The COLDCARD Hack and Future of Self-Custody

1
00:00:00,000 --> 00:00:06,000
And we're live. Welcome to another edition of the PBJ podcast.
2
00:00:06,280 --> 00:00:06,660
Hey guys.
3
00:00:06,820 --> 00:00:07,260
It's Friday.
4
00:00:07,560 --> 00:00:08,760
Good to have you back in the studio.
5
00:00:08,760 --> 00:00:10,440
I gotta come back. It's a special episode.
6
00:00:10,800 --> 00:00:13,780
I was like, I gotta get back in the studio for our special episode.
7
00:00:15,160 --> 00:00:16,520
Yeah, this is gonna be...
8
00:00:16,520 --> 00:00:17,160
Gonna be a big one.
9
00:00:17,320 --> 00:00:21,660
Yeah. Long day. I know Steve's got a few things here, so...
10
00:00:21,660 --> 00:00:22,840
Steve has been working hard.
11
00:00:22,940 --> 00:00:24,340
A lot. The last 24 hours.
12
00:00:24,560 --> 00:00:26,740
Max and I have been... What have we been doing?
13
00:00:27,500 --> 00:00:29,880
Well, actually, trying to survive the heat,
14
00:00:30,700 --> 00:00:33,400
I've been downing my electrolytes nonstop, actually.
15
00:00:33,560 --> 00:00:36,320
That's its own story, but that's been my number one.
16
00:00:36,640 --> 00:00:38,020
But the heat is not that hot here, is it?
17
00:00:39,220 --> 00:00:40,760
Sadly, I've become more and more effectively.
18
00:00:40,760 --> 00:00:41,720
Oh, you're sensitive to it now.
19
00:00:41,800 --> 00:00:44,120
Yeah, but anyways, more to come.
20
00:00:44,420 --> 00:00:46,740
But where do we jump in?
21
00:00:46,740 --> 00:00:48,780
I think we're going to spend the whole show basically
22
00:00:48,780 --> 00:00:52,300
talking about the Bitcoin security incident cold card.
23
00:00:52,300 --> 00:00:54,600
I feel like
24
00:00:54,600 --> 00:00:57,360
everyone in the community has been on
25
00:00:57,360 --> 00:00:59,260
Twitter the past 24 hours
26
00:00:59,260 --> 00:01:00,600
discussing it
27
00:01:00,600 --> 00:01:02,680
it's
28
00:01:02,680 --> 00:01:05,320
yeah I mean the first thing
29
00:01:05,320 --> 00:01:06,100
I'd say is just
30
00:01:06,100 --> 00:01:08,860
my heart goes out to everyone affected by this
31
00:01:08,860 --> 00:01:09,900
yeah it fucking sucks
32
00:01:09,900 --> 00:01:12,120
and
33
00:01:12,120 --> 00:01:14,780
we'll get into the details but unfortunately
34
00:01:14,780 --> 00:01:16,840
I think it's sort of only gotten started
35
00:01:16,840 --> 00:01:18,800
in terms of theft and
36
00:01:18,800 --> 00:01:19,740
impact on people
37
00:01:19,740 --> 00:01:22,160
but
38
00:01:22,160 --> 00:01:26,280
But the last count I saw, like over 1,000 BTC.
39
00:01:27,340 --> 00:01:33,260
So something like 65 million plus wouldn't surprise me if it's 100 million by this point.
40
00:01:33,880 --> 00:01:34,960
And that could keep exceeding.
41
00:01:35,020 --> 00:01:37,180
So this is a big security incident.
42
00:01:38,980 --> 00:01:41,780
So feel awful for those people.
43
00:01:43,680 --> 00:01:47,040
And I mean, one thing we can talk about throughout the show is
44
00:01:47,040 --> 00:01:51,580
this feels like the worst day in Bitcoin's history.
45
00:01:52,160 --> 00:01:54,180
really I'll make that claim
46
00:01:54,180 --> 00:01:55,320
that's a big claim yeah
47
00:01:55,320 --> 00:01:57,460
because there's been
48
00:01:57,460 --> 00:02:00,160
security incidents before
49
00:02:00,160 --> 00:02:02,260
there's been thefts and loss of funds
50
00:02:02,260 --> 00:02:04,180
all of those are tragic
51
00:02:04,180 --> 00:02:04,980
and horrible
52
00:02:04,980 --> 00:02:08,060
this one though is
53
00:02:08,060 --> 00:02:10,180
self custody and it's a very
54
00:02:10,180 --> 00:02:11,140
serious vulnerability
55
00:02:11,140 --> 00:02:14,320
it was people doing it the right way
56
00:02:14,320 --> 00:02:16,180
doing what they'd always been
57
00:02:16,180 --> 00:02:18,040
advised cold storage
58
00:02:18,040 --> 00:02:20,180
and cold card is a product
59
00:02:20,180 --> 00:02:21,700
that's known for its like
60
00:02:21,700 --> 00:02:24,640
power features and extra security features.
61
00:02:25,500 --> 00:02:29,300
So, you know, it's hard to blame the people
62
00:02:29,300 --> 00:02:32,440
who chose that path to secure their Bitcoin
63
00:02:32,440 --> 00:02:34,960
that they did something wrong.
64
00:02:35,880 --> 00:02:41,120
So later in the show, we'll talk about
65
00:02:41,120 --> 00:02:43,160
the future of self-custody and the impact on that.
66
00:02:43,360 --> 00:02:45,860
But there's, I mean, certainly part of that discussion
67
00:02:45,860 --> 00:02:50,380
is like how big a setback will this be for self-custody?
68
00:02:50,380 --> 00:02:55,540
And I would argue self-custody is an essential option.
69
00:02:56,040 --> 00:02:59,200
We need that option in Bitcoin for Bitcoin to be healthy long-term.
70
00:03:00,620 --> 00:03:07,300
So that's why I'd make the claim that it might be the worst 24 hours in Bitcoin's history.
71
00:03:07,320 --> 00:03:09,840
Totally. And I'm sure you've got a lot of this outline.
72
00:03:09,980 --> 00:03:16,460
But to me, the other biggest story, I mean, I think, I mean, there's a lot of ways to think about this.
73
00:03:16,460 --> 00:03:20,040
And I think it's good that everyone thinks through things with a calm and clear head, obviously.
74
00:03:20,040 --> 00:03:40,060
But I think to me, the even bigger story is, you know, while everyone's been kind of afraid of quantum coming, to me, the big story is just AI. Like you're saying there's a lot more to come with cold cards, sure. But I mean, I would say I have in my own personal experience, you know, as you guys saw yesterday, I actually did have some sort of anxiety during all of this.
75
00:03:40,060 --> 00:03:42,080
because I think, you know,
76
00:03:43,020 --> 00:03:44,100
it's cold card today,
77
00:03:44,280 --> 00:03:46,700
but what's to stop us from believing that,
78
00:03:47,040 --> 00:03:48,940
you know, the successor to Fable 5,
79
00:03:49,320 --> 00:03:50,640
and to be clear, like, make no mistake,
80
00:03:50,740 --> 00:03:52,520
this is a direct result of AI getting better,
81
00:03:52,900 --> 00:03:54,440
is going to be able to find some other weird vulnerability.
82
00:03:54,660 --> 00:03:56,820
Like, to me, the biggest takeaway from all of this
83
00:03:56,820 --> 00:03:59,260
is not an indictment of any single company or anything.
84
00:03:59,340 --> 00:04:00,320
The biggest thing to me is,
85
00:04:01,180 --> 00:04:02,460
and I know you guys have been pushing this,
86
00:04:02,460 --> 00:04:04,820
so kudos at Block and with Loop and everything,
87
00:04:05,020 --> 00:04:06,960
but, like, the Bitcoin ecosystem
88
00:04:06,960 --> 00:04:09,040
has to get access to the
89
00:04:10,060 --> 00:04:13,860
absolute cutting-edge frontier models in stealth, full stop.
90
00:04:14,200 --> 00:04:17,560
And to me, wherever I would feel comfortable custodying,
91
00:04:17,600 --> 00:04:19,420
whether it's self-custody, custodian,
92
00:04:19,920 --> 00:04:22,400
really comes down to who do I know is getting access
93
00:04:22,400 --> 00:04:24,960
to the latest Anthropic and OpenAI models, full stop.
94
00:04:25,300 --> 00:04:25,380
Yep.
95
00:04:25,680 --> 00:04:28,180
We'll definitely cover that in depth.
96
00:04:28,260 --> 00:04:29,420
I think that's a really important topic.
97
00:04:30,780 --> 00:04:33,740
But maybe we start with if people are listening
98
00:04:33,740 --> 00:04:36,960
and they haven't been watching Twitter,
99
00:04:37,400 --> 00:04:39,900
what are we even talking about?
100
00:04:40,060 --> 00:04:44,240
So we're talking about a very serious security incident.
101
00:04:45,440 --> 00:04:51,340
Coldcard is a popular brand and product that they've been around at least a decade, I think.
102
00:04:53,120 --> 00:04:56,120
It's a hardware wallet for people to store their Bitcoin.
103
00:04:57,120 --> 00:05:02,380
And yesterday, I think it was around morning or early afternoon Pacific time,
104
00:05:02,380 --> 00:05:09,800
I started seeing tweets by credible people like Kevin Loic, who I've known for a long time.
105
00:05:09,800 --> 00:05:11,260
credible person in the Bitcoin space
106
00:05:11,260 --> 00:05:13,780
tweeting that he had been hearing from people
107
00:05:13,780 --> 00:05:17,220
that they were noticing their funds were gone
108
00:05:17,220 --> 00:05:18,600
and they were using cold cards.
109
00:05:19,040 --> 00:05:22,660
And he's the kind of person who doesn't like troll tweets.
110
00:05:23,160 --> 00:05:26,340
And he runs a security advisory firm in Bitcoin.
111
00:05:26,700 --> 00:05:29,300
So if he's tweeting something,
112
00:05:29,540 --> 00:05:31,220
you kind of take it seriously.
113
00:05:32,900 --> 00:05:34,780
So that raised some awareness.
114
00:05:34,980 --> 00:05:35,940
And then throughout the day,
115
00:05:36,060 --> 00:05:37,620
more and more people are researching it,
116
00:05:37,620 --> 00:05:39,900
more data and evidence coming about.
117
00:05:40,720 --> 00:05:42,560
And it turns out that
118
00:05:42,560 --> 00:05:46,680
many cold card products are affected.
119
00:05:47,500 --> 00:05:49,000
We'll get into some of the details,
120
00:05:49,220 --> 00:05:52,260
but some are easier to hack than others.
121
00:05:53,220 --> 00:05:55,120
But as I already mentioned,
122
00:05:56,100 --> 00:05:57,520
the last count I saw,
123
00:05:57,520 --> 00:06:01,520
over 1,000 BTC have been transferred.
124
00:06:02,760 --> 00:06:04,820
And it's observable on chain
125
00:06:04,820 --> 00:06:08,240
through a very clear pattern
126
00:06:08,240 --> 00:06:10,560
that lots of independent people
127
00:06:10,560 --> 00:06:14,820
have discovered and shared.
128
00:06:14,980 --> 00:06:16,360
What is the TLDR and the pattern?
129
00:06:16,940 --> 00:06:18,860
Is it just that they're larger transaction sizes?
130
00:06:21,560 --> 00:06:24,100
I couldn't tell you all the details of that pattern.
131
00:06:24,100 --> 00:06:26,600
I mean, they're looking at certain signature types
132
00:06:26,600 --> 00:06:27,620
and stuff like that.
133
00:06:28,040 --> 00:06:33,500
There's also a fixed fee of 30 sats per vbyte
134
00:06:33,500 --> 00:06:35,600
or something that was common across all of them.
135
00:06:35,720 --> 00:06:37,760
But there were several other details.
136
00:06:38,040 --> 00:06:39,880
It was much higher than average, right?
137
00:06:39,960 --> 00:06:40,400
So you can look at it.
138
00:06:40,400 --> 00:06:41,600
But there were several other details
139
00:06:41,600 --> 00:06:43,140
that made it very clear.
140
00:06:43,400 --> 00:06:44,380
And so that pattern,
141
00:06:44,900 --> 00:06:47,760
but even today, I think there were more thefts,
142
00:06:47,820 --> 00:06:49,200
a different fee rate was used.
143
00:06:50,680 --> 00:06:55,160
But there's a pretty clear identifiable pattern.
144
00:06:56,680 --> 00:06:57,580
In any case,
145
00:06:59,580 --> 00:07:02,620
yeah, so it's a high-impact event.
146
00:07:02,620 --> 00:07:06,380
So what are some of the details of the vulnerability?
147
00:07:08,120 --> 00:07:16,280
So there's certain, you know, cold card, there's different hardware products and there's different firmware software versions for the product.
148
00:07:16,280 --> 00:07:32,240
So if you have like an MK2 or MK3 version of the hardware and you have firmware that is dated somewhere from like 2021 until I think current, you are at grave risk.
149
00:07:33,000 --> 00:07:36,600
It's a very easy, reproducible step.
150
00:07:36,760 --> 00:07:41,620
Like any of us could just use AI, steal these coins in minutes.
151
00:07:41,860 --> 00:07:43,820
It's very, very weak entropy.
152
00:07:43,820 --> 00:07:45,480
So what's the vulnerability?
153
00:07:45,480 --> 00:07:58,320
The keys generated on the device are very weak, meaning you don't need physical possession of the device to steal the coins.
154
00:07:58,740 --> 00:08:04,880
You don't need to have gained access to the seed phrase to steal the coins.
155
00:08:04,880 --> 00:08:07,1000
you can just look at the blockchain
156
00:08:07,1000 --> 00:08:12,880
and be able to regenerate the keys
157
00:08:12,880 --> 00:08:15,600
because the entropy is so weak.
158
00:08:15,760 --> 00:08:17,900
How does that actually functionally work?
159
00:08:18,440 --> 00:08:20,960
So they'll see a specific transaction on the blockchain
160
00:08:20,960 --> 00:08:25,140
and what operations, I guess, is the eye crunching to be able to...
161
00:08:25,140 --> 00:08:26,500
Well, here's what I know.
162
00:08:26,500 --> 00:08:32,680
So the bug in this is that the hardware random number generator
163
00:08:32,680 --> 00:08:35,840
for those devices isn't being used.
164
00:08:35,840 --> 00:08:38,960
And that was a compile time bug.
165
00:08:39,780 --> 00:08:41,160
When they built the software,
166
00:08:41,960 --> 00:08:44,780
they had the code to do it properly,
167
00:08:45,180 --> 00:08:47,140
but there was like an if def that was wrong,
168
00:08:47,520 --> 00:08:48,580
or if not def or whatever.
169
00:08:49,140 --> 00:08:50,980
It built the software incorrectly,
170
00:08:51,880 --> 00:08:55,360
and so it falls back to a pseudo random number generator
171
00:08:55,360 --> 00:08:57,420
that's software-based, not hardware-based.
172
00:08:58,500 --> 00:08:59,660
One question on that,
173
00:08:59,760 --> 00:09:02,320
and I'm sure everyone now is looking back
174
00:09:02,320 --> 00:09:05,200
I'm thinking for other hardware wallets and whatever.
175
00:09:05,660 --> 00:09:06,620
Why is that even an option?
176
00:09:06,800 --> 00:09:09,560
Like if we know that software random number generators
177
00:09:09,560 --> 00:09:12,400
are not seeded with sufficient entropy,
178
00:09:12,520 --> 00:09:14,240
why is that even an option for any of these?
179
00:09:15,700 --> 00:09:17,460
Well, I certainly can't speak to, you know,
180
00:09:17,680 --> 00:09:22,140
how they got to, you know, that point with their software builds.
181
00:09:23,140 --> 00:09:25,020
But, you know, there's, at least historically,
182
00:09:25,120 --> 00:09:26,080
there's been hardware that doesn't,
183
00:09:26,240 --> 00:09:28,220
there is no hardware random number generator.
184
00:09:28,480 --> 00:09:28,600
Right.
185
00:09:28,880 --> 00:09:30,500
You could also imagine building it,
186
00:09:30,500 --> 00:09:31,460
like while you're building it,
187
00:09:31,480 --> 00:09:34,860
before you have a hardware random number generator there,
188
00:09:34,920 --> 00:09:36,540
you might actually build a version of this
189
00:09:36,540 --> 00:09:38,920
that you can test on the local computer
190
00:09:38,920 --> 00:09:41,660
that doesn't have that specific piece of hardware
191
00:09:41,660 --> 00:09:42,600
you're planning to call.
192
00:09:43,260 --> 00:09:45,560
So maybe like while you're developing it,
193
00:09:45,760 --> 00:09:47,180
you develop it in one way with the intention
194
00:09:47,180 --> 00:09:49,100
to flip it on before you go into production.
195
00:09:49,300 --> 00:09:51,420
I wonder though if in the future of wallet development,
196
00:09:51,600 --> 00:09:53,540
people will just cut that off entirely.
197
00:09:53,680 --> 00:09:54,560
Like you don't even want that
198
00:09:54,560 --> 00:09:57,260
anywhere near your secure element as an option.
199
00:09:57,600 --> 00:09:58,000
Well, yeah.
200
00:09:58,000 --> 00:10:00,380
I mean, I think knowing what we know,
201
00:10:00,540 --> 00:10:01,480
what everyone knows now,
202
00:10:01,580 --> 00:10:02,040
if you look back,
203
00:10:02,260 --> 00:10:04,460
obviously you wouldn't have those errors,
204
00:10:04,620 --> 00:10:05,800
you wouldn't have that code, etc.
205
00:10:06,140 --> 00:10:08,180
So, but in any case,
206
00:10:08,760 --> 00:10:11,500
it led to a situation where there was,
207
00:10:11,500 --> 00:10:15,060
there was no like cryptographic entropy for the key.
208
00:10:16,100 --> 00:10:17,520
You still have to like,
209
00:10:17,800 --> 00:10:18,900
there's still some inputs,
210
00:10:19,320 --> 00:10:23,140
like the CPU timer cycle and a few other things,
211
00:10:23,240 --> 00:10:25,740
but all of that can be brute forced quite easily.
212
00:10:25,740 --> 00:10:28,880
and like the UID,
213
00:10:29,100 --> 00:10:30,860
like unique ID with the hardware.
214
00:10:31,380 --> 00:10:32,960
So each cold card comes,
215
00:10:33,360 --> 00:10:34,620
there's an ID that you can look,
216
00:10:35,200 --> 00:10:36,980
if you open it up,
217
00:10:37,120 --> 00:10:39,340
that was etched into it
218
00:10:39,340 --> 00:10:42,220
from the manufacturing process
219
00:10:42,220 --> 00:10:42,820
that's unique.
220
00:10:43,000 --> 00:10:45,200
That's also like an input into this key.
221
00:10:45,880 --> 00:10:46,340
But if you,
222
00:10:46,420 --> 00:10:48,260
you can brute force all these variables
223
00:10:48,260 --> 00:10:52,100
and it's the deterministic process
224
00:10:52,100 --> 00:10:54,460
to generate the keys.
225
00:10:54,460 --> 00:11:08,400
So for those products, and so the coins that have been stolen so far, at least as of the ones yesterday, I think were all from those hardware models, and it's this relatively easy attack.
226
00:11:08,400 --> 00:11:12,680
then there's been additional
227
00:11:12,680 --> 00:11:15,240
research on newer models
228
00:11:15,240 --> 00:11:16,1000
MK4, MK5 and Q
229
00:11:16,1000 --> 00:11:20,600
Locke did a lot of research
230
00:11:20,600 --> 00:11:22,880
on this and published a report
231
00:11:22,880 --> 00:11:27,060
yesterday late afternoon
232
00:11:27,060 --> 00:11:27,780
Pacific Time
233
00:11:27,780 --> 00:11:28,760
it was really good, it was really thorough
234
00:11:28,760 --> 00:11:31,760
so it spoke to the first vulnerability
235
00:11:31,760 --> 00:11:35,160
it also spoke to the second
236
00:11:35,160 --> 00:11:43,660
And these newer models, there is real cryptographic entropy introduced, I think like 32 bits worth.
237
00:11:44,260 --> 00:11:47,060
So it's much stronger than zero.
238
00:11:48,760 --> 00:11:52,820
And so it's not easily brute forced, but 32 bits.
239
00:11:53,440 --> 00:11:59,740
I think in total, even today I saw further analysis to try to narrow the window.
240
00:11:59,740 --> 00:12:04,860
Because I think in the published report, the range is given like 32 to 72 bits.
241
00:12:05,160 --> 00:12:08,040
depending on sort of what the attacker,
242
00:12:08,280 --> 00:12:09,940
what assumptions the attacker has,
243
00:12:10,340 --> 00:12:11,880
that's a very big range.
244
00:12:12,020 --> 00:12:13,880
That's a range from like trivial,
245
00:12:14,280 --> 00:12:16,780
or you know, like not that much money
246
00:12:16,780 --> 00:12:19,220
to millions and millions and millions of dollars.
247
00:12:21,440 --> 00:12:22,900
I think they're in the process
248
00:12:22,900 --> 00:12:24,740
of trying to narrow that range further.
249
00:12:25,640 --> 00:12:27,100
I'm sure if they do and they're confident,
250
00:12:27,320 --> 00:12:28,740
they'll publish that as well.
251
00:12:28,740 --> 00:12:34,960
But this one would be a substantial,
252
00:12:35,160 --> 00:12:37,240
financial cost to do the
253
00:12:37,240 --> 00:12:39,240
to do the attack
254
00:12:39,240 --> 00:12:40,520
but it
255
00:12:40,520 --> 00:12:43,300
would be that the MK4s
256
00:12:43,300 --> 00:12:44,620
MK5s and Q could
257
00:12:44,620 --> 00:12:45,860
result
258
00:12:45,860 --> 00:12:48,960
and so now I see
259
00:12:48,960 --> 00:12:50,960
reports on X of people saying that
260
00:12:50,960 --> 00:12:53,240
the MK4, 5 and Q
261
00:12:53,240 --> 00:12:55,280
have been hacked
262
00:12:55,280 --> 00:12:57,140
so
263
00:12:57,140 --> 00:12:58,960
one plausible explanation of that is like
264
00:12:58,960 --> 00:13:01,040
an attacker is trying
265
00:13:01,040 --> 00:13:03,000
to reproduce all those
266
00:13:03,000 --> 00:13:05,080
keys and while it might cost them
267
00:13:05,080 --> 00:13:06,860
a lot of money to reproduce all the keys,
268
00:13:06,1000 --> 00:13:08,780
as they're marching along in that process,
269
00:13:08,880 --> 00:13:10,540
they're going to find some keys early on.
270
00:13:11,340 --> 00:13:13,560
So it wouldn't take too much time
271
00:13:13,560 --> 00:13:15,660
and that much expense to find some early ones.
272
00:13:15,980 --> 00:13:17,300
I don't know if that's true or not.
273
00:13:17,360 --> 00:13:19,960
That's just a theory, but it could be the case.
274
00:13:22,040 --> 00:13:25,580
And then there's other variations to this.
275
00:13:25,580 --> 00:13:26,920
So I mentioned earlier,
276
00:13:27,040 --> 00:13:29,600
the cold card did have a bunch of security options
277
00:13:29,600 --> 00:13:31,800
for power users.
278
00:13:32,020 --> 00:13:34,960
So it had like a dice roll feature.
279
00:13:35,080 --> 00:13:37,580
built in. Like with any hardware wallet, you can
280
00:13:37,580 --> 00:13:40,960
generate your, I should say, with most hardware
281
00:13:40,960 --> 00:13:44,160
wallets, you can generate your own seed phrase,
282
00:13:44,280 --> 00:13:46,480
your own entropy fully. You can just skip
283
00:13:46,480 --> 00:13:49,680
their solution and you can just enter
284
00:13:49,680 --> 00:13:52,560
your seed phrase, which you can generate
285
00:13:52,560 --> 00:13:55,120
offline or any other process you want to use.
286
00:13:55,440 --> 00:13:58,180
And that could be a good process or a bad process.
287
00:13:58,360 --> 00:14:00,320
If you know what you're doing and it's a good process,
288
00:14:00,960 --> 00:14:03,740
that's one of the most secure ways you can go about
289
00:14:03,740 --> 00:14:05,020
securing your Bitcoin.
290
00:14:05,960 --> 00:14:08,000
But there's definitely foot guns in that process.
291
00:14:08,120 --> 00:14:10,040
But in any case, Coldcard had an
292
00:14:10,040 --> 00:14:12,060
explicit feature built in to
293
00:14:12,060 --> 00:14:12,680
roll dice.
294
00:14:14,080 --> 00:14:15,700
I believe their guidance is
295
00:14:15,700 --> 00:14:17,640
rolling like 99 or 100
296
00:14:17,640 --> 00:14:18,440
times.
297
00:14:20,060 --> 00:14:21,680
To get a sufficient
298
00:14:21,680 --> 00:14:24,060
entropy. Or maybe it's like 50 or 60 if you
299
00:14:24,060 --> 00:14:25,980
want like 128 bit and it's 100
300
00:14:25,980 --> 00:14:28,080
rolls for 256 bit
301
00:14:28,080 --> 00:14:29,980
or something like that. But it's not
302
00:14:29,980 --> 00:14:32,120
like a 10 second
303
00:14:32,120 --> 00:14:36,340
process you just roll dice a few times and you're done yeah you got to be pretty committed to it
304
00:14:36,340 --> 00:14:43,080
so add entropy to the whatever random number generation they came up with they kind of start
305
00:14:43,080 --> 00:15:03,172
with their own seed then you doing that role you applying additional entropy to so even if you had the mk2 or mk3 which had horrible weak entropy if you rolled a dice properly and rolled them 100 times or whatever then you actually have sufficient entropy and you actually not at risk
306
00:15:03,172 --> 00:15:08,652
And is it generally accepted that kind of any good-looking, reasonable dice is okay?
307
00:15:08,852 --> 00:15:14,152
Or do you really like the super casino-grade dice to get good enough entropy?
308
00:15:14,152 --> 00:15:16,132
It's been a while since I've studied that.
309
00:15:16,132 --> 00:15:23,172
I mean, clearly, if there are dice that aren't perfectly manufactured, they'll have some bias in their roles.
310
00:15:23,792 --> 00:15:27,492
And that can reduce the search space.
311
00:15:28,132 --> 00:15:31,552
What I don't recall offhand is like, how much?
312
00:15:32,772 --> 00:15:44,112
And like, for example, if you do 100 rolls and have 256-bit security and you have imperfect dice, does it just reduce it to like 192-bit security?
313
00:15:44,112 --> 00:15:54,612
Well, my instinct is yes, but I'm not, please don't, go verify yourself, but my instinct is yes, and then in which you still have very sufficient security.
314
00:15:54,732 --> 00:15:55,992
How does that actually work?
315
00:15:56,072 --> 00:16:00,652
So with the dice rolling, how do you convert that, the entropy from the dice rolls, into changing the words?
316
00:16:01,372 --> 00:16:02,292
I'll answer that in a second.
317
00:16:02,472 --> 00:16:03,452
But just one more thing.
318
00:16:04,672 --> 00:16:08,112
Clearly, I think clearly the most secure way, though, is to combine sources, right?
319
00:16:08,112 --> 00:16:13,072
So, like advice I'd give to someone who is going,
320
00:16:13,472 --> 00:16:16,132
that I'm confident is going to be diligent enough to do this properly,
321
00:16:16,952 --> 00:16:18,372
which is not everyone I know,
322
00:16:18,592 --> 00:16:21,032
but if they're diligent enough to do it properly,
323
00:16:21,492 --> 00:16:24,432
use the random number generator from the product you buy.
324
00:16:24,992 --> 00:16:29,172
And then also add your own entropy with like dice rolls or coin flips
325
00:16:29,172 --> 00:16:30,472
or things like that.
326
00:16:31,052 --> 00:16:33,052
And do both.
327
00:16:33,052 --> 00:16:36,292
Because then if you have imperfect dice,
328
00:16:36,292 --> 00:16:45,592
And let's say it only adds like 70 bits of security, which I wouldn't be comfortable with if that's all the security was.
329
00:16:45,672 --> 00:16:52,172
But if you combine that with a random number generator, hopefully one that is not horribly broken, then you're going to have sufficient security.
330
00:16:52,172 --> 00:17:22,032
But I mean, is it worth, you know, there's always problems that could emerge when you're sort of relying on somebody else's entropy. Like, what if there's something in how the combination works between a dice roll that you are increasing the entropy in the system, supposedly, but maybe the software that actually does the increasing of the entropy has a risk versus, is it not better, like, things can go wrong? Is it not better to just do the dice roll outside, generate the, you know, the words and then use those and import them?
331
00:17:22,032 --> 00:17:29,672
Yeah, meaning if you do perfect dice rolls, then you input it to some software that you've not audited, they might have a bug.
332
00:17:29,732 --> 00:17:30,912
Yeah, I agree that's a vulnerability.
333
00:17:31,292 --> 00:17:34,532
So, and I'll get back to your point too in a second.
334
00:17:35,432 --> 00:17:43,712
So ways to handle that though is, you know, if you have a multi-sig, and later on in the show we'll get into all the different options that I think are good.
335
00:17:43,872 --> 00:17:45,472
So I don't want to get into too much detail.
336
00:17:45,472 --> 00:17:50,792
But if you have a multi-sig, you could imagine, first of all, vendor diversity makes sense.
337
00:17:50,792 --> 00:17:53,952
So choose different vendors for your different keys.
338
00:17:54,592 --> 00:17:58,452
And then for some, you could use the random number generator.
339
00:17:58,632 --> 00:18:00,072
For others, you could use your own entropy.
340
00:18:00,352 --> 00:18:01,632
That's a way to get diversity.
341
00:18:02,492 --> 00:18:06,412
And you're not falling prey to what you mentioned, right?
342
00:18:06,632 --> 00:18:10,112
Because some of the keys would be fully my entropy from dice rolls.
343
00:18:10,352 --> 00:18:13,372
Other keys would be not my intervention whatsoever.
344
00:18:13,372 --> 00:18:14,052
That's a good point.
345
00:18:14,552 --> 00:18:15,752
Sorry, what was your point again?
346
00:18:15,752 --> 00:18:18,132
How do you convert the dice roll entropy to actually change the current?
347
00:18:18,132 --> 00:18:19,992
I mean, there's different methods.
348
00:18:19,992 --> 00:18:21,792
and now with AI
349
00:18:21,792 --> 00:18:22,552
you could ask AI
350
00:18:22,552 --> 00:18:23,732
to get recommendations
351
00:18:23,732 --> 00:18:25,772
but I would definitely
352
00:18:25,772 --> 00:18:27,632
search for best practices
353
00:18:27,632 --> 00:18:28,352
doing this
354
00:18:28,352 --> 00:18:29,732
but there's definitely
355
00:18:29,732 --> 00:18:31,732
word lists out there
356
00:18:31,732 --> 00:18:34,112
from credible organizations
357
00:18:34,112 --> 00:18:34,772
like EFF
358
00:18:34,772 --> 00:18:36,212
and other places
359
00:18:36,212 --> 00:18:39,172
that map like five
360
00:18:39,172 --> 00:18:41,212
if you have five dice
361
00:18:41,212 --> 00:18:42,832
and roll the five dice
362
00:18:42,832 --> 00:18:43,392
they map it
363
00:18:43,392 --> 00:18:45,132
no it's like one, four, six
364
00:18:45,132 --> 00:18:46,692
and then it maps to a word
365
00:18:46,692 --> 00:18:47,512
and so
366
00:18:47,512 --> 00:18:48,632
but you need
367
00:18:48,632 --> 00:18:51,932
you need to be very knowledgeable
368
00:18:51,932 --> 00:18:53,272
and deliberate about this process.
369
00:18:53,272 --> 00:18:55,312
So I think a very common mistake
370
00:18:55,312 --> 00:18:57,172
people would do is that you roll the dice
371
00:18:57,172 --> 00:18:59,412
and then you like sort of line them,
372
00:18:59,492 --> 00:19:01,052
you move them around and line them up.
373
00:19:01,292 --> 00:19:02,992
Then you're screwing with entropy.
374
00:19:03,872 --> 00:19:05,112
You can't order them.
375
00:19:05,292 --> 00:19:06,872
Or like maybe a common instinct would be like,
376
00:19:06,952 --> 00:19:08,552
oh, I'll order them like into,
377
00:19:08,792 --> 00:19:11,692
like one, three, four, five, five.
378
00:19:12,152 --> 00:19:15,032
Don't do that because then you've reduced the entropy.
379
00:19:15,452 --> 00:19:17,972
You need to, so like one piece of advice,
380
00:19:17,972 --> 00:19:19,972
if you can get different colored dice.
381
00:19:20,212 --> 00:19:21,612
Or different manufacturers even.
382
00:19:22,092 --> 00:19:23,552
Well, different colored dice.
383
00:19:24,192 --> 00:19:25,552
So each die is a different color.
384
00:19:25,692 --> 00:19:28,372
And then you write down like green goes first,
385
00:19:28,672 --> 00:19:29,592
then yellow, whatever,
386
00:19:29,992 --> 00:19:32,092
then roll and then always stick to that.
387
00:19:32,192 --> 00:19:32,912
That would be one way.
388
00:19:32,992 --> 00:19:34,392
Or just the way you roll,
389
00:19:34,792 --> 00:19:36,032
you just always go left to right.
390
00:19:36,332 --> 00:19:37,472
I think it's even simpler.
391
00:19:38,032 --> 00:19:39,652
Take the dice, throw them out,
392
00:19:39,772 --> 00:19:40,432
and don't look at it.
393
00:19:40,492 --> 00:19:40,952
Don't look at it.
394
00:19:41,212 --> 00:19:42,752
Organize them into a line.
395
00:19:43,392 --> 00:19:44,212
And then just read the line.
396
00:19:45,072 --> 00:19:46,692
Or don't even touch them, no?
397
00:19:46,692 --> 00:19:51,072
No, I'm saying organize them so while you're not looking, you don't see what you're doing.
398
00:19:51,152 --> 00:19:52,032
You're not trying to do something.
399
00:19:52,292 --> 00:19:53,312
Yeah, and then you want to be cautious.
400
00:19:53,472 --> 00:19:58,072
If you are looking on a word list, if you're doing it on an online computer,
401
00:19:58,332 --> 00:20:00,792
if you're super paranoid and there's malware in the computer,
402
00:20:00,932 --> 00:20:02,052
it might know what you're looking at.
403
00:20:02,132 --> 00:20:04,272
I mean, you can go to different levels of paranoia there.
404
00:20:04,512 --> 00:20:06,632
The cold card interface for this is actually quite simple.
405
00:20:06,732 --> 00:20:08,272
It's much simpler than that, if I recall correctly.
406
00:20:08,572 --> 00:20:14,592
I haven't used it in many years, but my recollection is that you can roll a dice
407
00:20:14,592 --> 00:20:16,552
and then input a number and roll a dice and input a number.
408
00:20:16,692 --> 00:20:18,352
That's why it requires 100.
409
00:20:18,532 --> 00:20:20,812
So the word list I'm talking about is a different process.
410
00:20:21,212 --> 00:20:22,532
You don't need to roll 100 times.
411
00:20:24,052 --> 00:20:28,432
If you want 24 words, then you'd roll five dice 24 times.
412
00:20:28,592 --> 00:20:31,432
And you get a ton of entropy as long as you follow a proper process.
413
00:20:31,592 --> 00:20:34,292
But yeah, cold card, it's a simpler process with a lot of rolls.
414
00:20:36,212 --> 00:20:39,932
So anyway, we'll get into more details in a bit.
415
00:20:39,932 --> 00:20:45,892
But I guess that's the very high-level description of the vulnerabilities here.
416
00:20:45,892 --> 00:20:52,552
And just to like do another zoom out on that, Bitcoin itself does not have any security problems discovered.
417
00:20:52,592 --> 00:20:53,032
Yeah, exactly.
418
00:20:53,292 --> 00:20:54,512
Bitcoin itself is fine.
419
00:20:54,712 --> 00:21:04,452
There's a specific vendor, CoinKite, specific product, ColdCard Mark II and III and maybe IV, V and Q that have random number generation vulnerabilities.
420
00:21:05,132 --> 00:21:14,532
So it's kind of a horrible thing where heart goes out to people who are involved in this, but it's more of a specific vendor problem with random number generation.
421
00:21:14,532 --> 00:21:15,532
It's not a Bitcoin problem.
422
00:21:15,532 --> 00:21:23,332
Yeah, and to be clear, when I said at the top that, you know, while everyone's like kind of worried about quantum, like quantum actually potentially could attack the base layer of Bitcoin itself.
423
00:21:23,732 --> 00:21:26,252
But I do think it's important to sort of recognize.
424
00:21:26,512 --> 00:21:36,592
I do think something in the Bitcoin community a lot of times will go very deep on like this one, like kind of like obscure this thing could happen when the biggest vulnerability is often just like the simple things.
425
00:21:36,732 --> 00:21:37,292
Like right in front of you.
426
00:21:37,332 --> 00:21:37,672
Exactly.
427
00:21:37,832 --> 00:21:39,352
What are the companies that are doing this?
428
00:21:39,532 --> 00:21:42,472
Like who's, you know, what AI do they have access to, et cetera.
429
00:21:43,372 --> 00:21:45,252
The one more thing I want to mention,
430
00:21:45,392 --> 00:21:46,652
we talked about the dice rolls.
431
00:21:47,072 --> 00:21:47,912
So again, just to be clear,
432
00:21:48,272 --> 00:21:51,612
so cold card customers who did the proper number of dice rolls
433
00:21:51,612 --> 00:21:53,212
and they're confident how they did that,
434
00:21:53,432 --> 00:21:55,052
they should not be at risk.
435
00:21:56,352 --> 00:21:57,592
Then there's another...
436
00:21:57,592 --> 00:21:58,492
So if you're in that category,
437
00:21:58,872 --> 00:22:01,532
like you have, say, an MK2 or 3,
438
00:22:01,652 --> 00:22:02,972
which are the bad ones,
439
00:22:03,092 --> 00:22:06,812
but you've done the dice roll to 100 dice rolls or something,
440
00:22:07,432 --> 00:22:08,332
you should just sit tight.
441
00:22:08,332 --> 00:22:09,072
Don't do anything?
442
00:22:09,152 --> 00:22:09,932
Is that kind of the advice?
443
00:22:10,992 --> 00:22:12,332
That is the expectation.
444
00:22:12,472 --> 00:22:25,012
If it were me, just to sleep better at night, even if I can intellectualize that I'm safe, I still might move funds.
445
00:22:25,932 --> 00:22:33,852
But the known vulnerability suggests that you'd be fine, that you're safe.
446
00:22:36,592 --> 00:22:41,832
I'll get to some other thinking around maybe advice for that type of person in that situation as well.
447
00:22:42,472 --> 00:22:44,492
But also let's talk about passphrases.
448
00:22:45,052 --> 00:22:48,752
So with the way a lot of these wallets work,
449
00:22:49,152 --> 00:22:53,592
you can generate like a 12 or 24 word passphrase,
450
00:22:53,692 --> 00:22:55,792
or I shouldn't say a seed.
451
00:22:55,892 --> 00:22:56,292
Seed phrase.
452
00:22:56,472 --> 00:22:56,712
Seed.
453
00:22:57,832 --> 00:23:01,732
And then there's a standard called BIP39.
454
00:23:02,232 --> 00:23:06,112
It allows for either a 13th or 25th word
455
00:23:06,112 --> 00:23:07,832
that's called the passphrase.
456
00:23:08,792 --> 00:23:11,272
It's different though than all the other words
457
00:23:11,272 --> 00:23:15,312
because the other words all conform to a specific dictionary.
458
00:23:15,652 --> 00:23:18,432
Like the words have to come from a publicly known dictionary.
459
00:23:20,032 --> 00:23:22,772
But if you have like 12 or 24 of them randomly generated
460
00:23:22,772 --> 00:23:27,052
in a certain order, it's like impossible to crack
461
00:23:27,052 --> 00:23:28,452
if it's strong entropy.
462
00:23:29,952 --> 00:23:33,632
With the passphrase, it can be any string at all.
463
00:23:34,732 --> 00:23:38,032
And the length of that string,
464
00:23:38,812 --> 00:23:40,492
it's not literally unbounded
465
00:23:40,492 --> 00:23:43,732
because it depends on the implementation of the wallet itself,
466
00:23:43,852 --> 00:23:44,832
how long it can be.
467
00:23:44,892 --> 00:23:46,712
But it can be a long string.
468
00:23:46,712 --> 00:23:49,612
We call it a 13th word, but it really could be a 13th word.
469
00:23:49,812 --> 00:23:51,692
It could be a 50-character string.
470
00:23:51,972 --> 00:23:53,012
Tons of other words in it.
471
00:23:53,032 --> 00:23:53,912
So it can be weak.
472
00:23:53,972 --> 00:23:54,472
It could be strong.
473
00:23:54,612 --> 00:23:56,192
I mean, it can be like 1, 2, 3, 4.
474
00:23:56,472 --> 00:23:57,992
Or I love you or something.
475
00:23:58,172 --> 00:23:59,212
It's like a password.
476
00:24:00,472 --> 00:24:02,372
Or it can be super, super strong.
477
00:24:02,372 --> 00:24:04,512
It could be like 256-bit or more.
478
00:24:04,852 --> 00:24:05,212
Security.
479
00:24:06,072 --> 00:24:09,352
So if a person is using a passphrase,
480
00:24:09,352 --> 00:24:14,192
if it's strong, secure, and well-guarded,
481
00:24:14,712 --> 00:24:17,992
then even if you're using this weak entropy seed,
482
00:24:18,672 --> 00:24:22,152
if it's like 128-bit or more security in the passphrase,
483
00:24:22,232 --> 00:24:24,532
you also should be safe.
484
00:24:26,912 --> 00:24:28,272
But again, it comes in the pocket,
485
00:24:28,532 --> 00:24:31,332
like, will you sleep better at night if you move your phones?
486
00:24:31,552 --> 00:24:33,932
So I think probably the right advice is
487
00:24:33,932 --> 00:24:37,912
if you have a product in the firmware version
488
00:24:37,912 --> 00:24:39,492
that is at most risk,
489
00:24:39,592 --> 00:24:42,332
you need to take immediate action.
490
00:24:43,272 --> 00:24:44,512
It's almost a certainty
491
00:24:44,512 --> 00:24:46,652
your coins are going to be stolen
492
00:24:46,652 --> 00:24:49,472
and like within minutes, hours, or days.
493
00:24:49,992 --> 00:24:53,932
If you have a passphrase
494
00:24:53,932 --> 00:24:56,872
that maybe it's weak or you're not sure,
495
00:24:57,312 --> 00:24:58,492
or if you rolled dice,
496
00:24:58,572 --> 00:25:00,752
but you forget how many times you rolled the dice,
497
00:25:01,352 --> 00:25:03,132
or if you have a newer product
498
00:25:03,132 --> 00:25:05,092
like MK4 or 5 or Q,
499
00:25:05,732 --> 00:25:07,792
you might not have to panic
500
00:25:07,792 --> 00:25:11,452
and rush to move the funds today,
501
00:25:12,272 --> 00:25:17,332
but probably within days moving the funds.
502
00:25:17,812 --> 00:25:20,712
But you have a little bit more time to get a clear head
503
00:25:20,712 --> 00:25:22,752
and think about where do you want to move them.
504
00:25:23,232 --> 00:25:24,772
Another piece of advice people are sharing,
505
00:25:24,872 --> 00:25:26,012
which I think is good advice too,
506
00:25:26,012 --> 00:25:32,212
which is if you need to or you feel like you need to
507
00:25:32,212 --> 00:25:34,052
or you want to move your funds immediately,
508
00:25:34,052 --> 00:25:38,852
it can be to a temporary location.
509
00:25:39,592 --> 00:25:43,532
You don't have to do a week's worth of research
510
00:25:43,532 --> 00:25:45,692
to figure out the perfect long-term solution
511
00:25:45,692 --> 00:25:47,092
for you moving forward.
512
00:25:48,012 --> 00:25:50,052
You should still do that,
513
00:25:50,132 --> 00:25:51,952
but you can have a temporary holding pattern.
514
00:25:51,952 --> 00:25:55,812
And that could be a number of different things.
515
00:25:55,992 --> 00:25:59,952
If you're comfortable moving it to a custodian or exchange
516
00:25:59,952 --> 00:26:03,792
and KYC-ing or moving back to the exchange
517
00:26:03,792 --> 00:26:04,952
you bought your coins or what have you,
518
00:26:05,312 --> 00:26:07,892
you can go to a river or a cash app
519
00:26:07,892 --> 00:26:09,552
or a Coinbase, etc.
520
00:26:10,632 --> 00:26:15,812
If you have like an old phone or old laptop,
521
00:26:16,432 --> 00:26:18,092
you could wipe the device
522
00:26:18,092 --> 00:26:20,752
so it's a clean slate of software,
523
00:26:21,432 --> 00:26:23,892
install some wallet software on it that you trust,
524
00:26:24,292 --> 00:26:26,932
move your coins there for a few days or weeks or whatever
525
00:26:26,932 --> 00:26:28,392
until you have a better system.
526
00:26:29,252 --> 00:26:32,132
Or if you have another hardware wallet,
527
00:26:32,132 --> 00:26:34,112
but you could do that as well.
528
00:26:34,532 --> 00:26:36,612
You could also, even with a cold card,
529
00:26:36,952 --> 00:26:39,212
you could, you know, if you don't have the budget
530
00:26:39,212 --> 00:26:41,692
to buy a new computer, you don't have an old laptop,
531
00:26:41,952 --> 00:26:45,432
or you're not in a situation to do that,
532
00:26:45,532 --> 00:26:48,632
you can still use your currently vulnerable cold card.
533
00:26:49,052 --> 00:26:50,932
You just need to create a new wallet
534
00:26:50,932 --> 00:26:56,272
using the suggestions from CoinKite
535
00:26:56,272 --> 00:26:57,372
and that others are providing,
536
00:26:57,372 --> 00:26:59,772
which is roll the session number of dice
537
00:26:59,772 --> 00:27:03,592
or create a very secure seed brace.
538
00:27:03,932 --> 00:27:06,632
Also, CoinKite did release new firmware versions.
539
00:27:06,952 --> 00:27:09,212
You created a sufficient seed brace.
540
00:27:09,412 --> 00:27:11,012
Sorry, I meant a pass brace. Thank you.
541
00:27:11,812 --> 00:27:12,752
But also seed brace.
542
00:27:13,152 --> 00:27:14,392
I'm skipping ahead to the...
543
00:27:14,392 --> 00:27:16,932
Because CoinKite did release new firmware last night
544
00:27:16,932 --> 00:27:19,752
that fixes this bug for most of their products,
545
00:27:20,252 --> 00:27:21,312
not for MK3.
546
00:27:21,832 --> 00:27:22,752
I don't...
547
00:27:22,752 --> 00:27:24,792
I think they said the reason why
548
00:27:24,792 --> 00:27:26,872
is because that was...
549
00:27:26,872 --> 00:27:29,332
They had sunset the software for that.
550
00:27:29,332 --> 00:27:38,552
So this is speculation, but my guess is it's going to take them a little bit more time to get that software ready because it probably hasn't been part of their process for a while.
551
00:27:39,292 --> 00:27:42,512
I think I saw where they committed to also doing that.
552
00:27:42,772 --> 00:27:56,152
So you could install new firmware and add additional entropy yourself and use your existing product and just send your funds to that new wallet.
553
00:27:56,152 --> 00:28:05,012
So I think that's the advice that I've sort of gathered from a lot of other people that makes sense to me.
554
00:28:05,232 --> 00:28:07,232
But that's one.
555
00:28:08,052 --> 00:28:16,892
Definitely, if you're a cold card customer, be immediately spending time thinking about what to do and taking action.
556
00:28:16,892 --> 00:28:20,212
in those cases just to be clear
557
00:28:20,212 --> 00:28:21,472
we're talking about
558
00:28:21,472 --> 00:28:23,992
applying like creating a new
559
00:28:23,992 --> 00:28:26,072
seed phrase
560
00:28:26,072 --> 00:28:27,652
and applying entropy
561
00:28:27,652 --> 00:28:29,332
and we're talking about
562
00:28:29,332 --> 00:28:31,872
using a passphrase and I think
563
00:28:31,872 --> 00:28:34,052
the advice I've seen at least
564
00:28:34,052 --> 00:28:36,092
says you could do either or
565
00:28:36,092 --> 00:28:37,912
do you think it's fine to do either or
566
00:28:37,912 --> 00:28:39,172
do you think like both is
567
00:28:39,172 --> 00:28:41,952
well again the suite of options you don't have to do that
568
00:28:41,952 --> 00:28:43,892
you can just move your funds to River
569
00:28:43,892 --> 00:28:45,972
or Cash App or whatever you don't have to
570
00:28:45,972 --> 00:28:48,412
generate any of this complicated stuff,
571
00:28:48,692 --> 00:28:49,912
but there's trade-offs of that too,
572
00:28:50,092 --> 00:28:52,372
and we'll get to those. Max will be a
573
00:28:52,372 --> 00:28:54,352
proponent later, and we'll talk about the trade-offs there,
574
00:28:54,432 --> 00:28:55,312
but that's one option.
575
00:28:56,872 --> 00:28:57,592
Yeah, and then
576
00:28:57,592 --> 00:28:59,612
if you're choosing some other solution,
577
00:29:00,992 --> 00:29:02,252
I mean, there's all kinds of different products, right?
578
00:29:02,312 --> 00:29:04,412
BitKey, you don't have to deal with seed braises, right?
579
00:29:04,552 --> 00:29:06,252
So it's just a very
580
00:29:06,252 --> 00:29:07,592
slick, easy user experience.
581
00:29:08,372 --> 00:29:10,132
If you're going to reuse your ColdGuard product,
582
00:29:10,132 --> 00:29:11,732
which I think is what you're asking,
583
00:29:12,072 --> 00:29:13,532
where you are focusing on,
584
00:29:13,532 --> 00:29:14,612
then
585
00:29:14,612 --> 00:29:16,032
well yeah
586
00:29:16,032 --> 00:29:17,852
so if I had a cold card
587
00:29:17,852 --> 00:29:18,472
and I
588
00:29:18,472 --> 00:29:20,192
for whatever reason
589
00:29:20,192 --> 00:29:20,932
I didn't want to choose
590
00:29:20,932 --> 00:29:21,932
one of these other options
591
00:29:21,932 --> 00:29:23,212
I wanted to reuse my cold card
592
00:29:23,212 --> 00:29:24,852
I would install
593
00:29:24,852 --> 00:29:25,872
the new firmware
594
00:29:25,872 --> 00:29:26,992
which fixes the bug
595
00:29:26,992 --> 00:29:29,452
if it's
596
00:29:29,452 --> 00:29:30,472
if I have an MK3
597
00:29:30,472 --> 00:29:31,332
and it's not available
598
00:29:31,332 --> 00:29:32,912
well then I can't do that
599
00:29:32,912 --> 00:29:34,372
but I would
600
00:29:34,372 --> 00:29:35,552
then
601
00:29:35,552 --> 00:29:37,712
either roll dice
602
00:29:37,712 --> 00:29:38,692
or generate
603
00:29:38,692 --> 00:29:39,692
a passphrase
604
00:29:39,692 --> 00:29:40,652
definitely
605
00:29:40,652 --> 00:29:50,624
one of them you can do both But both is probably better I would do both Again I would do both But well I would
606
00:29:50,764 --> 00:29:54,364
if I had an MK3
607
00:29:54,364 --> 00:29:55,745
and didn't have the new firmware,
608
00:29:56,084 --> 00:29:57,804
so I know that it simply
609
00:29:57,804 --> 00:30:00,124
is not providing any security at all,
610
00:30:00,245 --> 00:30:01,364
I would definitely do both.
611
00:30:01,584 --> 00:30:03,844
Because even if I mess up my dice rolls
612
00:30:03,844 --> 00:30:05,784
or whatever, or I guess it,
613
00:30:05,784 --> 00:30:07,284
yeah, if I mess up the dice rolls,
614
00:30:07,505 --> 00:30:09,344
it's no worse than no security.
615
00:30:09,764 --> 00:30:10,584
So I would do both.
616
00:30:10,705 --> 00:30:15,084
If I installed the new firmware update that uses the hardware random number generator,
617
00:30:16,084 --> 00:30:19,864
then it gets debatable whether I dice roll.
618
00:30:20,764 --> 00:30:25,424
Because arguably it's better to use the hardware random number generator
619
00:30:25,424 --> 00:30:30,485
plus I can apply dice rolls for my passphrase.
620
00:30:31,225 --> 00:30:33,445
So then it's a combination of both.
621
00:30:34,205 --> 00:30:37,184
Actually, this gets back to your previous question too, like if you mix the two.
622
00:30:37,184 --> 00:30:46,764
So if I'm using, rolling the dice or generating my own entropy for the passphrase, then there's no way their software can mess with that.
623
00:30:48,945 --> 00:30:49,344
Right.
624
00:30:49,624 --> 00:30:49,764
Right.
625
00:30:50,024 --> 00:30:52,784
So then, I think that's a pretty solid approach.
626
00:30:52,884 --> 00:30:57,924
Yeah, and your answer, someone had asked that exact question in the chat, so whoever asked that, you get your question answered.
627
00:30:57,924 --> 00:31:06,445
I would just maybe add the caveat that, I mean, everything Steve is saying, I know you're trying to cover the kind of like extreme cases for someone that doesn't have other options necessarily available.
628
00:31:06,445 --> 00:31:11,264
but I would just like have the caveat of like this is like definitely doing all this is like if you
629
00:31:11,264 --> 00:31:15,664
really want to put in the time really understand what you're doing you would consider yourself a
630
00:31:15,664 --> 00:31:21,364
super pro user and even then I don't know like increasingly you know I know we'll talk more
631
00:31:21,364 --> 00:31:26,564
about this in the future of custody and what this means for everyone but like you know I think it's
632
00:31:26,564 --> 00:31:30,544
always worth evaluating does it make sense at least until you really have a good plan together
633
00:31:30,544 --> 00:31:34,304
to have someone that's like full-time thinking about this like at a river or a block just watch
634
00:31:34,304 --> 00:31:39,005
it for you in time because i uh i know i don't know i i wouldn't feel comfortable telling almost
635
00:31:39,005 --> 00:31:45,005
anyone i know to do any of this just full disclosure for um for those of you who don't
636
00:31:45,005 --> 00:31:52,924
tune in every week uh a year and a half ago january 2025 maybe like our our eighth episode
637
00:31:52,924 --> 00:31:58,784
or something of this show we covered self-custody the whole show right and we talked about different
638
00:31:58,784 --> 00:32:06,064
categories including using a custodian another category was using like a trezor ledger cold card
639
00:32:06,784 --> 00:32:12,384
and and just following the instructions like keeping it simple like what most human beings
640
00:32:12,384 --> 00:32:17,505
do just follow the instructions and use the product a third bucket was one of those products
641
00:32:17,505 --> 00:32:21,985
like a self-costity hardware wallet but you spend like a hundred hours or more
642
00:32:21,985 --> 00:32:26,364
with a do-it-yourself solution.
643
00:32:27,364 --> 00:32:30,764
Which definitely has the ability
644
00:32:30,764 --> 00:32:32,764
or the potential to be way more secure,
645
00:32:33,124 --> 00:32:35,645
but also introduces the potential for foot guns
646
00:32:35,645 --> 00:32:37,784
and requires the patience and diligence
647
00:32:37,784 --> 00:32:38,745
of 100 plus hours.
648
00:32:38,745 --> 00:32:40,445
And then the fourth category is BitKey.
649
00:32:40,764 --> 00:32:42,725
Because BitKey is such a novel design.
650
00:32:43,044 --> 00:32:43,965
So we walked through all that.
651
00:32:44,624 --> 00:32:47,024
And I mentioned this to David earlier,
652
00:32:47,184 --> 00:32:50,745
but sadly, the second category,
653
00:32:50,745 --> 00:32:54,044
which is using a product out of the,
654
00:32:54,404 --> 00:32:58,044
at least like the non-BitKey product out of the box,
655
00:32:58,644 --> 00:33:00,844
doing as you were told,
656
00:33:01,064 --> 00:33:03,705
it took a huge hit the past 24 hours, sadly.
657
00:33:04,745 --> 00:33:05,705
But in any case.
658
00:33:05,764 --> 00:33:07,804
And I would argue even that third category too.
659
00:33:08,064 --> 00:33:08,965
Like, I mean, I get it.
660
00:33:09,004 --> 00:33:11,264
There's people that, many people in the community
661
00:33:11,264 --> 00:33:13,245
that are obsessed and like spend all their time.
662
00:33:13,384 --> 00:33:15,465
And yeah, but I mean, people are still fallible.
663
00:33:15,624 --> 00:33:17,745
At the end of the day, I would just say,
664
00:33:17,864 --> 00:33:18,945
you know, I saw the meme yesterday.
665
00:33:18,945 --> 00:33:20,485
And obviously, look, I'm not saying that
666
00:33:20,485 --> 00:33:22,364
there's always going to be trade-offs.
667
00:33:22,485 --> 00:33:24,304
But if I were advising most people I know,
668
00:33:25,824 --> 00:33:27,024
the reality is,
669
00:33:27,584 --> 00:33:29,145
and we'll talk about the honeypots
670
00:33:29,145 --> 00:33:30,384
that these exchanges become too,
671
00:33:30,465 --> 00:33:31,624
but things like River,
672
00:33:32,304 --> 00:33:35,145
I know Leishman and his team is 24-7 obsessed with this.
673
00:33:35,225 --> 00:33:36,884
I know Block has the best engineers in the world
674
00:33:36,884 --> 00:33:37,564
obsessed with this.
675
00:33:37,965 --> 00:33:40,145
And even if I'm a really smart engineer individually,
676
00:33:40,145 --> 00:33:41,145
I'm still just one person.
677
00:33:41,245 --> 00:33:42,624
I know this is dangerous for self-custody
678
00:33:42,624 --> 00:33:44,424
and where we think about that in the future,
679
00:33:44,524 --> 00:33:45,284
which is why I think, by the way,
680
00:33:45,324 --> 00:33:47,404
things like BitKey are going to become increasingly important.
681
00:33:48,005 --> 00:33:49,284
But I don't know, man.
682
00:33:49,284 --> 00:34:10,284
And I think that the thing that just keeps coming to mind for me is no matter how smart you are as an individual, the AI is so much smarter and getting so much better, so much faster that, you know, even with like Ledger and Trezor, like, you know, how do we know there's not going to be firmware or some weird, you know, I guess with Trezor, it's a little bit harder because it's totally open source, which again, kind of pushes the beauty of things being totally open source.
683
00:34:10,284 --> 00:34:11,224
I know Ledger is not.
684
00:34:12,124 --> 00:34:13,865
More eyeballs, more AIs on it.
685
00:34:14,084 --> 00:34:15,365
But I don't know.
686
00:34:15,425 --> 00:34:16,925
It's like the meme that showed yesterday.
687
00:34:17,084 --> 00:34:18,044
It's like left side of the curve,
688
00:34:18,264 --> 00:34:19,665
right side of the curve, exchange.
689
00:34:19,845 --> 00:34:21,244
And like increasingly, you know,
690
00:34:21,284 --> 00:34:22,524
and obviously exchanges have gone under,
691
00:34:22,644 --> 00:34:25,304
but I do think the industry has changed a lot since that.
692
00:34:25,724 --> 00:34:28,224
Right now, Steve is making the Michael Saylor,
693
00:34:28,365 --> 00:34:29,865
Scottie Pippen look at you.
694
00:34:30,144 --> 00:34:30,205
Yeah.
695
00:34:30,524 --> 00:34:32,084
Well, no, I mean, I have a lot to say to that,
696
00:34:32,185 --> 00:34:33,325
but I wanted to, so.
697
00:34:33,764 --> 00:34:35,144
And again, I'm saying for people
698
00:34:35,144 --> 00:34:36,604
that probably have not put in the $100.
699
00:34:37,224 --> 00:34:39,084
Maybe even those people too, but certainly.
700
00:34:39,084 --> 00:34:41,544
I think the message we've shared, I think probably that January
701
00:34:41,544 --> 00:34:43,104
2025 was about it. Diversification.
702
00:34:43,385 --> 00:34:45,185
Custody diversification is
703
00:34:45,185 --> 00:34:47,464
the best choice for almost anybody. 100%.
704
00:34:47,464 --> 00:34:49,464
Like you should not say, I'm only doing it this
705
00:34:49,464 --> 00:34:51,365
way. Because there's just, every way
706
00:34:51,365 --> 00:34:53,284
has risks and put guns and there's always
707
00:34:53,284 --> 00:34:55,044
problems that could emerge. So if you
708
00:34:55,044 --> 00:34:57,185
really care about this, having
709
00:34:57,185 --> 00:34:59,284
multiple different ways that you custody
710
00:34:59,284 --> 00:35:01,264
it, so that if one goes wrong, you're still
711
00:35:01,264 --> 00:35:03,464
in the game. Right. And put simply,
712
00:35:04,124 --> 00:35:06,764
have two solutions,
713
00:35:07,325 --> 00:35:08,845
put half your sack in one,
714
00:35:09,084 --> 00:35:12,264
Well, just put simply, I'm keeping it super simple.
715
00:35:12,504 --> 00:35:12,644
Yeah.
716
00:35:13,425 --> 00:35:15,464
Have two solutions, put half your stack in one,
717
00:35:15,644 --> 00:35:16,464
half your stack in the other,
718
00:35:16,725 --> 00:35:18,024
and ideally those two solutions,
719
00:35:18,564 --> 00:35:21,304
the risks of each don't intersect.
720
00:35:21,464 --> 00:35:21,784
Exactly.
721
00:35:22,225 --> 00:35:23,304
That's the diversification.
722
00:35:23,405 --> 00:35:24,205
100% agree with that.
723
00:35:24,665 --> 00:35:28,284
Like having one be a cold card MK3 with no dice rolls
724
00:35:28,284 --> 00:35:30,385
and the other being a cold card MK3 with no dice rolls,
725
00:35:30,544 --> 00:35:31,604
that's not diversification.
726
00:35:31,604 --> 00:35:31,885
That's not diversification.
727
00:35:32,504 --> 00:35:35,804
So, you know, one type of diversification is like
728
00:35:35,804 --> 00:35:38,205
half of your stack is on a third-party custodian
729
00:35:38,205 --> 00:35:39,724
and half your stack is self-custody.
730
00:35:40,405 --> 00:35:42,304
Two very different risk profiles.
731
00:35:43,304 --> 00:35:44,064
And so,
732
00:35:45,165 --> 00:35:45,885
then
733
00:35:45,885 --> 00:35:47,984
it's just, and keep it
734
00:35:47,984 --> 00:35:49,604
simply again, like with probabilities.
735
00:35:50,205 --> 00:35:51,224
What's the probability
736
00:35:51,224 --> 00:35:53,624
that one of those solutions
737
00:35:53,624 --> 00:35:55,944
will fail?
738
00:35:56,624 --> 00:35:57,385
And, you know,
739
00:35:57,764 --> 00:36:00,064
it's impossible for us to know the number.
740
00:36:00,825 --> 00:36:01,944
But let's say
741
00:36:01,944 --> 00:36:04,064
it's a 1% chance
742
00:36:04,064 --> 00:36:05,185
that
743
00:36:05,185 --> 00:36:07,604
one of them will fail.
744
00:36:08,205 --> 00:36:09,724
Well, if you only use that one,
745
00:36:09,845 --> 00:36:11,644
then you have 1% chance of losing all your money.
746
00:36:12,244 --> 00:36:13,784
If you split it this way,
747
00:36:14,544 --> 00:36:17,205
and if Bitcoin moons to a million or more
748
00:36:17,205 --> 00:36:19,925
and it's everything you imagined and dreamt and believed
749
00:36:19,925 --> 00:36:22,084
and you're huddling, you're doing all the right things,
750
00:36:22,084 --> 00:36:22,944
you're huddling well off regardless.
751
00:36:23,925 --> 00:36:25,685
It's a lot better to have half your stack
752
00:36:25,685 --> 00:36:27,665
than 0% of your stack, right?
753
00:36:27,744 --> 00:36:29,165
And being out of the game is the worst.
754
00:36:29,444 --> 00:36:29,885
Being out of the game.
755
00:36:29,925 --> 00:36:30,665
You just don't want to be out.
756
00:36:30,905 --> 00:36:34,084
So I think that was our biggest piece of advice
757
00:36:34,084 --> 00:36:35,444
a year and a half ago on that show.
758
00:36:35,444 --> 00:36:37,905
And it would stand now as well.
759
00:36:38,205 --> 00:36:45,524
So in other words, people, I think, get wrapped up into the not your keys, not your coins mantra.
760
00:36:46,165 --> 00:36:47,944
And I definitely espouse that.
761
00:36:48,064 --> 00:36:52,084
I said at the beginning of the show, if self-custody dies, I think Bitcoin dies.
762
00:36:52,724 --> 00:36:53,984
Well, the option for self-custody.
763
00:36:54,345 --> 00:36:54,804
Yeah.
764
00:36:55,345 --> 00:37:04,564
Like a realistic user experience and option for people to use self-custody, if that dies, if culturally people aren't doing it, then I think Bitcoin dies.
765
00:37:04,564 --> 00:37:08,845
So I'm an enormous advocate of not your keys, not your coins, self-custody.
766
00:37:10,165 --> 00:37:17,264
But it would be an incredible outcome if 50% or more of people with Bitcoin
767
00:37:17,264 --> 00:37:20,564
held half their stack in self-custody and half with a custodian.
768
00:37:20,845 --> 00:37:25,325
If that's a state we achieve, that's an incredible state.
769
00:37:25,644 --> 00:37:29,444
Really healthy, really resilient, strong network, and that would be amazing.
770
00:37:29,444 --> 00:37:36,205
But I have a list of advice we should give for what people should do.
771
00:37:36,345 --> 00:37:37,244
I think it's important.
772
00:37:37,584 --> 00:37:39,144
One question that came up in the chat.
773
00:37:39,845 --> 00:37:41,784
So I was talking about the open source versus closed source.
774
00:37:41,905 --> 00:37:43,544
He's saying cold cars for closed source.
775
00:37:43,685 --> 00:37:44,925
I don't think that it isn't cold car.
776
00:37:45,024 --> 00:37:45,964
Wasn't it mostly open source?
777
00:37:46,024 --> 00:37:47,084
And this was open source as well?
778
00:37:47,665 --> 00:37:48,964
Yeah, it's open source.
779
00:37:49,284 --> 00:37:52,764
I think maybe the confusion here, there was drama in 2021,
780
00:37:52,944 --> 00:37:54,705
which actually led to this vulnerability,
781
00:37:54,944 --> 00:37:56,744
this code being introduced with the vulnerability,
782
00:37:56,744 --> 00:38:01,124
where originally the cold card software
783
00:38:01,124 --> 00:38:02,944
was free open source license.
784
00:38:03,524 --> 00:38:06,304
I don't know if it was MIT or it was a permissive license.
785
00:38:06,944 --> 00:38:08,564
And then Foundation Devices was founded.
786
00:38:09,104 --> 00:38:12,504
They used that software as they had the right to
787
00:38:12,504 --> 00:38:14,224
because it was free open source software.
788
00:38:15,084 --> 00:38:19,424
And CoinKite didn't like that action
789
00:38:19,424 --> 00:38:21,044
and then changed course
790
00:38:21,044 --> 00:38:25,104
so that they wanted to have a license
791
00:38:25,104 --> 00:38:26,244
which was more restrictive.
792
00:38:26,744 --> 00:38:34,264
And, you know, so if you're going to commercially use that software, then you'd have to go do a partnership or license from CoinKite.
793
00:38:34,264 --> 00:38:44,264
And as part of that, my understanding is that, well, they had to, CoinKite had to get rid of the code that was permissive.
794
00:38:45,064 --> 00:38:45,665
Oh, got it.
795
00:38:45,685 --> 00:38:47,504
And have code that was under a new license.
796
00:38:47,764 --> 00:38:48,084
Got it.
797
00:38:48,244 --> 00:38:52,504
And as part of that transition, that was a pretty big transition to a lot of new code.
798
00:38:52,504 --> 00:38:53,325
and it introduced,
799
00:38:53,865 --> 00:38:54,385
or, you know,
800
00:38:54,544 --> 00:38:55,724
during that transition,
801
00:38:55,885 --> 00:38:56,325
this vulnerability
802
00:38:56,325 --> 00:38:56,885
was introduced.
803
00:38:57,144 --> 00:38:57,964
That's my understanding
804
00:38:57,964 --> 00:38:58,784
of that timeline.
805
00:38:59,685 --> 00:39:00,424
But yeah,
806
00:39:00,584 --> 00:39:02,604
I believe ColdCard
807
00:39:02,604 --> 00:39:03,124
is open,
808
00:39:03,304 --> 00:39:04,405
so it's open source,
809
00:39:04,544 --> 00:39:05,365
but like, you know,
810
00:39:06,665 --> 00:39:07,905
it's not free open source
811
00:39:07,905 --> 00:39:09,764
and people can bicker
812
00:39:09,764 --> 00:39:10,144
and debate
813
00:39:10,144 --> 00:39:11,104
whether the term
814
00:39:11,104 --> 00:39:11,984
open source
815
00:39:11,984 --> 00:39:13,524
implies free open source,
816
00:39:13,644 --> 00:39:14,264
but it is
817
00:39:14,264 --> 00:39:15,604
unquestionably
818
00:39:15,604 --> 00:39:16,604
published source.
819
00:39:16,604 --> 00:39:17,024
It's open source,
820
00:39:17,124 --> 00:39:17,804
so it's auditable.
821
00:39:18,205 --> 00:39:19,504
So free open source,
822
00:39:19,784 --> 00:39:20,444
no one debates.
823
00:39:20,744 --> 00:39:22,284
That means it's permissive.
824
00:39:22,504 --> 00:39:24,865
published source, no one debates.
825
00:39:24,964 --> 00:39:26,944
That means it's not permissive or some kind of
826
00:39:26,944 --> 00:39:28,845
restrictions. Then the term open source,
827
00:39:28,964 --> 00:39:30,345
you can debate what it implies.
828
00:39:31,144 --> 00:39:32,365
And they went from being
829
00:39:32,365 --> 00:39:34,405
full open source
830
00:39:34,405 --> 00:39:36,024
to published.
831
00:39:36,544 --> 00:39:38,825
So you can audit it, you can confirm that it
832
00:39:38,825 --> 00:39:40,804
does what it claims to do, but you can't
833
00:39:40,804 --> 00:39:42,205
use it to build a competitive product.
834
00:39:42,405 --> 00:39:44,644
Is that roughly why such a thing would
835
00:39:44,644 --> 00:39:45,744
be built that way?
836
00:39:46,444 --> 00:39:48,524
Yes. And it's not like
837
00:39:48,524 --> 00:39:50,644
BitKey initially was launched
838
00:39:50,644 --> 00:39:52,484
with published source with a restrictive license.
839
00:39:52,504 --> 00:39:57,084
and then later changed that to be MIT license.
840
00:39:57,205 --> 00:39:58,165
Oh, I didn't realize that.
841
00:39:58,345 --> 00:40:00,685
All the keys code is false?
842
00:40:01,424 --> 00:40:02,865
Wow, that's awesome.
843
00:40:03,244 --> 00:40:07,665
That was always the spirit and intention.
844
00:40:08,244 --> 00:40:09,004
I think there's just like,
845
00:40:09,284 --> 00:40:10,424
it's easy to go,
846
00:40:10,845 --> 00:40:12,405
it's like a one-way door, right?
847
00:40:12,405 --> 00:40:14,865
So it's easy to do what they ended up doing.
848
00:40:16,165 --> 00:40:19,564
And it's, well, as cold cards,
849
00:40:19,724 --> 00:40:20,544
the perfect example,
850
00:40:20,685 --> 00:40:22,024
it's really hard to go the other way.
851
00:40:22,024 --> 00:40:28,225
In fact, that's a very clear example of the risks of going the other direction.
852
00:40:29,385 --> 00:40:32,665
So I think there was just like a cautious move by the big key team.
853
00:40:32,685 --> 00:40:36,225
To me, the other sort of big meta point, and I know we'll talk more about the future in all this,
854
00:40:36,385 --> 00:40:40,744
but is the beauty and theory of open source is, what is it?
855
00:40:40,825 --> 00:40:42,365
All bugs are shallow with enough eyeballs.
856
00:40:42,784 --> 00:40:45,405
I mean, increasingly, that's going to be all bugs are shallow with enough tokens, right?
857
00:40:45,405 --> 00:40:57,484
And so the thing I keep coming back to is, as a user, the metric I actually am interested in for software I use, like critical software like this, I want to know who has audited this for how long.
858
00:40:57,544 --> 00:41:00,845
Because I know, I believe they did an audit with, I don't know what level of Frontier model.
859
00:41:00,944 --> 00:41:01,504
Maybe it was Kimmy.
860
00:41:01,584 --> 00:41:02,825
I don't know if it was one of the closed source models.
861
00:41:03,325 --> 00:41:03,784
Who did an audit?
862
00:41:03,964 --> 00:41:06,425
I think Cold Car did of this release.
863
00:41:06,504 --> 00:41:06,845
I'm assuming.
864
00:41:07,144 --> 00:41:08,064
I think I read somewhere.
865
00:41:08,624 --> 00:41:12,104
I read it maybe on Stacker News in a thread that they had had some audit with a model.
866
00:41:12,185 --> 00:41:12,885
I don't know what model.
867
00:41:12,885 --> 00:41:17,325
I don't know how long their last release to have this or whatever release.
868
00:41:17,905 --> 00:41:19,304
But my, and so I don't know that.
869
00:41:19,405 --> 00:41:20,405
That's what I wrote on Stacker News.
870
00:41:20,984 --> 00:41:21,865
No inside knowledge there.
871
00:41:22,325 --> 00:41:28,284
But what I'm realizing, a response in that thread in Stacker News,
872
00:41:28,385 --> 00:41:31,385
actually from now, I believe, which was an excellent response,
873
00:41:31,644 --> 00:41:35,424
was we're getting to the point now where the state of software is no longer,
874
00:41:35,885 --> 00:41:37,244
hey, I did an audit, ship.
875
00:41:37,705 --> 00:41:41,845
It's like the attackers are now running instances of the latest software.
876
00:41:41,845 --> 00:41:43,764
And you don't know if it's someone from inside a company.
877
00:41:43,845 --> 00:41:46,604
I'm not saying they would, but like that's theoretically possible at the very cutting
878
00:41:46,604 --> 00:41:46,884
edge.
879
00:41:47,084 --> 00:41:50,484
And they're doing it 24 seven and like recursively improving their prompts.
880
00:41:50,764 --> 00:41:54,644
And so the thing that I'm honestly interested in, and I think this is the much bigger story
881
00:41:54,644 --> 00:41:58,564
than an individual hack, because this can go so much bigger for everything in the industry
882
00:41:58,564 --> 00:42:02,304
is, you know, I know we have project loop and maybe loop needs to be able to do this
883
00:42:02,304 --> 00:42:07,584
sort of like verifiably is I want to see for every code base I use, is there a top of the
884
00:42:07,584 --> 00:42:09,384
line model that's running 24 seven?
885
00:42:09,384 --> 00:42:11,185
And like how many instances are there?
886
00:42:11,185 --> 00:42:13,385
because I think the only way to scale with this,
887
00:42:13,464 --> 00:42:14,385
I think the good news is,
888
00:42:14,424 --> 00:42:15,464
as we've said before on the show,
889
00:42:15,584 --> 00:42:17,205
is that we're going through the transition phase
890
00:42:17,205 --> 00:42:18,304
from sort of humans to machines,
891
00:42:18,604 --> 00:42:20,024
and it's not always going to be this bad.
892
00:42:20,444 --> 00:42:21,584
But during this transition,
893
00:42:21,705 --> 00:42:22,325
I actually think a lot,
894
00:42:22,385 --> 00:42:24,024
there was a beautiful talk from Drew Bansel,
895
00:42:24,084 --> 00:42:25,744
who's done many beautiful talks over the years,
896
00:42:26,044 --> 00:42:27,885
talking about how one of the biggest shifts from AI
897
00:42:27,885 --> 00:42:30,264
was going to basically be AI starting to hack
898
00:42:30,264 --> 00:42:31,464
more software with Bitcoin attachments.
899
00:42:31,665 --> 00:42:32,724
That's obviously economic incentive.
900
00:42:33,084 --> 00:42:34,764
And that's how like real software,
901
00:42:34,885 --> 00:42:36,345
the age of real software would begin
902
00:42:36,345 --> 00:42:38,325
is once, you know, sort of AI is hardened
903
00:42:38,325 --> 00:42:40,524
and Bitcoin is the incentive for that.
904
00:42:40,524 --> 00:42:43,064
but I want to see like some metric.
905
00:42:43,284 --> 00:42:45,144
I want to see like for any product I use,
906
00:42:45,225 --> 00:42:46,284
like I'm thinking right now,
907
00:42:46,304 --> 00:42:48,325
like if you're using a Trezor or a Ledger or whatever,
908
00:42:48,425 --> 00:42:51,144
like are they running top of the line models 24 seven?
909
00:42:51,264 --> 00:42:53,385
Do they have people that understand
910
00:42:53,385 --> 00:42:54,464
what the attackers are doing?
911
00:42:54,845 --> 00:42:56,624
And if not, then I don't feel comfortable.
912
00:42:57,004 --> 00:42:58,804
So I don't know if there's some way to open source that
913
00:42:58,804 --> 00:43:01,185
and like literally show kind of like wallet scrutiny.
914
00:43:01,304 --> 00:43:03,244
I think way more important than like old school scrutiny
915
00:43:03,244 --> 00:43:04,885
is, you know, who's running,
916
00:43:05,225 --> 00:43:08,464
who is the maestro organizing models
917
00:43:08,464 --> 00:43:09,865
to validate these things?
918
00:43:09,865 --> 00:43:11,444
how many instances are they running?
919
00:43:11,564 --> 00:43:12,584
How frequently is it running?
920
00:43:12,784 --> 00:43:14,244
Because I know the attackers are 24-7.
921
00:43:15,165 --> 00:43:17,984
So I can share some things we've learned with Project Loop.
922
00:43:18,104 --> 00:43:18,244
Yeah.
923
00:43:18,944 --> 00:43:21,744
Which makes what you're wanting complicated,
924
00:43:22,004 --> 00:43:24,064
but I 100% agree with what you're wanting.
925
00:43:24,205 --> 00:43:25,865
And definitely we're thinking through,
926
00:43:26,004 --> 00:43:27,084
can we achieve that?
927
00:43:28,144 --> 00:43:32,004
But the first point I'll make is around open source.
928
00:43:32,205 --> 00:43:32,284
Yeah.
929
00:43:32,804 --> 00:43:36,185
If code is not open, if it's not available,
930
00:43:36,424 --> 00:43:37,484
if it's not source available,
931
00:43:37,484 --> 00:43:39,064
then obviously something like Project Loop
932
00:43:39,064 --> 00:43:42,604
can't even help because Project Loop would not have.
933
00:43:42,804 --> 00:43:46,264
And just so people know, if you're not familiar with Project Loop, L-O-U-P-E,
934
00:43:46,624 --> 00:43:51,624
Spiral launched it a few months ago with the intention of doing AI security scans
935
00:43:51,624 --> 00:43:56,744
for open source Bitcoin repositories.
936
00:43:56,744 --> 00:44:01,524
We started with a set of eight projects for the initial wave.
937
00:44:02,325 --> 00:44:03,964
We've completed that first phase.
938
00:44:04,064 --> 00:44:07,044
Now we've welcomed another 10 or so projects,
939
00:44:07,044 --> 00:44:09,685
and we're doing regular security scans there.
940
00:44:10,284 --> 00:44:12,264
But the source has to be available.
941
00:44:12,744 --> 00:44:14,204
Now, if it's free open source software,
942
00:44:14,204 --> 00:44:18,424
I would argue that it's better for security
943
00:44:18,424 --> 00:44:21,064
because it's more likely other people
944
00:44:21,064 --> 00:44:23,124
will be adopting it and using it in their products.
945
00:44:23,724 --> 00:44:25,044
And you need people to care.
946
00:44:25,345 --> 00:44:29,165
The more people, companies, economic value
947
00:44:29,165 --> 00:44:32,304
cares about a piece of software,
948
00:44:32,825 --> 00:44:43,797
the more likely someone going to run an audit against it And someone going to look at the audit output which is what I want to get to Because even if let say Project Loop guaranteed
949
00:44:43,797 --> 00:44:47,877
or could verifiably say we audited the software,
950
00:44:48,517 --> 00:44:51,157
that alone, I mean, that should make you feel better
951
00:44:51,157 --> 00:44:52,217
than if you don't see that,
952
00:44:52,297 --> 00:44:54,557
but that shouldn't make you feel super warm and fuzzy.
953
00:44:54,757 --> 00:44:54,936
I agree.
954
00:44:55,097 --> 00:44:56,837
Because what Project Loop's doing
955
00:44:56,837 --> 00:45:00,077
is it's generating a bunch of GitHub issues
956
00:45:00,077 --> 00:45:03,277
and a private repo for the project
957
00:45:03,277 --> 00:45:08,737
So just the security people or maintainers of a project see the results.
958
00:45:08,877 --> 00:45:12,117
Like I don't even, I don't see the results even though I'm working on Project Loop.
959
00:45:14,377 --> 00:45:19,597
And for some projects that's dozens of issues being opened.
960
00:45:19,757 --> 00:45:21,917
Others is hundreds of issues being opened.
961
00:45:22,397 --> 00:45:23,717
So what's done with those issues?
962
00:45:23,777 --> 00:45:27,877
If nothing's done with them, then it didn't get any more secure.
963
00:45:27,877 --> 00:45:35,337
Or if some work was done and some fixes were done, but there was a critical vulnerability that slipped through, it's still problematic.
964
00:45:35,697 --> 00:45:42,617
So does Project Loop publish either the models used and or the prompts used?
965
00:45:42,697 --> 00:45:48,097
Or is that considered something that the like the harness and the prompts is all open source.
966
00:45:48,197 --> 00:45:48,956
Anyone can run it.
967
00:45:49,137 --> 00:45:50,137
And it's generic prompts.
968
00:45:50,377 --> 00:45:52,377
There's not specialized knowledge in the prompts.
969
00:45:52,797 --> 00:45:57,497
The AI security scanning is good enough that that works.
970
00:45:57,877 --> 00:46:15,597
That said, like if you know, like with the cold card incident, when you know the specific vulnerability and you know that and now you want to go test like Trezor, Ledger, BitKey, other products, like prompting it very specifically to focus on that can radically shrink the time and cost to find it.
971
00:46:16,157 --> 00:46:20,837
But with Project Lube, we don't go in knowing that, oh, this is where the bug is.
972
00:46:21,077 --> 00:46:22,257
We don't know where the bugs are.
973
00:46:22,797 --> 00:46:24,377
So it's a generic prompt.
974
00:46:24,577 --> 00:46:25,456
That's all open source.
975
00:46:25,456 --> 00:46:28,577
the models used is not published.
976
00:46:29,397 --> 00:46:30,877
And I'll just say that
977
00:46:30,877 --> 00:46:34,577
the project tries to get access
978
00:46:34,577 --> 00:46:36,277
to the most powerful models possible
979
00:46:36,277 --> 00:46:38,317
because we want to do as good a job as possible
980
00:46:38,317 --> 00:46:39,897
and help projects become,
981
00:46:40,577 --> 00:46:42,717
you know, we're the good guys.
982
00:46:43,037 --> 00:46:43,697
We're not trying to attack these projects.
983
00:46:43,697 --> 00:46:44,657
It almost seems like every time
984
00:46:44,657 --> 00:46:47,077
a new high-powered model comes out,
985
00:46:47,417 --> 00:46:49,117
it may be able to do...
986
00:46:49,117 --> 00:46:49,817
Yeah, crazy new stuff.
987
00:46:49,817 --> 00:46:51,657
It almost seems like you might need a new audit,
988
00:46:52,377 --> 00:46:53,017
a new project loop.
989
00:46:53,017 --> 00:46:55,897
I would even push further on that.
990
00:46:56,057 --> 00:46:58,097
To me, the big takeaway from that now comment was,
991
00:46:58,197 --> 00:47:00,057
I don't even think it should be dynamic process.
992
00:47:00,456 --> 00:47:01,977
Like, I think you should be able to see in real time
993
00:47:01,977 --> 00:47:03,257
for things that are important enough.
994
00:47:03,537 --> 00:47:05,037
It's not like, oh, there's a new model, new audit.
995
00:47:05,157 --> 00:47:07,677
It's like, no, I want to see 24-7, 365
996
00:47:07,677 --> 00:47:10,097
continually crunching on the prompt.
997
00:47:10,097 --> 00:47:13,117
So one thing is, or you mentioned,
998
00:47:13,217 --> 00:47:15,877
you saw that like CoinKite ran a scan or not.
999
00:47:15,936 --> 00:47:18,157
I didn't see that, but I don't know if they did or did not.
1000
00:47:18,657 --> 00:47:21,837
I do know, I have heard others did scan.
1001
00:47:22,057 --> 00:47:22,277
Yeah.
1002
00:47:23,017 --> 00:47:24,297
cold card firmware
1003
00:47:24,297 --> 00:47:26,157
earlier
1004
00:47:26,157 --> 00:47:27,637
this year.
1005
00:47:27,817 --> 00:47:29,737
And it did not find this bug
1006
00:47:29,737 --> 00:47:31,537
with a less powerful model.
1007
00:47:31,797 --> 00:47:33,157
Yeah, with a less powerful model.
1008
00:47:33,177 --> 00:47:34,117
With a less powerful model.
1009
00:47:34,157 --> 00:47:34,617
That's the issue.
1010
00:47:34,916 --> 00:47:35,337
And now,
1011
00:47:35,557 --> 00:47:36,817
I know many people,
1012
00:47:36,956 --> 00:47:38,637
like Instagibs on my own team
1013
00:47:38,637 --> 00:47:40,517
reproduced this yesterday.
1014
00:47:41,117 --> 00:47:43,416
It took an hour and 50 minutes
1015
00:47:43,416 --> 00:47:43,857
he published.
1016
00:47:44,177 --> 00:47:45,197
Like from the time he's like,
1017
00:47:45,197 --> 00:47:46,677
I wonder if I should try
1018
00:47:46,677 --> 00:47:47,737
to look into this.
1019
00:47:48,217 --> 00:47:49,597
Hour and 50 minutes later,
1020
00:47:50,297 --> 00:47:50,517
he,
1021
00:47:50,977 --> 00:47:52,697
and he also had an MK3.
1022
00:47:53,017 --> 00:47:54,577
hardware, actual hardware himself.
1023
00:47:54,897 --> 00:47:57,037
And he completely reproduced the vulnerability
1024
00:47:57,037 --> 00:47:58,956
and was able to generate
1025
00:47:58,956 --> 00:48:00,097
24 words.
1026
00:48:01,456 --> 00:48:02,956
And was that chat GPT
1027
00:48:02,956 --> 00:48:04,857
5.6 Sol or Fable 5 or
1028
00:48:04,857 --> 00:48:06,697
something that just came out in the last few weeks?
1029
00:48:07,537 --> 00:48:08,197
It's a
1030
00:48:08,197 --> 00:48:10,597
newer model.
1031
00:48:11,857 --> 00:48:12,797
Newer than those?
1032
00:48:13,897 --> 00:48:14,497
Private access?
1033
00:48:14,916 --> 00:48:15,557
I can't say.
1034
00:48:16,757 --> 00:48:18,916
But this is something that I do, I know
1035
00:48:18,916 --> 00:48:20,837
you can't say this now, but like, especially
1036
00:48:20,837 --> 00:48:22,597
for the exchanges also, like seriously,
1037
00:48:22,597 --> 00:48:31,557
I want to see specifically, I would use an exchange that says I got access to GPC6 or Mythos competitor before anyone else over anything else.
1038
00:48:31,817 --> 00:48:32,936
That'd be my number one requirement.
1039
00:48:33,157 --> 00:48:35,657
But then they might lose access to that model.
1040
00:48:37,057 --> 00:48:40,977
Then these are going to be a hard thing to do.
1041
00:48:40,977 --> 00:48:47,277
So what we're seeing so far, these models,
1042
00:48:47,697 --> 00:48:52,177
obviously everyone's seeing that they can do magical stuff
1043
00:48:52,177 --> 00:48:55,477
and find bugs that humans haven't found for many years or decades.
1044
00:48:56,737 --> 00:48:59,877
But that doesn't mean, and also generic prompts work.
1045
00:49:00,537 --> 00:49:04,537
But that doesn't mean that it's perfectly spitting out,
1046
00:49:05,157 --> 00:49:08,037
like there's three vulnerabilities.
1047
00:49:08,037 --> 00:49:10,637
One's critical, one's medium, one's low.
1048
00:49:10,637 --> 00:49:11,297
and then that's it.
1049
00:49:11,697 --> 00:49:13,377
It's like there's a lot of noise
1050
00:49:13,377 --> 00:49:16,217
and it'll misclassify the severity both ways.
1051
00:49:16,557 --> 00:49:18,217
It'll say some things are high when they're low.
1052
00:49:18,337 --> 00:49:20,377
It'll say some things are medium when they're severe.
1053
00:49:20,936 --> 00:49:23,597
So it requires humans to look at the output still
1054
00:49:23,597 --> 00:49:26,657
and it requires humans who understand the code base.
1055
00:49:26,997 --> 00:49:29,557
That requires attention and people who care
1056
00:49:29,557 --> 00:49:31,456
and people who are prioritizing the stuff,
1057
00:49:31,916 --> 00:49:37,217
which gets back to the more people
1058
00:49:37,217 --> 00:49:38,416
that are using a piece of software,
1059
00:49:38,416 --> 00:49:40,857
it's more likely to get more attention.
1060
00:49:41,557 --> 00:49:42,537
So that's still important.
1061
00:49:42,936 --> 00:49:44,317
I mean, the models are going to keep getting better.
1062
00:49:44,697 --> 00:49:46,157
So, I mean, it wouldn't surprise me
1063
00:49:46,157 --> 00:49:48,017
if by the end of the year or a year from now,
1064
00:49:48,377 --> 00:49:50,677
it will be, it spits out the,
1065
00:49:51,157 --> 00:49:51,916
oh, because other things,
1066
00:49:51,997 --> 00:49:53,357
it'll find legitimate bugs,
1067
00:49:53,577 --> 00:49:55,177
but they're not security vulnerabilities, right?
1068
00:49:55,297 --> 00:49:57,617
So among these like dozens or hundreds of issues,
1069
00:49:58,157 --> 00:50:00,977
it's actually, it's pretty accurate in all of them.
1070
00:50:01,137 --> 00:50:02,637
Like they're all like issues,
1071
00:50:02,637 --> 00:50:03,977
but sometimes it's like, well,
1072
00:50:04,097 --> 00:50:06,037
but actually the compiled,
1073
00:50:06,297 --> 00:50:07,877
that's only if you compile it this way
1074
00:50:07,877 --> 00:50:10,157
and that's for debug or no one does that
1075
00:50:10,157 --> 00:50:11,137
or things like that.
1076
00:50:11,537 --> 00:50:14,197
In any case, the other thing I was going to say,
1077
00:50:14,257 --> 00:50:16,977
you're saying like 24-7, 365 or whatever.
1078
00:50:19,057 --> 00:50:20,137
For important enough.
1079
00:50:20,237 --> 00:50:20,657
Well, that's it.
1080
00:50:20,797 --> 00:50:23,057
So another thing we're learning with Project Loop
1081
00:50:23,057 --> 00:50:27,617
is that it costs money to run these models.
1082
00:50:27,777 --> 00:50:27,877
Right.
1083
00:50:28,537 --> 00:50:30,517
And we're also learning that different models
1084
00:50:30,517 --> 00:50:32,037
cost a different amount of money
1085
00:50:32,037 --> 00:50:33,557
and some cost a lot of money.
1086
00:50:34,117 --> 00:50:37,717
So we're having to establish tiers
1087
00:50:37,717 --> 00:50:40,237
and frequency of rescans
1088
00:50:40,237 --> 00:50:41,637
and also what you're rescanning.
1089
00:50:41,757 --> 00:50:43,917
So like, are you rescanning the entire code base
1090
00:50:43,917 --> 00:50:45,297
every hour?
1091
00:50:45,697 --> 00:50:47,517
Or, you know, are you only
1092
00:50:47,517 --> 00:50:49,997
scanning like new code?
1093
00:50:50,237 --> 00:50:52,257
So there's all these variables we can play with
1094
00:50:52,257 --> 00:50:54,057
to make it so that
1095
00:50:54,057 --> 00:50:55,597
the quality is still high
1096
00:50:55,597 --> 00:50:57,977
but it's like practical
1097
00:50:57,977 --> 00:50:59,977
economically. There's also a lot of like
1098
00:50:59,977 --> 00:51:01,777
third-party library inclusion
1099
00:51:01,777 --> 00:51:03,057
in software, right?
1100
00:51:03,837 --> 00:51:05,857
And so there might be things that are
1101
00:51:05,857 --> 00:51:07,577
you kind of like accidentally include
1102
00:51:07,577 --> 00:51:09,436
is like, oh, how do you compute thing X?
1103
00:51:09,557 --> 00:51:11,377
Oh, there's some library that computes thing X.
1104
00:51:11,477 --> 00:51:12,397
I'll just include it.
1105
00:51:12,456 --> 00:51:13,397
You trust it kind of works.
1106
00:51:13,897 --> 00:51:14,777
So being able to figure out
1107
00:51:14,777 --> 00:51:17,757
how much of those libraries get included
1108
00:51:17,757 --> 00:51:20,057
and should those get their own loop projects.
1109
00:51:21,277 --> 00:51:23,037
The whole way software works.
1110
00:51:23,097 --> 00:51:23,897
Well, and to be clear,
1111
00:51:24,057 --> 00:51:25,557
I mean, as much as the loop thing is amazing,
1112
00:51:25,677 --> 00:51:27,477
I really think you guys can play
1113
00:51:27,477 --> 00:51:30,857
a world-changing role through this transition.
1114
00:51:31,477 --> 00:51:33,297
But also, there's also the economic actors.
1115
00:51:33,436 --> 00:51:34,416
I want to see,
1116
00:51:34,557 --> 00:51:35,916
and I know we're going to talk about this as well,
1117
00:51:36,037 --> 00:51:36,637
but Coinbase,
1118
00:51:37,357 --> 00:51:38,337
Those guys have a lot of money.
1119
00:51:38,416 --> 00:51:38,936
Public company.
1120
00:51:39,217 --> 00:51:40,456
How much is being spent right now?
1121
00:51:40,517 --> 00:51:42,557
Are they getting access to GPT-6
1122
00:51:42,557 --> 00:51:44,657
or whatever Fable's successor is?
1123
00:51:44,677 --> 00:51:46,177
And if so, I would hope they're running
1124
00:51:46,177 --> 00:51:49,577
10,000 instances of these models 24-7.
1125
00:51:49,777 --> 00:51:50,477
I bet they are too,
1126
00:51:50,717 --> 00:51:52,097
but it would be cool to know that.
1127
00:51:52,237 --> 00:51:52,297
Right.
1128
00:51:53,037 --> 00:51:54,277
Yeah, they might not be able to say that.
1129
00:51:54,357 --> 00:51:54,757
Exactly.
1130
00:51:54,757 --> 00:51:56,077
But then it's like,
1131
00:51:56,377 --> 00:51:58,117
so one thing that I come in,
1132
00:51:59,097 --> 00:52:01,737
part of the spirit of self-custody,
1133
00:52:01,817 --> 00:52:03,257
it's not just the self-custody piece.
1134
00:52:03,677 --> 00:52:05,477
I think to me what's very important,
1135
00:52:05,477 --> 00:52:07,777
obviously the right to self-custody, but also is transparency.
1136
00:52:08,377 --> 00:52:10,237
I think one of the most valuable things that we could do
1137
00:52:10,237 --> 00:52:12,817
or could be done in general for individuals
1138
00:52:12,817 --> 00:52:14,897
is just be able to know, okay,
1139
00:52:14,956 --> 00:52:16,637
of all of your different options out there,
1140
00:52:16,757 --> 00:52:19,377
who's cutting edge getting this and who's not?
1141
00:52:19,477 --> 00:52:21,657
Because I really think it's that black and white.
1142
00:52:21,797 --> 00:52:22,877
Maybe they can't tell us,
1143
00:52:22,877 --> 00:52:27,777
but then that kind of puts the end user in not a great position.
1144
00:52:29,817 --> 00:52:32,317
We're an hour in, and I really want to get to advice.
1145
00:52:32,317 --> 00:52:32,877
Already, wow.
1146
00:52:32,956 --> 00:52:33,757
Everything we're talking about is great.
1147
00:52:33,757 --> 00:52:44,877
But just advice for people, because we already gave some advice that we've learned and makes sense to us, I think, for people who use a cold card themselves.
1148
00:52:45,497 --> 00:52:51,916
But I think I have advice for everyone in Bitcoin who cares about Bitcoin, who holds the Bitcoin or believes in its future.
1149
00:52:52,237 --> 00:52:55,217
Every single person should take action right now.
1150
00:52:55,217 --> 00:53:05,377
And that action is that if you know someone who uses a cold card, or even if you just know someone who has Bitcoin, reach out to them.
1151
00:53:06,097 --> 00:53:16,517
Obviously, if they have a cold card, reach out to your family members, your friends, your coins immediately, and then collect the most reliable information you can find and share with them.
1152
00:53:17,817 --> 00:53:22,456
But if you have friends that you don't know what they use for storing their Bitcoin, ask them.
1153
00:53:22,477 --> 00:53:23,477
To make sure they're aware, yeah.
1154
00:53:23,477 --> 00:53:34,896
Because while I've been on Twitter constantly the last 24 hours, even you guys haven't, but you're aware, but there's a lot of people who live a life, don't work in the space full time.
1155
00:53:35,157 --> 00:53:37,277
And they're on summer vacation or something, right?
1156
00:53:37,317 --> 00:53:38,837
And they are at risk right now.
1157
00:53:38,936 --> 00:53:41,017
So reach out to your friends and family members.
1158
00:53:41,317 --> 00:53:42,817
Everyone can do that.
1159
00:53:42,997 --> 00:53:44,396
So I think that's really important.
1160
00:53:45,677 --> 00:53:52,837
So, I mean, anyone listening to this show, you're probably equipped to go find the one, two, or three credible pieces of information.
1161
00:53:52,837 --> 00:53:59,416
I mean, you can certainly, like, what CoinKite has published and what Block has published, I think is highly credible and reliable.
1162
00:54:00,396 --> 00:54:03,677
So, you know, you can share that with people.
1163
00:54:10,717 --> 00:54:11,197
Sorry.
1164
00:54:11,857 --> 00:54:13,077
I have my notes here.
1165
00:54:13,416 --> 00:54:14,456
A lot of information to share.
1166
00:54:15,717 --> 00:54:16,757
We got...
1167
00:54:16,757 --> 00:54:18,337
Oh, yeah, sorry.
1168
00:54:18,456 --> 00:54:18,757
Go ahead.
1169
00:54:19,297 --> 00:54:19,737
Do you mind?
1170
00:54:19,737 --> 00:54:22,477
some people listening
1171
00:54:22,477 --> 00:54:25,037
or people listening that you know folks
1172
00:54:25,037 --> 00:54:27,997
I mean if you are like a security researcher
1173
00:54:27,997 --> 00:54:31,396
or you nerd out on like blockchain analytics
1174
00:54:31,396 --> 00:54:32,817
whatever I mean if you
1175
00:54:32,817 --> 00:54:36,617
or you're toying with your AI LLM
1176
00:54:36,617 --> 00:54:39,977
if you do any kind of analysis that you think would be
1177
00:54:39,977 --> 00:54:42,977
valuable that's delivering new
1178
00:54:42,977 --> 00:54:45,396
technical information about this exploit that
1179
00:54:45,396 --> 00:54:48,857
you haven't seen revealed yet from CoinKite or Block
1180
00:54:48,857 --> 00:54:51,717
or anyone else, please report that.
1181
00:54:52,477 --> 00:54:54,936
Security at Coinkite.com is their email.
1182
00:54:55,837 --> 00:54:59,857
And NVK has told me that's the place people should share information
1183
00:54:59,857 --> 00:55:04,277
because they're getting DMs from, they're being flooded with DMs.
1184
00:55:04,357 --> 00:55:09,277
So the best resource to contact Coinkite is security at Coinkite.com.
1185
00:55:10,436 --> 00:55:14,277
Block is doing a tremendous amount of analysis.
1186
00:55:14,916 --> 00:55:16,436
And we haven't really talked about that yet,
1187
00:55:16,436 --> 00:55:19,197
but I had a firsthand view
1188
00:55:19,197 --> 00:55:20,577
at everything done yesterday.
1189
00:55:21,477 --> 00:55:23,657
I already knew the block security team
1190
00:55:23,657 --> 00:55:24,577
was A+.
1191
00:55:24,577 --> 00:55:26,597
Because several of them,
1192
00:55:26,717 --> 00:55:27,956
I've worked with it for seven years now.
1193
00:55:28,077 --> 00:55:29,936
So I know how good they are.
1194
00:55:30,157 --> 00:55:31,797
But yesterday was a shiny example.
1195
00:55:33,337 --> 00:55:35,197
Incredibly intelligent, knowledgeable.
1196
00:55:36,057 --> 00:55:37,117
They really care.
1197
00:55:37,697 --> 00:55:39,217
This isn't blocked product, right?
1198
00:55:39,977 --> 00:55:41,897
And block didn't do this
1199
00:55:41,897 --> 00:55:44,257
as like a gotcha to a competitor
1200
00:55:44,257 --> 00:55:45,497
because BitKey and ColdGuard
1201
00:55:45,497 --> 00:55:46,577
are somewhat competitors.
1202
00:55:47,416 --> 00:55:49,537
Block cares about Bitcoin,
1203
00:55:49,777 --> 00:55:50,577
cares about the people
1204
00:55:50,577 --> 00:55:51,277
who are affected,
1205
00:55:53,117 --> 00:55:55,477
including the Block security engineers.
1206
00:55:55,877 --> 00:55:57,097
Also, a couple engineers
1207
00:55:57,097 --> 00:55:58,217
from Spiral as well
1208
00:55:58,217 --> 00:55:59,117
put a bunch of time
1209
00:55:59,117 --> 00:55:59,897
and thought into this.
1210
00:56:00,497 --> 00:56:02,517
So, huge shout out to them.
1211
00:56:03,597 --> 00:56:05,436
It was also cool
1212
00:56:05,436 --> 00:56:07,237
to watch them use Buzz
1213
00:56:07,237 --> 00:56:08,477
in this process.
1214
00:56:08,477 --> 00:56:09,717
They literally use Buzz.
1215
00:56:09,916 --> 00:56:10,637
And I saw first,
1216
00:56:10,737 --> 00:56:13,397
this isn't a marketing thing for Buzz.
1217
00:56:13,517 --> 00:56:15,137
It's just they actually use this tool
1218
00:56:15,137 --> 00:56:18,317
because each of the security engineers
1219
00:56:18,317 --> 00:56:21,797
were doing their AI research
1220
00:56:21,797 --> 00:56:23,436
on these vulnerabilities,
1221
00:56:23,436 --> 00:56:25,356
but they were doing it in a collaborative environment.
1222
00:56:25,697 --> 00:56:28,157
Previously, they'd be doing it solo in an island.
1223
00:56:29,057 --> 00:56:30,037
And then they'd have to determine
1224
00:56:30,037 --> 00:56:31,837
when are they going to share results?
1225
00:56:31,977 --> 00:56:34,977
What are they going to share with others?
1226
00:56:35,237 --> 00:56:36,456
And there's friction.
1227
00:56:36,757 --> 00:56:37,956
Where here, there was no friction.
1228
00:56:38,717 --> 00:56:39,876
Everyone's doing the research.
1229
00:56:40,017 --> 00:56:41,297
It's all searchable.
1230
00:56:41,637 --> 00:56:44,817
It's all available to your own agent to see it all.
1231
00:56:44,817 --> 00:56:46,436
and it accelerated the process.
1232
00:56:46,577 --> 00:56:47,797
And then for people who aren't
1233
00:56:47,797 --> 00:56:50,197
the biggest technical expert,
1234
00:56:50,357 --> 00:56:51,897
but like comms people and executives
1235
00:56:51,897 --> 00:56:52,977
and everyone else,
1236
00:56:53,416 --> 00:56:56,397
can follow along and quickly get summaries
1237
00:56:56,397 --> 00:56:57,537
and have their questions answered.
1238
00:56:57,697 --> 00:56:59,436
It was really powerful and amazing.
1239
00:56:59,877 --> 00:57:01,477
So that's block corporate internal.
1240
00:57:01,697 --> 00:57:03,397
Obviously, that's not, you know,
1241
00:57:03,436 --> 00:57:04,837
not meant for public consumption,
1242
00:57:04,956 --> 00:57:06,077
but it seems like it could be useful
1243
00:57:06,077 --> 00:57:07,757
to have a buzz,
1244
00:57:07,897 --> 00:57:08,936
I don't know if this is on Flint
1245
00:57:08,936 --> 00:57:10,117
or somewhere, a new community
1246
00:57:10,117 --> 00:57:10,817
that gets popped up,
1247
00:57:10,877 --> 00:57:12,436
but a way that people in the public
1248
00:57:12,436 --> 00:57:13,697
who want to participate
1249
00:57:13,697 --> 00:57:16,017
and not necessarily all the security research,
1250
00:57:16,157 --> 00:57:17,557
but start to ask questions
1251
00:57:17,557 --> 00:57:18,977
and try to get the best information.
1252
00:57:19,356 --> 00:57:20,217
Yeah, that's a good idea.
1253
00:57:20,456 --> 00:57:22,177
I know of like several signal channels
1254
00:57:22,177 --> 00:57:23,157
that are private doing that,
1255
00:57:23,317 --> 00:57:26,817
but I think that's a good idea.
1256
00:57:27,017 --> 00:57:30,177
Like a lot of this stuff can be public.
1257
00:57:30,677 --> 00:57:31,817
I mean, especially if you're just like
1258
00:57:31,817 --> 00:57:35,177
an armchair quarterback at home or whatever,
1259
00:57:35,477 --> 00:57:37,456
or just an independent person,
1260
00:57:37,597 --> 00:57:39,037
you're not working for a company,
1261
00:57:39,037 --> 00:57:41,197
you have to sort of follow the guidelines
1262
00:57:41,197 --> 00:57:42,416
or rules of the company.
1263
00:57:43,697 --> 00:57:49,577
You can join forces with other like-minded and caring Bitcoin people to do that.
1264
00:57:49,657 --> 00:57:50,477
I think that's a great idea.
1265
00:57:51,956 --> 00:57:56,396
So, yeah, big shout out to the Block team.
1266
00:57:58,237 --> 00:58:03,997
And all that information also was shared ahead of time with CoinKite.
1267
00:58:05,037 --> 00:58:08,477
But then we felt it was important to publish it as well.
1268
00:58:08,477 --> 00:58:22,097
So if you were trying to contact your friends who you think might be affected or loved ones who have Bitcoin, you want to just make sure they understand, is the block post on that pretty exhaustive about what to do and the advice?
1269
00:58:22,177 --> 00:58:29,317
Like if somebody doesn't have a chance to listen to the show, or they're not going to share the whole show, would it be to share that one link to the block post or is there another?
1270
00:58:30,117 --> 00:58:35,177
I think the two most credible things would be CoinKite Communications.
1271
00:58:35,177 --> 00:58:37,057
That is the company behind this.
1272
00:58:37,057 --> 00:58:38,777
so they're official communications.
1273
00:58:38,956 --> 00:58:41,137
I mean, they do a blog post outline.
1274
00:58:41,297 --> 00:58:43,916
I mean, and you pointed this out before the show.
1275
00:58:44,017 --> 00:58:45,177
I mean, NVK literally says,
1276
00:58:45,477 --> 00:58:47,077
move your funds off his own product.
1277
00:58:47,456 --> 00:58:50,217
So you know it's serious.
1278
00:58:50,757 --> 00:58:52,637
If the founder and CEO and a person
1279
00:58:52,637 --> 00:58:55,297
who has spent a decade of his life
1280
00:58:55,297 --> 00:58:56,456
working on this product is saying,
1281
00:58:56,637 --> 00:58:58,537
move your funds off of it, it's serious.
1282
00:58:59,097 --> 00:59:01,357
So just that alone, I think, is what I'd share.
1283
00:59:01,357 --> 00:59:05,637
But I think the Block engineering post
1284
00:59:05,637 --> 00:59:12,637
it has more and better detail and also highlights risks with some of the other products that
1285
00:59:12,637 --> 00:59:19,356
I haven't yet seen CoinKite, you know, emphasize yet. So I think that would be a good share as well.
1286
00:59:19,356 --> 00:59:23,497
On the Buzz community idea, I wonder, maybe it's under a project loop or something else, but
1287
00:59:23,497 --> 00:59:27,597
it'd also be interesting to see a community where people could come together, propose projects that
1288
00:59:27,597 --> 00:59:31,356
need to get audited, companies that need to get audited, help reach out to those companies, bring those people
1289
00:59:31,409 --> 00:59:54,049
And because I'm just thinking right now, like, you know, there's whatever dozens of companies in the Bitcoin space. Now, granted, most don't have that, you know, that great of an asset center management or whatever. But like, for like the long tail of these companies, like, is there a community to like identify them, reach out to them, bring them into the mix? See, are you running, you know, the latest models? If you're not, can we find a way to help you do that? That seems to me like that would be extremely high leverage.
1290
00:59:54,049 --> 00:59:57,089
so let's talk about
1291
00:59:57,089 --> 00:59:59,249
what can be done
1292
00:59:59,249 --> 01:00:01,089
to try to minimize
1293
01:00:01,089 --> 01:00:01,949
the damage
1294
01:00:01,949 --> 01:00:05,009
the blast radius
1295
01:00:05,009 --> 01:00:05,469
if you will
1296
01:00:05,469 --> 01:00:08,989
if you're using too many LLMs
1297
01:00:08,989 --> 01:00:12,929
what makes this so challenging
1298
01:00:12,929 --> 01:00:15,069
is that most software
1299
01:00:15,069 --> 01:00:15,749
vulnerabilities
1300
01:00:15,749 --> 01:00:19,009
like zero days or something
1301
01:00:19,009 --> 01:00:20,289
they can be really
1302
01:00:20,289 --> 01:00:22,689
severe problems but
1303
01:00:22,689 --> 01:00:28,689
a software patch can be released and you update and then you're protected.
1304
01:00:29,229 --> 01:00:29,329
Right?
1305
01:00:29,429 --> 01:00:33,929
That's, I think for most people, that's what you experience.
1306
01:00:33,929 --> 01:00:38,349
And, you know, so hopefully you weren't impacted in that narrow window of time
1307
01:00:38,349 --> 01:00:41,509
in which it was publicly known and exposed.
1308
01:00:42,229 --> 01:00:44,269
This is not the case here, though, right?
1309
01:00:44,729 --> 01:00:49,889
If you're vulnerable and you just upgrade the firmware, you're still vulnerable.
1310
01:00:49,889 --> 01:00:53,089
because your key was generated in the past.
1311
01:00:53,669 --> 01:00:54,189
It's vulnerable.
1312
01:00:54,369 --> 01:00:56,629
And just upgrading the firmware doesn't change that key,
1313
01:00:56,729 --> 01:00:57,729
doesn't make it strong.
1314
01:00:57,949 --> 01:01:00,869
You have to sweep your funds into a new wallet,
1315
01:01:01,049 --> 01:01:02,229
transfer your funds into a new wallet.
1316
01:01:02,449 --> 01:01:04,569
It makes it really tricky.
1317
01:01:04,569 --> 01:01:08,389
And that's also why it's so important for people to communicate
1318
01:01:08,389 --> 01:01:11,069
with friends and family and those that they know.
1319
01:01:11,949 --> 01:01:15,109
And influencers on, I mean, many podcasters
1320
01:01:15,109 --> 01:01:19,269
took sponsorship money from CoinKite.
1321
01:01:19,269 --> 01:01:21,109
and I'm not shitting in them for doing that
1322
01:01:21,109 --> 01:01:24,229
but I do think it's their responsibility
1323
01:01:24,229 --> 01:01:27,169
to tell their audiences about this issue
1324
01:01:27,169 --> 01:01:28,389
to spread the word.
1325
01:01:28,649 --> 01:01:29,689
The more people that know the better
1326
01:01:29,689 --> 01:01:31,129
because otherwise there might be people
1327
01:01:31,129 --> 01:01:34,169
who don't learn of this for days, weeks, months later
1328
01:01:34,169 --> 01:01:35,229
and they're going to wake,
1329
01:01:35,509 --> 01:01:36,689
three months from now,
1330
01:01:36,989 --> 01:01:38,589
they go check and all their money is gone
1331
01:01:38,589 --> 01:01:40,009
and they're like, what's going on?
1332
01:01:40,029 --> 01:01:40,869
And they'll be like, oh my God,
1333
01:01:40,889 --> 01:01:42,029
this was discovered three months ago.
1334
01:01:42,169 --> 01:01:43,629
I could have done something about it.
1335
01:01:43,629 --> 01:01:44,029
Right?
1336
01:01:44,669 --> 01:01:46,029
So spreading the word is super important.
1337
01:01:47,229 --> 01:01:48,589
So what else can be done?
1338
01:01:48,589 --> 01:01:59,509
Well, in an ideal world, the attacker can be discovered and found, right?
1339
01:01:59,809 --> 01:02:01,029
That would be step one.
1340
01:02:01,349 --> 01:02:02,769
So maybe that happens.
1341
01:02:03,849 --> 01:02:09,249
Okay, even if that happens, number two is getting the Bitcoin back from the attacker.
1342
01:02:09,469 --> 01:02:10,449
And that's not a given.
1343
01:02:11,549 --> 01:02:15,549
I mean, they could literally just burn the money and it's gone forever.
1344
01:02:15,549 --> 01:02:19,209
or they moved it somewhere
1345
01:02:19,209 --> 01:02:23,169
and tossed the key in some farmland
1346
01:02:23,169 --> 01:02:25,149
and will never give it up.
1347
01:02:25,609 --> 01:02:26,349
So it's not a given,
1348
01:02:26,489 --> 01:02:28,149
but hopefully the person's apprehended.
1349
01:02:29,269 --> 01:02:31,689
The money is claimed back.
1350
01:02:32,089 --> 01:02:33,229
But what's the third step?
1351
01:02:34,069 --> 01:02:35,669
So if the authorities or whatever
1352
01:02:35,669 --> 01:02:36,589
are able to get the Bitcoin,
1353
01:02:36,809 --> 01:02:40,189
how do they get it back to the rightful owners?
1354
01:02:41,569 --> 01:02:43,449
Showproof, I guess, original ownership.
1355
01:02:43,869 --> 01:02:44,629
Well, I have some answers,
1356
01:02:44,689 --> 01:02:45,129
but I'm curious.
1357
01:02:45,129 --> 01:02:48,389
Do you have any thoughts before I share some of the ideas?
1358
01:02:48,529 --> 01:02:51,569
I mean, if you could sign a message from your, I guess, original,
1359
01:02:52,129 --> 01:02:56,009
when you created the coin, it's like from that key.
1360
01:02:56,209 --> 01:02:58,089
But any of us can sign the same message.
1361
01:02:58,769 --> 01:02:59,649
Oh, now you can. You're right.
1362
01:03:00,789 --> 01:03:01,469
That's a good question.
1363
01:03:01,769 --> 01:03:05,369
It's kind of like, you know, we've been going through these mental exercises of quantum.
1364
01:03:05,609 --> 01:03:06,749
Like, what if there's a quantum computer?
1365
01:03:07,049 --> 01:03:10,829
How do we distinguish between you, the rightful owner, and me, the quantum attacker?
1366
01:03:10,949 --> 01:03:12,189
It's the same thing, right?
1367
01:03:12,189 --> 01:03:22,089
So, fortunately, there's some smart folks, including folks on Spiral, but others on Twitter that are in the Bitcoin community who are coming up with ideas.
1368
01:03:22,369 --> 01:03:23,749
So I'll summarize some ideas.
1369
01:03:23,849 --> 01:03:35,889
And I think there's three types of evidence that collectively we've come up with that I think do a pretty good job at demonstrating ownership.
1370
01:03:36,669 --> 01:03:39,949
Good enough that I think it could be used if needed.
1371
01:03:39,949 --> 01:03:43,149
one is that
1372
01:03:43,149 --> 01:03:45,149
I mentioned earlier
1373
01:03:45,149 --> 01:03:47,389
that each device has a unique ID
1374
01:03:47,389 --> 01:03:48,009
UID
1375
01:03:48,009 --> 01:03:51,089
and it's used in the generation
1376
01:03:51,089 --> 01:03:52,609
of the key
1377
01:03:52,609 --> 01:03:55,509
so if I come to you
1378
01:03:55,509 --> 01:03:57,209
and I'm like here's my hardware
1379
01:03:57,209 --> 01:03:59,389
it's the only hardware
1380
01:03:59,389 --> 01:04:01,029
in the world that has UID
1381
01:04:01,029 --> 01:04:03,209
73 or whatever
1382
01:04:03,209 --> 01:04:03,829
the number is
1383
01:04:03,829 --> 01:04:07,609
and the only
1384
01:04:07,609 --> 01:04:09,529
way to generate the key
1385
01:04:09,529 --> 01:04:12,249
from my wallet is to use the number 73
1386
01:04:12,249 --> 01:04:14,209
because it was used in the key generation.
1387
01:04:14,869 --> 01:04:16,809
And so then we can all look at the public blockchain
1388
01:04:16,809 --> 01:04:18,929
and see how much money was in that wallet.
1389
01:04:18,929 --> 01:04:20,789
So I can come to you with hardware
1390
01:04:20,789 --> 01:04:25,169
that seems strong evidence that I had that much money.
1391
01:04:26,149 --> 01:04:30,369
Okay, but it is possible to create fake hardware
1392
01:04:30,369 --> 01:04:32,729
that has UID 73.
1393
01:04:33,569 --> 01:04:37,449
So an additional piece of evidence would be
1394
01:04:37,449 --> 01:04:40,889
I have an email from CoinKite,
1395
01:04:41,409 --> 01:04:45,169
you know, from like the year 2023,
1396
01:04:45,909 --> 01:04:47,449
showing that I'm a customer
1397
01:04:47,449 --> 01:04:49,729
and I purchased a, you know,
1398
01:04:49,829 --> 01:04:51,449
cold card MK3 or whatever.
1399
01:04:51,729 --> 01:04:51,829
Right.
1400
01:04:52,669 --> 01:04:56,889
And it's a DKIM signed email,
1401
01:04:57,189 --> 01:04:58,869
which means it was signed
1402
01:04:58,869 --> 01:05:00,669
by a cold card server key.
1403
01:05:01,509 --> 01:05:02,109
Did they do that?
1404
01:05:02,649 --> 01:05:04,449
We demonstrated it this morning.
1405
01:05:04,449 --> 01:05:09,889
So I actually bought some OpenDimes in 2023.
1406
01:05:10,409 --> 01:05:11,869
So I grabbed my email from there,
1407
01:05:12,109 --> 01:05:13,229
looked at the email header,
1408
01:05:13,809 --> 01:05:14,889
and it has the signature.
1409
01:05:15,489 --> 01:05:18,329
And that signature, it signs for all the contents.
1410
01:05:18,889 --> 01:05:21,849
So you can't forge a different email recipient
1411
01:05:21,849 --> 01:05:23,089
or a different date.
1412
01:05:24,049 --> 01:05:26,549
So it's cryptographically secure.
1413
01:05:26,629 --> 01:05:26,869
That's good.
1414
01:05:27,169 --> 01:05:31,029
And then we also checked the current server key
1415
01:05:31,029 --> 01:05:32,029
from CoinKite.
1416
01:05:32,029 --> 01:05:35,509
it does verify my email from 2023.
1417
01:05:36,509 --> 01:05:40,289
So, I mean, we should also prove that it goes back to 2021
1418
01:05:40,289 --> 01:05:42,949
and still that same key works.
1419
01:05:43,109 --> 01:05:47,049
But that's another piece of evidence.
1420
01:05:47,349 --> 01:05:50,189
So if I come to you with this hardware wallet
1421
01:05:50,189 --> 01:05:54,029
and this proof that I actually bought a hardware wallet,
1422
01:05:54,169 --> 01:05:56,849
I don't think the UID is in the email.
1423
01:05:57,049 --> 01:05:59,349
So I don't have proof that it's that hardware wallet,
1424
01:05:59,349 --> 01:06:00,349
but I do have proof
1425
01:06:00,349 --> 01:06:02,349
that three years ago
1426
01:06:02,349 --> 01:06:03,429
I bought a hardware wallet.
1427
01:06:03,809 --> 01:06:05,089
That's very strong evidence.
1428
01:06:05,569 --> 01:06:06,449
The third piece of evidence
1429
01:06:06,449 --> 01:06:07,089
that could be used,
1430
01:06:07,829 --> 01:06:09,649
I suspect a lot of customers
1431
01:06:09,649 --> 01:06:10,849
transferred coins
1432
01:06:10,849 --> 01:06:12,049
that they bought on an exchange
1433
01:06:12,049 --> 01:06:12,929
to Coldcard.
1434
01:06:13,089 --> 01:06:13,889
That's a great one.
1435
01:06:13,949 --> 01:06:14,089
Right?
1436
01:06:14,329 --> 01:06:14,609
And then,
1437
01:06:14,849 --> 01:06:15,869
so there can be proof
1438
01:06:15,869 --> 01:06:17,129
from that exchange
1439
01:06:17,129 --> 01:06:18,689
where you had to KYC
1440
01:06:18,689 --> 01:06:20,789
that I'm so-and-so.
1441
01:06:21,189 --> 01:06:21,369
Yeah.
1442
01:06:21,489 --> 01:06:22,049
And then there's
1443
01:06:22,049 --> 01:06:23,209
blockchain analysis
1444
01:06:23,209 --> 01:06:24,429
that shows those coins
1445
01:06:24,429 --> 01:06:25,809
move from the exchange
1446
01:06:25,809 --> 01:06:27,649
to the address
1447
01:06:27,649 --> 01:06:28,289
of the Coldcard.
1448
01:06:28,289 --> 01:06:35,269
So I think there's like, that's pretty credible that you could have a process to fairly give people their Bitcoin back.
1449
01:06:35,429 --> 01:06:38,089
So I don't want to create false hopes for people.
1450
01:06:38,409 --> 01:06:41,349
We do not know if all three of these steps can happen.
1451
01:06:41,509 --> 01:06:42,009
But it's promising.
1452
01:06:42,709 --> 01:06:44,709
But it's possible.
1453
01:06:44,929 --> 01:06:48,989
If you had to guess, is it like 1% chance or 80% chance?
1454
01:06:48,989 --> 01:06:50,069
I don't want to give a false hope.
1455
01:06:50,229 --> 01:06:57,549
But I think it's plausible enough that again, a call to action to everyone in the ecosystem.
1456
01:06:57,549 --> 01:06:59,709
if you learn of anything
1457
01:06:59,709 --> 01:07:02,109
that can help in any of those three steps.
1458
01:07:02,329 --> 01:07:04,489
Like if you think of an additional piece of evidence
1459
01:07:04,489 --> 01:07:07,169
that could help on the third phase,
1460
01:07:07,629 --> 01:07:09,289
please share that, you know, again,
1461
01:07:09,389 --> 01:07:11,009
security at coin kite dot com
1462
01:07:11,009 --> 01:07:13,529
or share it with, you know,
1463
01:07:13,649 --> 01:07:14,649
publish it on Twitter.
1464
01:07:14,869 --> 01:07:16,329
If you don't think it's, you know,
1465
01:07:16,349 --> 01:07:17,289
if you think it's responsible
1466
01:07:17,289 --> 01:07:19,889
and it's not sensitive, get the word out.
1467
01:07:21,729 --> 01:07:25,009
Yeah, I think it's worth people's time, at least.
1468
01:07:25,229 --> 01:07:26,049
I mean, and hackers,
1469
01:07:26,049 --> 01:07:27,749
There is precedent for getting apprehended.
1470
01:07:28,349 --> 01:07:29,609
You know, the Gox coins eventually.
1471
01:07:29,729 --> 01:07:30,769
I don't know how long that'll take.
1472
01:07:30,829 --> 01:07:32,649
It could take forever to distribute it.
1473
01:07:33,789 --> 01:07:38,089
Now, here's another very interesting thing to discuss.
1474
01:07:39,529 --> 01:07:43,069
As stated earlier, there's a lot of vulnerable coins
1475
01:07:43,069 --> 01:07:46,809
that are likely to be stolen still.
1476
01:07:49,049 --> 01:07:50,449
So it could be new thieves now.
1477
01:07:51,149 --> 01:07:52,409
But there's a ton of vulnerable coins
1478
01:07:52,409 --> 01:07:53,269
that have not been stolen.
1479
01:07:53,429 --> 01:07:53,589
Right.
1480
01:07:53,589 --> 01:07:56,089
and well I say a ton
1481
01:07:56,089 --> 01:07:58,169
like there are people looking into
1482
01:07:58,169 --> 01:07:59,449
trying to quantify that
1483
01:07:59,449 --> 01:08:01,829
like how much, how many wallets
1484
01:08:01,829 --> 01:08:03,029
how many coins
1485
01:08:03,029 --> 01:08:06,169
but with what little I know about
1486
01:08:06,169 --> 01:08:08,169
that question I think it's
1487
01:08:08,169 --> 01:08:08,949
material
1488
01:08:08,949 --> 01:08:12,149
like it
1489
01:08:12,149 --> 01:08:14,129
wouldn't surprise me if it's hundreds of millions of
1490
01:08:14,129 --> 01:08:16,089
dollars more that could be stolen
1491
01:08:16,089 --> 01:08:18,329
so if there are
1492
01:08:18,329 --> 01:08:20,169
there are people
1493
01:08:20,169 --> 01:08:22,129
who know how to steal these
1494
01:08:22,129 --> 01:08:27,869
coins. The methods are known now. So if you have, so obviously we don't want black hat hackers
1495
01:08:27,869 --> 01:08:33,589
stealing the coins. That's criminality. That's theft. It's bad. What if white hat hackers did it?
1496
01:08:33,889 --> 01:08:40,189
Yeah. And combined with, we know if we have, if we have a process that, um, like the community
1497
01:08:40,189 --> 01:08:45,889
feels overall is like fair and just to get the coins back to the rightful owners, then you can,
1498
01:08:45,889 --> 01:08:50,669
To me, my own opinion is it's ethical to do that.
1499
01:08:50,669 --> 01:09:04,321
The white hat hacking and the transfer of funds into a system and a process that would get it to the rightful owners But you know who does that right Is it even legal I haven done the legal analysis
1500
01:09:04,501 --> 01:09:09,021
I haven't spoken to lawyers yet and seen legal analysis on that.
1501
01:09:09,101 --> 01:09:12,261
So I don't even know what the precedent is on that, the legality.
1502
01:09:13,581 --> 01:09:20,421
But, you know, to me in an ideal world, like U.S. authorities would be the best position.
1503
01:09:20,421 --> 01:09:23,681
because they, well, I mean, whether it's legal or not,
1504
01:09:23,741 --> 01:09:25,581
if the U.S. does it, then the U.S., you know.
1505
01:09:25,841 --> 01:09:26,481
It's legal now.
1506
01:09:26,581 --> 01:09:27,281
It's legal now.
1507
01:09:29,501 --> 01:09:32,101
You know, and if it was a transparent process,
1508
01:09:32,721 --> 01:09:34,321
and if everything I said earlier is like,
1509
01:09:34,381 --> 01:09:36,681
it pans out to be true and accurate,
1510
01:09:36,801 --> 01:09:39,361
like that there is a just fair process for people
1511
01:09:39,361 --> 01:09:41,121
and at least the majority of people
1512
01:09:41,121 --> 01:09:42,941
correctly get their Bitcoin back.
1513
01:09:43,061 --> 01:09:45,001
And there's not like a vulnerability.
1514
01:09:45,281 --> 01:09:46,981
What we want to avoid is a vulnerability in the process
1515
01:09:46,981 --> 01:09:50,701
where an attacker can exploit that process
1516
01:09:50,701 --> 01:09:52,501
and get all the Bitcoin themselves.
1517
01:09:52,661 --> 01:09:54,281
Well, obviously, that would be a disaster
1518
01:09:54,281 --> 01:09:56,001
and doesn't solve the problem.
1519
01:09:57,661 --> 01:09:59,541
But yeah, ideally it's then,
1520
01:09:59,621 --> 01:10:04,101
because then it's not any cowboy doing it
1521
01:10:04,101 --> 01:10:07,341
or global cowboy trying to do this
1522
01:10:07,341 --> 01:10:09,681
or even a company trying to do it.
1523
01:10:11,681 --> 01:10:13,601
I think that's challenging.
1524
01:10:13,941 --> 01:10:14,361
Yeah, it's interesting.
1525
01:10:14,361 --> 01:10:16,641
but in any case
1526
01:10:16,641 --> 01:10:17,801
like
1527
01:10:17,801 --> 01:10:19,521
I think
1528
01:10:19,521 --> 01:10:21,441
that could have
1529
01:10:21,441 --> 01:10:23,341
that could really mitigate
1530
01:10:23,341 --> 01:10:24,421
the problem
1531
01:10:24,421 --> 01:10:25,121
or it could like
1532
01:10:25,121 --> 01:10:27,121
it could reduce the blast radius tremendously
1533
01:10:27,121 --> 01:10:28,181
so I think that's worth
1534
01:10:28,181 --> 01:10:29,221
a lot of people
1535
01:10:29,221 --> 01:10:30,181
thinking about more
1536
01:10:30,181 --> 01:10:30,541
and coming
1537
01:10:30,541 --> 01:10:30,921
you know
1538
01:10:30,921 --> 01:10:32,841
smarter ideas
1539
01:10:32,841 --> 01:10:33,701
and new ideas
1540
01:10:33,701 --> 01:10:35,741
bringing to bear on that
1541
01:10:35,741 --> 01:10:37,861
how about
1542
01:10:37,861 --> 01:10:38,281
should we
1543
01:10:38,281 --> 01:10:39,341
maybe the last topic
1544
01:10:39,341 --> 01:10:40,361
is like the impact
1545
01:10:40,361 --> 01:10:41,261
on self-custard
1546
01:10:41,261 --> 01:10:41,461
yeah
1547
01:10:41,461 --> 01:10:42,061
it's a big one
1548
01:10:42,061 --> 01:10:43,181
I mean you've already teed up
1549
01:10:43,181 --> 01:10:44,161
several of those conversations
1550
01:10:44,161 --> 01:10:46,581
and I had down here like talking about Project Loop.
1551
01:10:46,661 --> 01:10:47,881
We talked about that,
1552
01:10:47,881 --> 01:10:50,201
but custody diversification,
1553
01:10:50,401 --> 01:10:51,841
we already, we talked about that.
1554
01:10:52,841 --> 01:10:54,101
But yeah, I had down here,
1555
01:10:54,341 --> 01:10:56,141
are custodians safer?
1556
01:10:56,401 --> 01:10:58,461
And that's sort of like your mantra.
1557
01:10:59,141 --> 01:10:59,821
Well, and again,
1558
01:10:59,961 --> 01:11:01,601
I'm not saying all custodians are safer.
1559
01:11:02,161 --> 01:11:03,541
I'm saying, you know,
1560
01:11:04,421 --> 01:11:06,001
everyone has their own unique calculus
1561
01:11:06,001 --> 01:11:07,141
and like how deep they are,
1562
01:11:07,241 --> 01:11:08,881
what risks they're willing to understand.
1563
01:11:09,181 --> 01:11:10,781
But what I am increasingly saying,
1564
01:11:10,941 --> 01:11:13,001
perhaps the one slightly different angle
1565
01:11:13,001 --> 01:11:13,861
I'm taking on this
1566
01:11:13,861 --> 01:11:16,041
than most people I've seen or talked to so far
1567
01:11:16,041 --> 01:11:17,541
is, again, I think the big story is AI.
1568
01:11:18,201 --> 01:11:20,281
And so my big takeaway is
1569
01:11:20,281 --> 01:11:23,641
if there are custodians that are, you know,
1570
01:11:24,601 --> 01:11:28,201
very, let's say, well-resourced
1571
01:11:28,201 --> 01:11:30,481
and are getting access to the absolute latest
1572
01:11:30,481 --> 01:11:31,321
cutting-edge models,
1573
01:11:31,441 --> 01:11:33,521
at least during this transition phase,
1574
01:11:33,841 --> 01:11:36,741
that to me seems like a pretty, honestly,
1575
01:11:37,281 --> 01:11:38,641
you know, promising trade-off.
1576
01:11:39,521 --> 01:11:40,361
But then again, I don't know.
1577
01:11:40,541 --> 01:11:41,281
It's hard to know.
1578
01:11:41,381 --> 01:11:42,221
Are they? How are they?
1579
01:11:42,221 --> 01:11:45,181
I think there's also a range of custodians.
1580
01:11:45,521 --> 01:11:47,121
Like an ETF is a custodian.
1581
01:11:47,821 --> 01:11:50,341
A Coinbase or a River are custodians.
1582
01:11:51,121 --> 01:11:54,821
So there's a range of what kinds of risks you're taking.
1583
01:11:54,921 --> 01:11:57,401
All but one ETF is not a custodian.
1584
01:11:58,741 --> 01:12:01,021
Fidelity is the only ETF that's a custodian.
1585
01:12:01,221 --> 01:12:04,961
I think the ETFs have diversified a bit under the hood.
1586
01:12:05,061 --> 01:12:05,341
Some.
1587
01:12:05,441 --> 01:12:05,941
Is my understanding.
1588
01:12:06,821 --> 01:12:10,321
But still, they all launched with Coinbase.
1589
01:12:10,321 --> 01:12:12,621
I think one used another one.
1590
01:12:13,221 --> 01:12:13,561
Gemini.
1591
01:12:13,701 --> 01:12:15,261
I mean, did they use Fidelity?
1592
01:12:15,721 --> 01:12:16,601
Fidelity used Fidelity.
1593
01:12:16,601 --> 01:12:17,181
Fidelity used Fidelity.
1594
01:12:17,181 --> 01:12:17,981
Everyone else used Fidelity.
1595
01:12:18,321 --> 01:12:20,041
And then Gemini used Fidelity?
1596
01:12:20,301 --> 01:12:20,601
No, no, no.
1597
01:12:20,641 --> 01:12:22,081
Someone, one of the ETFs, I don't remember.
1598
01:12:22,081 --> 01:12:22,961
Oh, used Gemini.
1599
01:12:23,121 --> 01:12:23,261
Right.
1600
01:12:23,521 --> 01:12:25,261
And then I think Anchorage has gotten added in.
1601
01:12:25,421 --> 01:12:28,401
There are a few of these, but the point is it's like five or fewer.
1602
01:12:28,521 --> 01:12:28,981
Highly centralized.
1603
01:12:29,201 --> 01:12:29,381
Yeah, yeah.
1604
01:12:29,481 --> 01:12:36,321
So the, and by the way, I'm not, in this discussion, I'm not trying to FUD any particular solution.
1605
01:12:37,121 --> 01:12:37,361
Yeah, yeah.
1606
01:12:37,361 --> 01:12:37,501
Right.
1607
01:12:37,541 --> 01:12:38,921
Because we can go, there's trade-offs everywhere.
1608
01:12:38,921 --> 01:12:39,881
There's trade-offs everywhere.
1609
01:12:40,321 --> 01:12:44,521
whether it's BitKey or Coinbase or Cash App or River
1610
01:12:44,521 --> 01:12:46,181
or any of these trade-offs.
1611
01:12:46,801 --> 01:12:49,801
So I'd say that a new,
1612
01:12:50,061 --> 01:12:54,361
I mean, let's not rehash traditional risks with custodians,
1613
01:12:54,501 --> 01:12:58,981
but a new risk would be the rise of AI can impact this too.
1614
01:12:58,981 --> 01:13:01,861
So we just saw a week ago,
1615
01:13:02,261 --> 01:13:08,361
like OpenAI, a company that is like the world expert at AI systems
1616
01:13:08,361 --> 01:13:11,761
had an AI escape from their company.
1617
01:13:12,881 --> 01:13:17,261
Unknowingly, it escaped and attacked other companies for a week.
1618
01:13:17,741 --> 01:13:19,161
OpenAI wasn't even aware.
1619
01:13:19,621 --> 01:13:21,121
And they're the experts in the world.
1620
01:13:21,541 --> 01:13:25,121
And what happens if the stealer, the thief, becomes an AI, not a human?
1621
01:13:25,421 --> 01:13:25,621
Right.
1622
01:13:25,741 --> 01:13:28,981
So this wasn't even, that wasn't like an intentional,
1623
01:13:30,021 --> 01:13:31,801
I mean, it was intentional in some ways.
1624
01:13:31,801 --> 01:13:36,221
They were intentionally trying to test a new model for its cyber abilities,
1625
01:13:36,221 --> 01:13:41,801
But they didn't like, but they didn't like intent, my understanding, they didn't intend
1626
01:13:41,801 --> 01:13:43,821
it to go attack Hugging Face or other companies.
1627
01:13:43,941 --> 01:13:44,021
Right.
1628
01:13:45,641 --> 01:13:51,541
So if like a world expert can't even control their AI, you can imagine a world where, in
1629
01:13:51,541 --> 01:13:57,721
fact, I mean, it's hard to imagine a world not where Coinbase and Block and River and
1630
01:13:57,721 --> 01:14:00,601
you name it, all these companies are going to be attacked by AIs.
1631
01:14:00,741 --> 01:14:00,921
Yes.
1632
01:14:01,301 --> 01:14:02,621
So I think that's inevitable.
1633
01:14:03,461 --> 01:14:07,161
Now the question is, what's the fortress look like?
1634
01:14:07,161 --> 01:14:10,641
The good news is, unlike with Hugging Face,
1635
01:14:10,701 --> 01:14:12,121
I don't know exactly what got hacked,
1636
01:14:12,241 --> 01:14:16,161
but I assume only online computers were accessed
1637
01:14:16,161 --> 01:14:17,781
and something happened.
1638
01:14:18,321 --> 01:14:23,281
With cold storage, that does prevent a huge barrier for AI.
1639
01:14:23,921 --> 01:14:27,141
So that's one reason to sleep better at night
1640
01:14:27,141 --> 01:14:29,641
if you have coins at River or Cash App or Coinbase
1641
01:14:29,641 --> 01:14:30,201
or what have you.
1642
01:14:30,201 --> 01:14:33,141
because they have a closed door
1643
01:14:33,141 --> 01:14:35,081
yeah because the AI can't like
1644
01:14:35,081 --> 01:14:37,241
leap into the air and like
1645
01:14:37,241 --> 01:14:39,321
access not yet you know
1646
01:14:39,321 --> 01:14:41,161
the robots are coming but something
1647
01:14:41,161 --> 01:14:42,641
that's not connected to the internet
1648
01:14:42,641 --> 01:14:45,281
but I mean the robots are coming
1649
01:14:45,281 --> 01:14:47,441
but like these systems
1650
01:14:47,441 --> 01:14:47,861
at
1651
01:14:47,861 --> 01:14:51,221
you know more sophisticated companies like
1652
01:14:51,221 --> 01:14:53,421
a coinbase or a block and I'm for a rib or two
1653
01:14:53,421 --> 01:14:55,581
they have layers of security
1654
01:14:55,581 --> 01:14:57,921
and
1655
01:14:57,921 --> 01:15:03,901
And I guess my guess would be like the heightened risk of AI
1656
01:15:03,901 --> 01:15:09,581
probably penetrates some of those layers much more successfully than the past.
1657
01:15:09,961 --> 01:15:11,941
So it's just a question of like, you know,
1658
01:15:12,401 --> 01:15:15,541
are they still blocked at the final layers or not?
1659
01:15:16,141 --> 01:15:20,401
Are these companies hardening their systems, etc.?
1660
01:15:20,401 --> 01:15:21,881
So there's still that risk.
1661
01:15:22,781 --> 01:15:24,021
Which again, I'll just come back to,
1662
01:15:24,141 --> 01:15:27,181
if there is some way to report to the public
1663
01:15:27,181 --> 01:15:29,701
from these companies or from a third party,
1664
01:15:29,841 --> 01:15:31,161
like a wallet scrutiny type thing or something,
1665
01:15:31,461 --> 01:15:33,981
I would like to know, to the extent that we can know,
1666
01:15:34,641 --> 01:15:36,241
what are all of these companies doing?
1667
01:15:36,441 --> 01:15:38,141
Like, are they, you know, 24-7,
1668
01:15:38,301 --> 01:15:39,601
recursively changing their prompts?
1669
01:15:39,801 --> 01:15:41,801
Like, do these guys have not just like
1670
01:15:41,801 --> 01:15:42,881
old school security people,
1671
01:15:42,981 --> 01:15:44,241
but do they have people that are like
1672
01:15:44,241 --> 01:15:46,381
using the absolute latest models
1673
01:15:46,381 --> 01:15:47,921
and are, you know, native to that?
1674
01:15:47,921 --> 01:15:51,241
I do agree that it's good marketing
1675
01:15:51,241 --> 01:15:52,181
from these companies
1676
01:15:52,181 --> 01:15:55,401
if they have like engineering blog posts
1677
01:15:55,401 --> 01:15:57,881
and content that outline in detail,
1678
01:15:58,581 --> 01:15:59,581
here's what we do,
1679
01:15:59,901 --> 01:16:01,421
or here's what we're doing new
1680
01:16:01,421 --> 01:16:02,541
because of the new threats.
1681
01:16:02,741 --> 01:16:04,221
That would give me more confidence
1682
01:16:04,221 --> 01:16:07,401
versus a vendor that isn't doing that.
1683
01:16:08,121 --> 01:16:09,461
So I agree with that.
1684
01:16:09,521 --> 01:16:10,361
But at the same time,
1685
01:16:10,361 --> 01:16:12,821
it's still trusting them.
1686
01:16:13,321 --> 01:16:14,021
It's not like,
1687
01:16:14,081 --> 01:16:14,661
I don't think these companies
1688
01:16:14,661 --> 01:16:16,761
are going to lie about what they're doing.
1689
01:16:17,261 --> 01:16:21,021
But even if you're a security expert
1690
01:16:21,021 --> 01:16:22,441
and you read all the details,
1691
01:16:22,861 --> 01:16:24,421
because you still might not be aware
1692
01:16:24,421 --> 01:16:27,041
of all the holes still that are not being covered.
1693
01:16:27,621 --> 01:16:28,581
Nothing is perfect.
1694
01:16:29,981 --> 01:16:33,101
But so future of self-custody,
1695
01:16:33,301 --> 01:16:35,781
does this mean the end of self-custody?
1696
01:16:35,981 --> 01:16:37,581
Well, on one hand,
1697
01:16:37,901 --> 01:16:39,241
well, I don't think it's the end of self-custody,
1698
01:16:39,421 --> 01:16:41,241
but I do think on one hand,
1699
01:16:41,361 --> 01:16:42,981
it's definitely a major step backwards
1700
01:16:42,981 --> 01:16:47,321
because what percentage of cold card customers
1701
01:16:47,321 --> 01:16:52,261
are going to move to exchanges now or ETFs?
1702
01:16:52,261 --> 01:16:52,621
Most.
1703
01:16:53,941 --> 01:16:54,381
Okay.
1704
01:16:54,421 --> 01:16:56,961
I don't know what the answer will be,
1705
01:16:57,061 --> 01:16:59,001
but I'm not going to tell you that that's wrong.
1706
01:16:59,821 --> 01:17:01,441
I hope it's not most,
1707
01:17:01,541 --> 01:17:02,781
but even if half do,
1708
01:17:03,201 --> 01:17:06,081
we don't have that many people self-custodying.
1709
01:17:06,281 --> 01:17:08,041
We have a lot of Bitcoin that self-custody,
1710
01:17:08,181 --> 01:17:08,801
but not people.
1711
01:17:09,001 --> 01:17:11,061
A small percentage of Bitcoin is self-custody.
1712
01:17:11,741 --> 01:17:15,561
So if we lose half of cold car customers
1713
01:17:15,561 --> 01:17:18,581
or majority, that would be bad.
1714
01:17:21,001 --> 01:17:23,641
And then even if you're using another solution,
1715
01:17:23,641 --> 01:17:30,401
You're saying, well, how vulnerable is Trezor, Ledger, BitKey, Jade, any other solution.
1716
01:17:31,101 --> 01:17:33,621
This will probably cause a lot of Bitcoiners who are doing that to,
1717
01:17:34,401 --> 01:17:40,421
they might have spent a good chunk of time five years ago looking into this and picking the product.
1718
01:17:40,661 --> 01:17:42,821
And then they're like, this product's the best for me.
1719
01:17:43,461 --> 01:17:45,221
And then they haven't thought about it since.
1720
01:17:45,421 --> 01:17:47,161
That probably describes a lot of people.
1721
01:17:48,001 --> 01:17:53,501
An event like this, I mean, a horrific event, but a positive is it does encourage
1722
01:17:53,641 --> 01:17:56,221
a lot of people to go re-evaluate
1723
01:17:56,221 --> 01:17:57,401
their security environment.
1724
01:17:57,941 --> 01:17:59,181
A lot of things have improved.
1725
01:17:59,321 --> 01:18:00,381
Like products exist today
1726
01:18:00,381 --> 01:18:02,781
that didn't exist five to seven years ago.
1727
01:18:03,381 --> 01:18:05,241
And improvements have been made.
1728
01:18:05,581 --> 01:18:07,261
Plus you might have just learned yourself,
1729
01:18:07,441 --> 01:18:08,981
your own setup that, you know,
1730
01:18:09,101 --> 01:18:10,801
you would do things differently.
1731
01:18:12,421 --> 01:18:15,581
So I think that's a positive.
1732
01:18:15,581 --> 01:18:18,341
But then I do wonder how many will,
1733
01:18:18,781 --> 01:18:19,881
even if you didn't have a cold card,
1734
01:18:20,161 --> 01:18:36,113
but you just be like I don want to take the risk that my treasure ledger big key whatever is going to be next And then you move it to a custodian Again though I think we get back to custody diversification is probably the biggest piece of advice we could share
1735
01:18:37,013 --> 01:18:38,873
Because there's risks with the custodians, too.
1736
01:18:39,013 --> 01:18:40,053
There's risks everywhere.
1737
01:18:41,053 --> 01:18:43,333
Choose two, split your stack.
1738
01:18:43,673 --> 01:18:49,133
Some ratio that you're comfortable with between the two solutions, I think, is the biggest piece of advice we could share.
1739
01:18:49,133 --> 01:19:18,633
And do you think that there's something, like this was a random number generator kind of problem, like that was the crux of it. Do you think that that is a big enough sort of thing that you say, look, I'm going to be much better about my random number generation process, but then maybe I'll use it with other tools that are not, you know, may have different types of risks, but I'm not going to, I'm going to isolate the random number generation risk and use sort of other existing tools.
1740
01:19:18,633 --> 01:19:23,073
Is that a good enough way to stay in the self-custody game, do you think?
1741
01:19:23,273 --> 01:19:27,053
Or do you think it's like, oh, some other exploit can be discovered just as well
1742
01:19:27,053 --> 01:19:28,153
and some other aspects of the program?
1743
01:19:28,153 --> 01:19:32,353
I don't know. There's a lot of attack vectors, right?
1744
01:19:32,653 --> 01:19:35,893
So the one that'll be top of mind for people now is key generation.
1745
01:19:36,313 --> 01:19:42,773
So a lot of people before who weren't maybe even thinking about that attack vector
1746
01:19:42,773 --> 01:19:45,053
deeply at all
1747
01:19:45,053 --> 01:19:47,233
or considering it
1748
01:19:47,233 --> 01:19:49,553
as a similar weighted attack vector
1749
01:19:49,553 --> 01:19:50,873
as like dozens of others,
1750
01:19:51,573 --> 01:19:52,893
now they're going to be thinking...
1751
01:19:52,893 --> 01:19:53,433
Now it seems more prominent.
1752
01:19:53,673 --> 01:19:54,833
I mean, I've always thought
1753
01:19:54,833 --> 01:19:58,373
it's like one of the most important things,
1754
01:19:58,433 --> 01:19:59,413
if not the most important,
1755
01:19:59,733 --> 01:20:01,873
because, well, for the reasons we just saw,
1756
01:20:01,993 --> 01:20:03,693
like you generate it and you can't...
1757
01:20:03,693 --> 01:20:05,453
Like it's hard to change it.
1758
01:20:05,593 --> 01:20:06,413
You're just...
1759
01:20:06,413 --> 01:20:09,553
Because you're exposed to remote attacks.
1760
01:20:09,653 --> 01:20:10,633
In fact, I think we talked about
1761
01:20:10,633 --> 01:20:12,433
in that episode a year and a half ago,
1762
01:20:12,773 --> 01:20:15,193
there's so much attention
1763
01:20:15,193 --> 01:20:17,133
and some companies market this daily,
1764
01:20:17,193 --> 01:20:19,253
daily emails from some companies
1765
01:20:19,253 --> 01:20:20,513
talking about wrench attacks
1766
01:20:20,513 --> 01:20:22,393
and physical assaults and stuff.
1767
01:20:22,853 --> 01:20:24,873
And that's very scary.
1768
01:20:25,013 --> 01:20:28,413
And like none of us want that to happen to anyone.
1769
01:20:28,993 --> 01:20:29,733
It's easy to visualize.
1770
01:20:29,933 --> 01:20:30,593
It's easy to visualize,
1771
01:20:30,713 --> 01:20:32,113
but it's kind of like being worried
1772
01:20:32,113 --> 01:20:33,373
about crashing in an airplane.
1773
01:20:33,573 --> 01:20:35,213
The probability is very low.
1774
01:20:35,313 --> 01:20:36,493
There's not that many incidents.
1775
01:20:37,053 --> 01:20:38,753
What's, you're far more likely,
1776
01:20:39,333 --> 01:20:40,173
I think by far,
1777
01:20:40,253 --> 01:20:41,593
you're most likely to either
1778
01:20:41,593 --> 01:20:47,113
make a mistake yourself. I think that's the most biggest risk is you make a mistake yourself.
1779
01:20:47,493 --> 01:20:52,713
The second biggest risk is a remote attack because then the person doesn't have the person
1780
01:20:52,713 --> 01:20:56,893
could be anywhere in the world and can attack the whole world. That's much more powerful
1781
01:20:56,893 --> 01:21:02,673
than if you have to like come to everyone's home physically. It's logistically harder,
1782
01:21:02,673 --> 01:21:10,173
puts you at more risk of being caught, right? So the key generation,
1783
01:21:10,173 --> 01:21:13,093
I think is going to rise
1784
01:21:13,093 --> 01:21:14,433
in people's view of like
1785
01:21:14,433 --> 01:21:16,633
being a really critical attack vector
1786
01:21:16,633 --> 01:21:18,213
and you want a solid solution there.
1787
01:21:18,753 --> 01:21:19,493
I think it's one of these things
1788
01:21:19,493 --> 01:21:20,973
like a wrench attack is very,
1789
01:21:21,133 --> 01:21:21,973
you can visualize it.
1790
01:21:22,033 --> 01:21:22,953
You know what that feels like.
1791
01:21:23,033 --> 01:21:23,993
You can imagine what it looks like.
1792
01:21:24,633 --> 01:21:25,613
Random number generation,
1793
01:21:25,953 --> 01:21:27,373
we have a hard time recognizing
1794
01:21:27,373 --> 01:21:28,573
how bad we are
1795
01:21:28,573 --> 01:21:29,813
at generating random numbers.
1796
01:21:29,933 --> 01:21:31,633
They're just like really hard to generate.
1797
01:21:31,753 --> 01:21:31,973
It's not like,
1798
01:21:32,053 --> 01:21:33,213
oh, I'll just make up a few things,
1799
01:21:33,573 --> 01:21:34,673
but those are not actually random.
1800
01:21:34,993 --> 01:21:36,793
We're just horrible at making random things
1801
01:21:36,793 --> 01:21:37,913
even though we think they're random.
1802
01:21:38,373 --> 01:21:39,533
I think it's just hard to visualize
1803
01:21:39,533 --> 01:21:42,533
and digest how tough that piece of it is.
1804
01:21:43,033 --> 01:21:46,053
Steve, I don't know how much you can speak to this,
1805
01:21:46,433 --> 01:21:48,693
but specifically you mentioned,
1806
01:21:48,813 --> 01:21:50,553
okay, we want self-custody to keep thriving.
1807
01:21:52,113 --> 01:21:55,593
What are your thoughts on these other,
1808
01:21:55,673 --> 01:21:57,293
the major hardware wall players?
1809
01:21:57,413 --> 01:22:01,913
Obviously, you know more about BitKey, Ledger, Trezor.
1810
01:22:02,273 --> 01:22:03,593
What is your advice?
1811
01:22:03,653 --> 01:22:05,133
If someone's listening to this and saying,
1812
01:22:05,133 --> 01:22:08,113
hey, I got a Trezor, I got a BitKey.
1813
01:22:08,113 --> 01:22:09,793
how should they evaluate
1814
01:22:09,793 --> 01:22:13,553
when I think about AI getting a lot better
1815
01:22:13,553 --> 01:22:15,893
is this device going to be safe
1816
01:22:15,893 --> 01:22:17,113
in a week, in a month, in a year
1817
01:22:17,113 --> 01:22:19,553
versus should I just go with it
1818
01:22:19,553 --> 01:22:21,293
because I think a lot of people are in that boat right now.
1819
01:22:21,573 --> 01:22:24,113
So I think something that was true before
1820
01:22:24,113 --> 01:22:26,313
true two days ago
1821
01:22:26,313 --> 01:22:27,373
still true today
1822
01:22:27,373 --> 01:22:32,753
is multi-sig is strong and powerful.
1823
01:22:33,753 --> 01:22:35,853
And when I say multi-sig
1824
01:22:35,853 --> 01:22:37,313
that could be like a two of three
1825
01:22:37,313 --> 01:22:40,773
or three of five, or you have like three keys or five keys
1826
01:22:40,773 --> 01:22:43,933
and you need two of three to sign.
1827
01:22:43,933 --> 01:22:44,293
Yeah.
1828
01:22:44,513 --> 01:22:45,573
Or three of five to sign.
1829
01:22:46,013 --> 01:22:49,553
Now, what's very important about such a setup, though,
1830
01:22:49,633 --> 01:22:52,913
is that you have diversity across your different keys.
1831
01:22:53,733 --> 01:22:55,533
And I think Coldcard, if I recall,
1832
01:22:55,733 --> 01:22:58,213
I hadn't used a Coldcard for several years,
1833
01:22:58,313 --> 01:23:01,253
but I think they had a multi-sig feature built in
1834
01:23:01,253 --> 01:23:05,093
and even like a product, a packaged product
1835
01:23:05,093 --> 01:23:07,153
where you bought like five Coldcards.
1836
01:23:07,313 --> 01:23:08,673
for your multi-sig.
1837
01:23:09,133 --> 01:23:10,273
Now, I think that's,
1838
01:23:10,773 --> 01:23:12,293
and I thought at the time too,
1839
01:23:12,433 --> 01:23:13,333
that's a dumb idea.
1840
01:23:13,593 --> 01:23:14,833
Because then you get no diversity.
1841
01:23:15,113 --> 01:23:16,513
So if you have one vulnerability
1842
01:23:16,513 --> 01:23:18,153
in a particular product,
1843
01:23:18,393 --> 01:23:20,973
and this is not a slight on cold card,
1844
01:23:21,133 --> 01:23:22,953
this would apply to any company
1845
01:23:22,953 --> 01:23:23,653
or any product.
1846
01:23:23,653 --> 01:23:25,553
If you're only using
1847
01:23:25,553 --> 01:23:26,733
that one company's product,
1848
01:23:27,013 --> 01:23:28,853
especially if it's the same model
1849
01:23:28,853 --> 01:23:29,813
and same software,
1850
01:23:30,473 --> 01:23:33,093
then you eliminate
1851
01:23:33,093 --> 01:23:34,973
the vast majority of value of multi-sig.
1852
01:23:35,373 --> 01:23:37,113
So having diversity of hardware
1853
01:23:37,113 --> 01:23:41,233
and software across the keys is really important
1854
01:23:41,233 --> 01:23:43,253
and then really valuable too.
1855
01:23:43,653 --> 01:23:44,513
Because in such a setup,
1856
01:23:44,913 --> 01:23:46,513
if one of your five keys
1857
01:23:46,513 --> 01:23:48,733
was one of these vulnerable cold cards,
1858
01:23:48,933 --> 01:23:50,993
you'd be perfectly safe, right?
1859
01:23:51,173 --> 01:23:55,713
So multi-sig remains very safe and secure.
1860
01:23:56,153 --> 01:23:59,753
Now, historically, it's been complicated to set up.
1861
01:24:00,433 --> 01:24:02,573
It's much more, if you're going to do it yourself
1862
01:24:02,573 --> 01:24:05,693
or, yeah, if you're going to do it yourself
1863
01:24:05,693 --> 01:24:07,233
Or use a service like Unchained.
1864
01:24:07,453 --> 01:24:07,673
Right.
1865
01:24:08,513 --> 01:24:09,693
But if you do it yourself,
1866
01:24:09,773 --> 01:24:10,733
it's definitely more complicated
1867
01:24:10,733 --> 01:24:12,553
than doing Signal Signature.
1868
01:24:14,053 --> 01:24:14,753
And so,
1869
01:24:15,273 --> 01:24:16,913
and I don't think a lot of people do that.
1870
01:24:17,313 --> 01:24:18,513
So I always get a little nervous
1871
01:24:18,513 --> 01:24:19,693
using a solution
1872
01:24:19,693 --> 01:24:21,733
that few other people do.
1873
01:24:21,833 --> 01:24:22,033
Yeah.
1874
01:24:22,313 --> 01:24:23,193
Because then it's novel.
1875
01:24:23,493 --> 01:24:23,933
Fewer eyeballs.
1876
01:24:24,413 --> 01:24:26,173
Fewer eyeballs on the process, right?
1877
01:24:26,173 --> 01:24:28,613
And if you're developing your own process,
1878
01:24:28,913 --> 01:24:31,973
you better be an expert and diligent.
1879
01:24:32,433 --> 01:24:33,653
And even then, it's risky.
1880
01:24:33,653 --> 01:24:36,813
So then, like you said, okay, great.
1881
01:24:37,053 --> 01:24:39,833
There's services like Unchained, Kasa,
1882
01:24:41,153 --> 01:24:46,293
like Nunchuck is a software wallet that supports multi-sig.
1883
01:24:47,173 --> 01:24:49,733
And there's other products too, but all of those,
1884
01:24:49,973 --> 01:24:52,533
they're all different, but they all help you
1885
01:24:52,533 --> 01:24:54,233
and assist you along the way.
1886
01:24:55,073 --> 01:24:59,513
And then there's BitKey, which is multi-sig by default.
1887
01:24:59,513 --> 01:25:01,753
So I believe it's the only product in the market
1888
01:25:01,753 --> 01:25:03,953
that is multi-sig out of the box.
1889
01:25:05,353 --> 01:25:06,793
Like Kasa and Unchained,
1890
01:25:06,893 --> 01:25:07,813
I mean, you have to go,
1891
01:25:08,173 --> 01:25:09,673
they can ship you these products,
1892
01:25:09,773 --> 01:25:13,533
but it's a collection of different vendors' products
1893
01:25:13,533 --> 01:25:15,693
that you're kind of mish-mashing together.
1894
01:25:16,593 --> 01:25:20,813
And their software helps you do that.
1895
01:25:21,873 --> 01:25:24,713
With BitKey, there's pros and cons to the approach.
1896
01:25:25,353 --> 01:25:26,293
The con is it's one company.
1897
01:25:26,533 --> 01:25:27,233
It's one company.
1898
01:25:27,233 --> 01:25:29,733
So you lose some of the diversification benefit.
1899
01:25:30,113 --> 01:25:30,133
Right.
1900
01:25:30,133 --> 01:25:37,393
But the simplicity in UX, no one questions it's by far the easiest product to use.
1901
01:25:37,633 --> 01:25:46,673
The main criticism or thing you'd want to evaluate with Viki is the security, privacy, etc., both of which have improved tremendously over the years.
1902
01:25:47,173 --> 01:25:53,913
And my opinion is that the overall safety value is high.
1903
01:25:53,913 --> 01:25:59,473
because part of safety is not just like raw security,
1904
01:25:59,633 --> 01:26:01,253
but are you using the product correctly?
1905
01:26:01,773 --> 01:26:05,093
In elimination of foot guns, BitKey does very, very well at.
1906
01:26:05,233 --> 01:26:09,273
So that user, if someone agrees with like user error
1907
01:26:09,273 --> 01:26:11,233
is probably the biggest risk to themselves,
1908
01:26:11,353 --> 01:26:12,533
like they're going to make a mistake.
1909
01:26:12,533 --> 01:26:16,373
If they're like egoless enough to admit
1910
01:26:16,373 --> 01:26:17,753
that they might make a mistake,
1911
01:26:18,273 --> 01:26:20,513
then BitKey is a big win there.
1912
01:26:21,373 --> 01:26:23,473
And then if you look at all the traditional metrics
1913
01:26:23,473 --> 01:26:26,493
of security. BitKey does really well on those
1914
01:26:26,493 --> 01:26:29,353
too. It's not without weaknesses though.
1915
01:26:29,353 --> 01:26:32,593
And I tell the team this, like key generation
1916
01:26:32,593 --> 01:26:35,033
is an area that
1917
01:26:35,033 --> 01:26:38,573
could improve. The good news is
1918
01:26:38,573 --> 01:26:40,513
that there's three keys.
1919
01:26:41,393 --> 01:26:44,413
And each of the keys are generated by a different piece
1920
01:26:44,413 --> 01:26:47,313
of hardware and software stack. So you do get
1921
01:26:47,313 --> 01:26:50,293
that diversity. Because one is generated on a block
1922
01:26:50,293 --> 01:26:52,313
server using
1923
01:26:52,313 --> 01:26:54,173
a suite of software.
1924
01:26:54,473 --> 01:26:56,533
Another is generated on your phone using a different
1925
01:26:56,533 --> 01:26:58,633
set of software on a phone.
1926
01:26:58,853 --> 01:27:00,713
And the third is generated on the
1927
01:27:00,713 --> 01:27:02,613
big-key hardware device using a different set
1928
01:27:02,613 --> 01:27:04,533
of software. So if
1929
01:27:04,533 --> 01:27:06,513
there's a vulnerability in one
1930
01:27:06,513 --> 01:27:08,193
piece of software or one piece of hardware,
1931
01:27:09,093 --> 01:27:10,673
you're still safe.
1932
01:27:11,493 --> 01:27:12,513
It would require
1933
01:27:12,513 --> 01:27:14,413
multiple vulnerabilities, but
1934
01:27:14,413 --> 01:27:16,573
it's possible, but it's just far, far
1935
01:27:16,573 --> 01:27:18,573
less likely. But
1936
01:27:18,573 --> 01:27:19,613
being a single vendor,
1937
01:27:19,613 --> 01:27:23,253
if there's some kind of rogue
1938
01:27:23,253 --> 01:27:25,473
employee or external
1939
01:27:25,473 --> 01:27:27,113
attack into the company,
1940
01:27:27,433 --> 01:27:29,532
you could insert yourself into the overall
1941
01:27:29,532 --> 01:27:30,853
process and potentially
1942
01:27:30,853 --> 01:27:33,253
modify multiple
1943
01:27:33,253 --> 01:27:35,373
pieces of that puzzle. So there's
1944
01:27:35,373 --> 01:27:37,813
that type of risk. But the
1945
01:27:37,813 --> 01:27:39,433
key generation part that
1946
01:27:39,433 --> 01:27:41,532
I'd like to see improved is it's not
1947
01:27:41,532 --> 01:27:43,073
independently verifiable right now.
1948
01:27:44,532 --> 01:27:45,613
And the...
1949
01:27:45,613 --> 01:27:46,913
How is that that you should resolve loss?
1950
01:27:47,693 --> 01:27:58,684
Well the key in the server is running on a server Got it okay It is possible to run on a server And we talked about like you know like you know remote attestable software
1951
01:27:58,684 --> 01:27:59,204
on a server.
1952
01:27:59,304 --> 01:27:59,884
That's possible.
1953
01:28:00,104 --> 01:28:01,164
So they can do that.
1954
01:28:01,204 --> 01:28:03,664
But it's right now you can't independently know
1955
01:28:03,664 --> 01:28:06,564
this software was run on that server
1956
01:28:06,564 --> 01:28:07,504
to generate my key.
1957
01:28:07,644 --> 01:28:07,944
Got it.
1958
01:28:08,004 --> 01:28:08,704
You can't do that yet.
1959
01:28:10,044 --> 01:28:12,784
And then on the phone, on an Android phone,
1960
01:28:12,904 --> 01:28:14,344
yes, it's a reproducible build.
1961
01:28:14,344 --> 01:28:15,964
You can do that.
1962
01:28:15,964 --> 01:28:19,804
You yourself, or if there's a security researcher you trust,
1963
01:28:20,144 --> 01:28:21,484
like Wall of Scrutiny or whatever,
1964
01:28:23,264 --> 01:28:25,804
and they have reviewed and approved
1965
01:28:25,804 --> 01:28:28,224
that they've independently verified that software.
1966
01:28:29,564 --> 01:28:32,244
On iPhone, you can't do that because of Apple.
1967
01:28:32,724 --> 01:28:33,984
You could jailbreak your iPhone.
1968
01:28:34,084 --> 01:28:36,744
So it's possible that a security researcher could jailbreak an iPhone
1969
01:28:36,744 --> 01:28:39,184
and prove that, yes, it's also reproducible there.
1970
01:28:39,504 --> 01:28:42,084
So the phones, I'd give that a thumbs up.
1971
01:28:42,084 --> 01:28:43,984
But then on the big key device itself,
1972
01:28:43,984 --> 01:28:46,464
yes, all the software Block has written is open source.
1973
01:28:47,424 --> 01:28:50,964
But the fingerprint sensor is a third-party vendor,
1974
01:28:51,164 --> 01:28:53,244
and sadly, they have contracts with the government,
1975
01:28:53,644 --> 01:28:55,644
and they can't open source their software.
1976
01:28:56,544 --> 01:29:00,524
Even if they agree with Block and want to open source it,
1977
01:29:00,564 --> 01:29:03,644
they can't because of government contracts, security contracts.
1978
01:29:04,184 --> 01:29:09,384
So it's possible that there could be malware
1979
01:29:09,384 --> 01:29:11,504
in the fingerprint sensor software
1980
01:29:11,504 --> 01:29:14,584
that could impact the key generation.
1981
01:29:16,664 --> 01:29:18,824
So anyway, there's trade-offs
1982
01:29:18,824 --> 01:29:20,664
with all these different products,
1983
01:29:20,804 --> 01:29:23,084
but I mean, getting back to advice to people
1984
01:29:23,084 --> 01:29:24,464
who still want to self-custody,
1985
01:29:24,544 --> 01:29:25,724
I don't think self-custody is dead.
1986
01:29:26,264 --> 01:29:28,764
Multi-sig has legs still,
1987
01:29:28,824 --> 01:29:30,244
and I think will have legs forever,
1988
01:29:30,484 --> 01:29:32,904
and it keeps getting better user experience.
1989
01:29:34,844 --> 01:29:41,464
And then doing offline entropy generation
1990
01:29:41,464 --> 01:29:44,604
for key generation or using offline
1991
01:29:44,604 --> 01:29:47,324
entropy for key generation is, I think,
1992
01:29:47,684 --> 01:29:48,504
super important.
1993
01:29:49,624 --> 01:29:53,544
Is that for everyone? That's a harder call, right?
1994
01:29:53,664 --> 01:29:56,224
Do people have the time
1995
01:29:56,224 --> 01:29:59,424
and patience to learn the process and then the diligence
1996
01:29:59,424 --> 01:30:02,504
to follow through in that process in a safe, secure
1997
01:30:02,504 --> 01:30:02,824
way?
1998
01:30:04,984 --> 01:30:08,164
For people who have a lot of Bitcoin, I hope the answer is yes
1999
01:30:08,164 --> 01:30:11,384
because it's important enough to them, but unclear. But for those that
2000
01:30:11,464 --> 01:30:14,964
can commit the time and diligence and patience,
2001
01:30:15,404 --> 01:30:17,144
then you can generate your own entropy.
2002
01:30:17,944 --> 01:30:19,564
And if you're confident in that process,
2003
01:30:19,704 --> 01:30:22,144
you can sleep much better at night
2004
01:30:22,144 --> 01:30:26,324
than solely using some third party.
2005
01:30:26,424 --> 01:30:27,824
Because even if the software is open source,
2006
01:30:27,944 --> 01:30:30,444
as we just learned, it could have bugs.
2007
01:30:30,864 --> 01:30:32,624
And the vast majority of people
2008
01:30:32,624 --> 01:30:34,464
can't go inspect the code themselves
2009
01:30:34,464 --> 01:30:36,384
and effectively audit it.
2010
01:30:37,144 --> 01:30:38,024
I mean, I guess...
2011
01:30:38,024 --> 01:30:38,784
They could have their agents do it.
2012
01:30:38,784 --> 01:30:40,104
They could have their agents do it now.
2013
01:30:40,104 --> 01:30:41,884
the only thing
2014
01:30:41,884 --> 01:30:44,044
like I think we're quickly moving
2015
01:30:44,044 --> 01:30:45,584
into a world where
2016
01:30:45,584 --> 01:30:48,264
the vendors themselves or project loops of the world
2017
01:30:48,264 --> 01:30:50,144
or other like security experts
2018
01:30:50,144 --> 01:30:51,824
are going to be scanning all the software
2019
01:30:51,824 --> 01:30:54,144
with LLMs so there's no marginal value
2020
01:30:54,144 --> 01:30:55,844
to a retail customer
2021
01:30:55,844 --> 01:30:57,404
doing that. I agree.
2022
01:30:58,504 --> 01:30:59,344
It's good to have the option.
2023
01:30:59,604 --> 01:31:00,244
Yeah so
2024
01:31:00,244 --> 01:31:03,784
I think those are the
2025
01:31:03,784 --> 01:31:05,664
I have one other question which
2026
01:31:05,664 --> 01:31:08,204
I guess is kind of probably the best thing I was reading here
2027
01:31:08,204 --> 01:31:09,404
and I know this will be less popular
2028
01:31:09,404 --> 01:31:11,364
but still I think it's just worth asking.
2029
01:31:11,484 --> 01:31:12,884
If someone wanted to get the custodial work,
2030
01:31:13,024 --> 01:31:15,524
we talked about the rivers, coin bases, squares,
2031
01:31:15,764 --> 01:31:16,724
Krakens, whatever of the world.
2032
01:31:17,584 --> 01:31:19,404
If they wanted to say, you know what,
2033
01:31:19,524 --> 01:31:21,564
I've had enough of this for now during the transition,
2034
01:31:21,704 --> 01:31:22,464
I'm ready to go IBIT.
2035
01:31:23,084 --> 01:31:24,824
Do we know, have the rules changed
2036
01:31:24,824 --> 01:31:28,344
on in-kind transfer of BTC to IBIT
2037
01:31:28,344 --> 01:31:30,524
and what the minimum threshold you need is now
2038
01:31:30,524 --> 01:31:31,564
to do that non-taxably?
2039
01:31:31,944 --> 01:31:32,464
I don't believe so.
2040
01:31:32,544 --> 01:31:34,684
It's not, no, it's only authorized participants.
2041
01:31:34,824 --> 01:31:35,624
Which is still like...
2042
01:31:35,624 --> 01:31:36,244
Like Jane Street.
2043
01:31:36,604 --> 01:31:38,664
So that's like tens of millions of dollars or something.
2044
01:31:38,664 --> 01:31:52,224
Well, yeah, you have to be an authorized participant, which is a separate role from retail. So I continue to be optimistic that this will happen over some timeframe. But I think there's nothing that suggests it's imminent.
2045
01:31:52,304 --> 01:32:08,104
So if a normal retail user wanted to, I'm assuming if you had a large enough stack, you potentially go through the qualified purchase, maybe. But for most retail users, if they wanted to say, I want to set this out during the AI transition, you would have to sell and then rebuy and take the taxable event.
2046
01:32:08,104 --> 01:32:08,504
Yeah.
2047
01:32:09,064 --> 01:32:09,284
Yeah.
2048
01:32:09,944 --> 01:32:10,184
Okay.
2049
01:32:10,504 --> 01:32:10,864
That's our job.
2050
01:32:11,844 --> 01:32:16,784
Well, and why not just move your coins to Coinbase?
2051
01:32:17,064 --> 01:32:18,624
That's where they're going to sit anyway,
2052
01:32:18,964 --> 01:32:19,884
if you buy any ETF.
2053
01:32:20,204 --> 01:32:21,004
That's a good point, actually.
2054
01:32:21,284 --> 01:32:22,404
Yeah, why do they need a level of indirect?
2055
01:32:22,764 --> 01:32:22,944
Yeah.
2056
01:32:23,604 --> 01:32:25,564
And, you know, or River, Cashflow, or others.
2057
01:32:25,664 --> 01:32:27,644
But if you literally want the same protection,
2058
01:32:28,084 --> 01:32:28,944
just move them to Coinbase.
2059
01:32:29,004 --> 01:32:30,804
That's where they're stored if you buy it.
2060
01:32:30,804 --> 01:32:32,724
I think the reason they would say this,
2061
01:32:32,744 --> 01:32:35,624
and I think it's a fair question,
2062
01:32:35,624 --> 01:32:37,504
the answer hopefully is the same for Coinbase,
2063
01:32:37,504 --> 01:32:42,384
So as an example, obviously, you'd much rather have your coins back.
2064
01:32:42,524 --> 01:32:52,644
But let's say in a case where if Coinbase got hacked, in theory, they would figure out some way to pay back at least part of the fiat value of that.
2065
01:32:53,064 --> 01:32:55,044
I bet potentially more if it's BlackRock backed.
2066
01:32:55,344 --> 01:32:59,304
So I think you're also thinking not only about the technological piece, but also the sort of the...
2067
01:32:59,304 --> 01:32:59,944
Something political.
2068
01:33:00,404 --> 01:33:04,724
Basically, these guys are too big to fail, which I know is the whole reason Bitcoin was created in the first place was to stop that.
2069
01:33:04,724 --> 01:33:06,784
But is Coinbase too big to fail yet by itself?
2070
01:33:06,864 --> 01:33:07,224
I don't know.
2071
01:33:07,224 --> 01:33:12,904
maybe maybe not ibit and blackrock yes would be my guess i'm just saying these are all different
2072
01:33:12,904 --> 01:33:17,064
factors that people probably think are more money just print more money that's what they'll do
2073
01:33:18,264 --> 01:33:23,864
it's a cynical take but coinbase has two million coins hacked and it's like
2074
01:33:24,824 --> 01:33:28,984
200 billion dollars or something it's probably too big to fail but i mean blackrock we know is
2075
01:33:28,984 --> 01:33:33,224
right so probably i don't know i'm just saying that that might be one reason but black i mean
2076
01:33:33,224 --> 01:33:34,964
I mean, this is a small portion of their business.
2077
01:33:35,164 --> 01:33:35,764
Yeah, exactly.
2078
01:33:36,244 --> 01:33:37,464
So they could just be like,
2079
01:33:37,844 --> 01:33:39,624
like Bitcoin go away and then move on.
2080
01:33:39,764 --> 01:33:41,164
But you don't think they would make their users whole,
2081
01:33:41,284 --> 01:33:41,784
at least in fiat?
2082
01:33:42,184 --> 01:33:43,964
If they got, if let's say Coinbase got hacked,
2083
01:33:44,384 --> 01:33:45,064
their underlying shares.
2084
01:33:45,084 --> 01:33:45,324
I don't know.
2085
01:33:45,384 --> 01:33:49,264
I'm skeptical US taxpayers would bail out BlackRock
2086
01:33:49,264 --> 01:33:51,004
for their Bitcoin customers.
2087
01:33:51,365 --> 01:33:53,084
Because BlackRock's not going to fail.
2088
01:33:53,164 --> 01:33:53,824
Right, that's what I'm saying.
2089
01:33:53,924 --> 01:33:55,504
So BlackRock may not even need that piece.
2090
01:33:55,504 --> 01:33:58,604
They may be able just to like other funds somehow.
2091
01:34:00,324 --> 01:34:01,724
Or just be like, sorry.
2092
01:34:01,904 --> 01:34:02,344
Or just be sorry.
2093
01:34:02,344 --> 01:34:05,564
I mean, I'm sure it's in their perspective
2094
01:34:05,564 --> 01:34:08,324
or you can read the terms and it says this is a risk, right?
2095
01:34:09,664 --> 01:34:10,964
I'm not saying they're going to do it either way,
2096
01:34:11,024 --> 01:34:12,664
but I'm saying it's a lot of incentives.
2097
01:34:12,664 --> 01:34:15,564
I would guess if I'm then, I'd look at like,
2098
01:34:15,664 --> 01:34:20,084
well, if we think Bitcoin still has legs
2099
01:34:20,084 --> 01:34:21,484
and we want to be in this business
2100
01:34:21,484 --> 01:34:24,004
and it's going to cost us a few million dollars
2101
01:34:24,004 --> 01:34:24,904
and it's not that bad,
2102
01:34:25,084 --> 01:34:26,764
yeah, let's make our customers whole
2103
01:34:26,764 --> 01:34:28,664
or make them happy or give them pennies and a dollar,
2104
01:34:29,004 --> 01:34:29,684
do it, that's right.
2105
01:34:29,784 --> 01:34:32,224
But if it looks like Bitcoin's over
2106
01:34:32,224 --> 01:34:34,104
with and it's going to cost us
2107
01:34:34,104 --> 01:34:35,164
I doubt that would be the case.
2108
01:34:35,324 --> 01:34:38,184
$100 billion, then I think they just conclude
2109
01:34:38,184 --> 01:34:39,384
like, sorry.
2110
01:34:40,044 --> 01:34:42,144
Do we know which, like I know
2111
01:34:42,144 --> 01:34:43,944
that Coinbase is a major
2112
01:34:43,944 --> 01:34:46,204
custodian for a lot of the ETFs.
2113
01:34:46,464 --> 01:34:48,184
My understanding is some of those are moving
2114
01:34:48,184 --> 01:34:50,384
towards more diversified underlying
2115
01:34:50,384 --> 01:34:52,264
custodians, but do we know much about
2116
01:34:52,264 --> 01:34:54,384
where like BitGo
2117
01:34:54,384 --> 01:34:56,264
or Paxos or
2118
01:34:56,264 --> 01:34:58,064
Anchorage may show up in
2119
01:34:58,064 --> 01:35:00,044
other exchanges? They kind of provide
2120
01:35:00,044 --> 01:35:01,404
more diversification.
2121
01:35:02,224 --> 01:35:06,384
which exchanges use Paxos or Bitco?
2122
01:35:06,524 --> 01:35:07,664
I don't know.
2123
01:35:08,044 --> 01:35:09,365
Because that would be another way that you...
2124
01:35:09,365 --> 01:35:13,144
I do remember back in 2017,
2125
01:35:14,604 --> 01:35:16,244
I think at Curious Conversations,
2126
01:35:16,524 --> 01:35:20,084
some exchange owner was there
2127
01:35:20,084 --> 01:35:23,284
and had a ledger necklace around his neck.
2128
01:35:23,584 --> 01:35:26,004
And he's like, yeah, these are the keys to the exchange.
2129
01:35:26,164 --> 01:35:26,444
And I'm like...
2130
01:35:26,444 --> 01:35:26,944
Are you serious?
2131
01:35:27,264 --> 01:35:28,044
He was joking.
2132
01:35:29,404 --> 01:35:31,264
You're not talking about the guy you got,
2133
01:35:31,264 --> 01:35:38,024
is it no that was a no that's a different guess yeah anyway i hope we're beyond those days i hope
2134
01:35:38,024 --> 01:35:41,244
we're beyond those days and again like i do think you know just to be fair when people are always
2135
01:35:41,244 --> 01:35:44,744
like well the exchanges like they've all gone on it's like true but like the exchanges have also
2136
01:35:44,744 --> 01:35:50,464
come a long way since mount gox and like even 2017 era like it's it's it's a different breed now
2137
01:35:50,464 --> 01:35:55,424
um that but again i i do think and i think we called for this you know way back in the the
2138
01:35:55,424 --> 01:35:59,624
custody episode but some kind of wallet screening like project which i think you launched or someone
2139
01:35:59,624 --> 01:36:00,884
that you helped launch
2140
01:36:00,884 --> 01:36:02,244
to look at the different exchanges.
2141
01:36:03,944 --> 01:36:04,784
And I think at that time...
2142
01:36:04,784 --> 01:36:05,364
The Wall of Scrutiny team
2143
01:36:05,364 --> 01:36:06,924
put together like a prototype.
2144
01:36:07,284 --> 01:36:07,464
Yeah.
2145
01:36:07,604 --> 01:36:08,924
I don't know if that ever...
2146
01:36:08,924 --> 01:36:10,144
This is like exchange scrutiny?
2147
01:36:10,424 --> 01:36:10,764
Yeah.
2148
01:36:10,864 --> 01:36:11,684
And it would be super cool
2149
01:36:11,684 --> 01:36:12,624
to see that come back
2150
01:36:12,624 --> 01:36:13,924
with even more detail now.
2151
01:36:14,004 --> 01:36:14,604
So it's like, all right,
2152
01:36:14,684 --> 01:36:16,624
what exactly for your cold storage setup?
2153
01:36:16,884 --> 01:36:18,324
Is it just you doing self-custody?
2154
01:36:18,404 --> 01:36:19,865
Do you have other exchanges brought in?
2155
01:36:19,984 --> 01:36:20,764
What kind of models
2156
01:36:20,764 --> 01:36:21,724
are you getting access to?
2157
01:36:21,824 --> 01:36:23,044
How frequently are you running them?
2158
01:36:23,164 --> 01:36:24,104
Like, these are all things
2159
01:36:24,104 --> 01:36:25,544
I would be very interested to see.
2160
01:36:25,764 --> 01:36:25,924
Yeah.
2161
01:36:26,784 --> 01:36:27,824
Yeah, I do think that,
2162
01:36:27,924 --> 01:36:29,264
I mean, your call for like
2163
01:36:29,624 --> 01:36:31,744
services and products and brands to
2164
01:36:31,744 --> 01:36:34,344
be transparent, mark what they have.
2165
01:36:34,424 --> 01:36:35,924
I mean, it's similar to proof of reserves.
2166
01:36:36,204 --> 01:36:38,524
Proof of reserves is also not bulletproof.
2167
01:36:38,905 --> 01:36:40,405
There's gaps in it, but
2168
01:36:40,405 --> 01:36:42,624
everything else being equal,
2169
01:36:43,004 --> 01:36:44,264
I'm going to choose the vendor
2170
01:36:44,264 --> 01:36:46,304
that has that service and
2171
01:36:46,304 --> 01:36:47,644
describes how they use it.
2172
01:36:48,224 --> 01:36:49,865
It's definitely a
2173
01:36:49,865 --> 01:36:50,844
win.
2174
01:36:52,664 --> 01:36:53,804
Anything else to cover?
2175
01:36:53,804 --> 01:36:55,944
I think, you know, as I think NVK
2176
01:36:55,944 --> 01:36:57,885
is famous for saying, secure your Bitcoin
2177
01:36:57,885 --> 01:37:00,064
like it's worth 10 times as much as it is
2178
01:37:00,064 --> 01:37:01,044
because someday it might be.
2179
01:37:02,144 --> 01:37:03,885
Yeah, I mean, I know we didn't get to today.
2180
01:37:04,084 --> 01:37:05,284
There were a lot of other crazy things
2181
01:37:05,284 --> 01:37:05,824
happening in the world.
2182
01:37:06,044 --> 01:37:07,684
So some, you know,
2183
01:37:07,905 --> 01:37:09,304
while there's been turbulence in Bitcoin,
2184
01:37:09,464 --> 01:37:09,844
believe it or not,
2185
01:37:09,865 --> 01:37:11,584
there's also been turbulence in the AI world too.
2186
01:37:11,704 --> 01:37:13,244
So maybe tune in next week.
2187
01:37:13,264 --> 01:37:14,024
We'll have more to say on that.
2188
01:37:14,344 --> 01:37:14,824
Yeah. All right.
2189
01:37:15,365 --> 01:37:15,684
Thank you.
2190
01:37:15,684 --> 01:37:16,365
Yeah, thank you.
2191
01:37:16,464 --> 01:37:16,524
Bye.