which is what ForkMonitor uses to decide if there's a fork.
https://blossom.primal.net/332859327ef6a639edb8a09a50dd0d09a46c144e02aa06e0d563dbf724e04776.jpg

2026-08-08 19:53:02 GMT

Happy K-Day!
https://forkmonitor.info/
https://blossom.primal.net/d129b53f00c3a86baf1002cfe17ebdd76d5615dc499cd2ca8f42f0e452f65031.jpg

2026-08-07 15:55:49 GMT

- reply

Dammit, I was about to get dinner.

2026-08-06 14:17:48 GMT

https://img.sprovoost.nl/3169a0ed4c9d1021309c6d5bb5fd2f6846c2616d729117d0309164e4a70be1ab.webp

2026-08-06 06:49:43 GMT

- reply

Nostr is not known for a focus on code quality. I think he's confusing a lack of attackers with strength.
Specifically for nsec entropy it's somewhat easy to so this correctly on a modern phone. And relays are redundant and no funds are lost if they go down due to some DoS, sure. And having your wallet drained to an NWC bug, might be fine if there's not much in it. But the latter is another example of low stakes.

2026-08-06 06:37:12 GMT

- reply

Not sure about the US, but in many countries the government can prosecute a crime even if nobody presses charges. So indeed they'd have to reframe their action as some sort of responsible disclosure ("we took it because someone else would have"). Offering to keep half wouldn't be a good look.
I guess they could frame the lack of early public announcement as an opsec thing.

2026-08-05 18:49:07 GMT

- reply

Voice memo?
Lopp: Hey Siri, note .... ...
Neighbor: shut the f*** up!

2026-08-05 14:51:29 GMT

- reply

It's sad that this is probably the first time millions of people see this device.
https://image.nostr.build/51138a0424c53a5bb5eead3a21df33583f1256ee601ba688d483ab6d90899965.jpg

2026-08-05 14:48:39 GMT

Took less than a week for the mainstream media to pick this up: https://nos.nl/artikel/2625622-aanval-op-apparaat-dat-bitcoins-beveiligt-voor-tientallen-miljoenen-euro-s-gestolen #ColdCard

2026-08-05 08:19:14 GMT

- reply

By the way, I'm talking purely about capability, of doing this as a long term rug pull plan, not about motive.
My reasoning above doesn't rule out opportunistic theft by one of the founders. Like I said, any ration law enforcement would investigate them too.
I do think that scenario is unlikely, but not conspiracy level unlikely.

2026-08-05 08:05:05 GMT

- reply

Well you've seen Keonne and Bill :-)

2026-08-05 08:04:08 GMT

- reply

Many entrepreneurs have such a personality though.

2026-08-05 07:43:02 GMT

- reply

switck == doc-hex wasn't a secret in the first place? The ngu library was a ColdCard project, so obviously written by their developer. Who cares what nicknames they use?
What's a conspiracy is the allegation that this was a five year prepared plan.

2026-08-05 07:40:14 GMT

- reply

And they would not have used an API to gather information. Plenty of time to set up such infrastructure locally.
You can't have someone sophisticated enough to hide underhanded C in a sloppy commit for give years, have that much patience, yet not rug everyone at once.
The game theory just doesn't work: anyone else could have found the bug in the mean time, and their sweep would come as a total surprise to the original exploiter.
A well prepared rug pull by the founders would have had to continuously maintain multiple alternative sweep transactions, in order to fee bump whatever a surprise third party attacker was doing.
I would also expect an attacker with five years of preparation to have immediately coinjoined, spread the loot across utxos and quickly exit some of it.
They would have keep to these logistics secret too for years, yet spend time continuously maintaining it, not accidentally revealing expertise in this, etc.
That's why I consider this a classic conspiracy theory. People look for supporting evidence and clues, assume an all powerful "they", and then either ignore practical logistics like I describe above, or come up with another complicated explanation how that could be avoided.
Meanwhile the simpler explanations remain: some 15 year old in his bedroom with an LLM, someone with a bit more expierence in hacking, or even the North Korea crew that wants a break from hacking shitcoin smart contracts. All of them would simply take the money as quickly as they can. The latter two would realize that people will notice the attack, so they have to make a big sweep first. And that someone else might find the bug too, or even that their LLM provider was eavesdropping and might rat them out, so they can't prepare a total sweep.

2026-08-04 19:26:13 GMT

- reply

Memes are fun, but you do realize that unless the FBI manages to quickly track down the hacker, they're going to investigate the (co)founders anyway? And in parallel, not later, they'll investigate every competent wallet developer and security researcher.
So priority should probably be to make sure they find the real thief, fast.

2026-08-04 17:50:20 GMT

- reply

Please publish audio podcast afterwards, I haven't seen a new The Rage episode in my feed for a while...

2026-08-04 16:54:33 GMT

Not quite the usual $5 wrench attack, but still concerning. Google translation:
"Man urinated on during kidnapping and extortion in Nieuw-Vennep following crypto dispute"
A man was kidnapped, extorted, and "severely humiliated" in Nieuw-Vennep in May, it emerged on Tuesday during an initial public hearing. One of the three suspects appeared before the judge. Among other things, the victim had urine urinated on.
The Public Prosecution Service (OM) described the acts on Tuesday as "very frightening and also humiliating." The victim was allegedly taken to a forest in Nieuw-Vennep on May 1 by Youssef B., among others. There, he was allegedly held in a car to extort 60,000 euros from him.
According to the OM, B. committed a "violent, intimidating form of vigilantism." The victim was allegedly forced to undress. The file contains screenshots of a video showing urine being urinated on the naked body and into the victim's mouth.
There were also alleged threats involving a firearm. According to the victim's statement, he was also told that his eyes would be gouged out and that his fingers would be cut off. Additionally, the suspects allegedly threatened to post the video of the urination online.
The incident was reportedly preceded by an argument regarding crypto investments. B. allegedly invested money in cryptocurrency with the victim. "It was a kind of pyramid scheme that he fell for," says his lawyer.
B. had a 'close friendship' with the victim
The 23-year-old B. denies that there were threats involving a firearm. According to his lawyer, there is no evidence of this. During the hearing, B. read a short letter to the victim in which he expressed remorse.
He described the victim as someone he had known for years and with whom he had "a close friendship." "I took the law into my own hands; I should never have done that. What I have done cannot be justified."
The court ruled that B. will be released on conditions starting Wednesday. His brother is also a suspect in the case and remains in custody. The court had already released the third suspect earlier.
https://www.nu.nl/binnenland/6405326/man-in-mond-geplast-bij-ontvoering-en-afpersing-in-nieuw-vennep-na-cryptoruzie.html

2026-08-04 16:02:57 GMT

- reply

Consensus by X, wonderful.

2026-08-04 07:41:19 GMT

- reply

I did a splice-in with CLN recently, with funds from an external wallet. Needed chatgpt to help me figure out the commands, but it worked!

2026-08-03 13:28:29 GMT

- reply

Alternative site: https://nitter.poast.org/FarsideInsights/status/2084211693266067514

2026-08-03 13:25:59 GMT

- reply

Partial screenshot since social media preview doesn't seem to work.
https://image.nostr.build/198e021ae0814830fef6c89960da2830bb0cbbe808c733c8ca7fbe758245fa31.jpg

2026-08-03 13:24:23 GMT

End of an era: https://xcancel.com/i/article/2084211693266067514

2026-08-03 11:07:23 GMT

- reply

You can check each individual source against its required minimum entropy. That seems better than not using a source entirely just because it's not enough on its own.

2026-08-03 11:01:51 GMT

- reply

Re (1): perhaps devices should collect 128 bits of entropy independent of their secure enclave. That can be from multiple sources, but they have to add up. Like in the old PGP days, it could require the user to do something random. A cheap accelerometer plus the instruction to shake, should be intuitive and quick enough?

2026-08-03 07:32:08 GMT

And the phishing begins...
https://image.nostr.build/062c7d56c0074bbf6f4f46acb046d751a96045a7bc8fd3f526a1039b9e8f3ed8.jpg
https://image.nostr.build/bdda01709f908e87e33c647cbb22d5041138f34cd978a5b2dd20ca2ba0655f8c.jpg
https://image.nostr.build/0f286d5213ccfc46a6b158ee939c3d870d700943aad35d4c080def2fd2a7b306.jpg

2026-08-02 19:56:51 GMT

- reply

It's hard to reason about such probabilities. From what I can find, the odds of dying from the war in Kyiv, as a non-combatant, is about 10x that of a deadly car accident. This increases dramatically as you go east. And even more when you're in range of fiber optic drones that hunt for civilians near the frontlines. And for combatants it's obviously a lot higher.
https://ukraine.ohchr.org/en/reports/protection-of-civilians
https://en.wikipedia.org/wiki/List_of_countries_by_traffic-related_death_rate

2026-08-02 17:16:08 GMT

- reply

I also didn't check what Trezor did, back in their pre-secure-enclave days.

2026-08-02 16:59:54 GMT

- reply

I suppose there's still a way for people to prove ownership, since the device id is used in generating the seed. I would not want to be Coinkite's Mandrik though.
#nevent1q…z9yu

2026-08-02 16:48:39 GMT

- reply

For us non-xers, link to tweet?

2026-08-02 16:12:02 GMT

- reply

How defensive do you want it?
#nevent1q…t5hc

2026-08-02 16:09:52 GMT

- reply

* I don't think

2026-08-02 16:09:41 GMT

- reply

I don't the un-FOSS-ness was a critical factor, because Google Zero and Ledger will happily look at it for the opportunity to find a bug. That said, it's worth considering whether FOSS needs to be a minimum security requirement for critical code paths.
But remember that the secure enclaves are proprietary black boxes. So you'd have to treat them as unsafe and somehow get enough entropy to be cryptographically secure without it. That's a bar no hardware wallet today meets, afaik.

2026-08-02 13:14:04 GMT

- reply

I don't think there is any, just full seed collisions. Unless the software pseudo-RNG is so bad that a few million device runs would reveal a lack of uniformity. Perhaps the lesson is that test code random number generators should yield sequential numbers, that are easier to guard against.

2026-08-02 11:49:26 GMT

- reply

Entropy is hard to test. You'd have to have to run it on (slow) real hardware billions of times to find a 32 bit collision. In a simulator you don't expect to use the real TRNG.
Checking library symbols is not a typical thing for an integration test, though it might have been good (in hindsight) to run a whitelist of symbols you want to see. Bitcoin Core does some of that:
https://github.com/bitcoin/bitcoin/blob/master/contrib/guix/symbol-check.py

2026-08-02 11:33:12 GMT

- reply

I also agree that writing your own cryptographic library merely for licensing reasons is a big risk. Since nobody else will use it, review is limited to security researchers, but that's still a big audience and they didn't find the issue. A random competitor that slavishly copies your firmware is not going to help you find the bug.
But it's mainly a wrapper around well respected libraries like libsecp and libwally. And it's quite common for libraries to make you bring your own entropy, since that depends on the target environment.

2026-08-02 11:25:47 GMT

- reply

Link
#nevent1q…g54y

2026-08-02 11:24:41 GMT

- reply

I just read that article. Assuming there's no documentation *elsewhere* then I agree that the process was bad. And I would have noticed the huge commits and brought that up.
This is a case where you need multiple developers in a team to keep each other honest, and say: hey, you're touching critical (build system) code that's over your head. We need to hire someone external to help with this library migration.

2026-08-02 09:33:34 GMT

- reply

Would be nice to keep the relay running for at least paid subscribers to post to it and read-only for everyone else. Bonus points to allow anyone followed by subscribers to post to it.

2026-08-02 09:27:57 GMT

- reply

Maybe, if the audit was specifically about seed generation. I've worked (mostly as a reviewer) on this exact topic for two wallets:
1. the blockchain .info web wallet (after the Android incident, but that was a different codebase)
2. the Bitcoin Core wallet (when openssl was replaced)
In both cases, part of the strategy was to xor the kitchen sink. So had I reviewed the code, I probably would have suggested to xor more. That would have reduced the impact. But unlike browsers and desktop operating systems, hardware wallets don't have a lot "more".
I don't think I would have caught the main issue. Maybe, because I've played with esp32 before, with the help of an LLM* and a few days of time, I would have fully understood the build pipeline, and maybe realized the problem. Just maybe.
Let's also not forget that hardware wallet competitors spend a lot of time and energy "hacking" each other to show off at conferences. They probably focussed more on seed extraction than generation, and therefore missed it. Security researchers also, even without a bounty, had an incentive to find and responsibly disclose this, for prestige alone. Yet they didn't.
* = that's not actively hostile to security research

2026-08-02 08:11:54 GMT

Very detailed write-up by @npub1c2d…kxar and the Liana team. Useful in general if you want to understand the impact on multisig setups. Of particular note:
1. having a hard to predict descriptor, doesn't always protects your coins (especially with SegWit)
2. in-flight attacks on the mempool haven't happened yet, but could be a matter of time
https://wizardsardine.com/blog/coldcard-rng-vulnerability/

2026-08-01 15:09:01 GMT

- reply

Between just Ledger and ColdCard, you'd have to forgive Ledger for:
- the backup service
- a plethora of shitcoins that potentially require more complexity in the base layer, even if you only install the bitcoin app
- having ~one engineer still working on that app, so just hope it continues to be maintained
- terrible privacy by default (xpub plus marketing analytics)
- database breach(es) that probably got people killed
That said, when used correctly, i.e. with HWI or some other custom software, it's great. And they supported MuSig2 long before ColdCard (the rest of the hardware wallets don't at all, last time I checked).
And hopefully it doesn't also have an entropy bug, though I didn't check :-)

2026-08-01 15:00:51 GMT

- reply

What's the impact on MuSig2 if *one* of two signers has bad entropy for their nonce? Assuming the signers are not adversaries.

2026-08-01 05:56:19 GMT

- reply

I briefly tried yesterday and it didn't find it without cheating. It did find some of the other entropy issues mentioned in the write up. But I didn't try very hard.
#nevent1q…mkca

2026-07-31 13:06:29 GMT

- reply

Anthropic ran Mythos themselves against all sorts of projects, but I guess they haven't published a full list. Would be useful to know.
I briefly tried to quiz Sol 5.6 (high) on master before the fixes. It's tricky doing this after the fact. The first time it pulled the remote, saw the fix and obviously "found" the vulnerability.
I then tried again while insisting it doesn't look at the internet. Then it didn't find it.
In general, prompts like "is this code secure" won't give useful answers. The attacker probably had a better hunch of what to look for.

2026-07-31 12:30:14 GMT

- reply

You need to store the seed somewhere though. Going from a secure enclave to storing it on paper and typing on use, comes with its own trade-offs. For that scenario, booting an OS in RAM would prevent leaving anything behind.

2026-07-31 12:27:14 GMT

- reply

#nevent1q…pf8d

2026-07-31 12:10:05 GMT

- reply

One problem with offline laptops is that you can't update the OS. Maybe that's fine if you rely fully on disk encryption.