110 - The Singularity Stack: Energy, AI Agents, & Bitcoin Settlement with Liam Nelson

1
00:00:00,840 --> 00:00:04,770
This isn't for everyone. It's for the ones who know the
2
00:00:04,920 --> 00:00:07,590
old system is broken and are ready to build something that
3
00:00:07,680 --> 00:00:22,400
lasts. Welcome to the Build with Bitcoin podcast. We're co-host
4
00:00:22,490 --> 00:00:25,430
Iraiden Lynne, and today we're doing a deep dive with Early
5
00:00:25,550 --> 00:00:28,370
Writers partner Liam Nelson. We're going to get into your
6
00:00:28,850 --> 00:00:31,760
research report. We'll be unpacking everything called the
7
00:00:31,790 --> 00:00:35,240
Singularity Stack. But first of all, welcome to the podcast, and
8
00:00:35,450 --> 00:00:36,110
thanks for joining us.
9
00:00:36,230 --> 00:00:38,720
Thanks for having me. First time, long time. Really
10
00:00:38,750 --> 00:00:41,320
appreciate the show and all the knowledge that you guys have put
11
00:00:41,410 --> 00:00:42,910
out over the years, so thanks for having me.
12
00:00:43,059 --> 00:00:45,459
We we appreciate you guys, and early congrats on the
13
00:00:45,549 --> 00:00:47,739
report because it's a fascinating research project
14
00:00:47,769 --> 00:00:52,299
that you guys put out. And before that, let's maybe get to
15
00:00:52,359 --> 00:00:54,819
know you a little bit. I mean, what can you share about your
16
00:00:54,879 --> 00:00:57,279
background, Liam, and what brought you to Bitcoin?
17
00:00:57,399 --> 00:01:01,019
Yeah, for sure. Just really quick background on
18
00:01:01,019 --> 00:01:04,859
was finance major at Georgetown Business School, and
19
00:01:05,549 --> 00:01:09,449
then started off my career at Point 72. For those who don't
20
00:01:09,479 --> 00:01:13,589
know, it's founded by Steve Cohen, and they are about a $40
21
00:01:13,649 --> 00:01:17,579
billion long short hedge fund, primarily trading U.S. equities.
22
00:01:18,179 --> 00:01:22,579
And I was there from 2019 to 2022, and the time there was
23
00:01:22,669 --> 00:01:26,449
particularly interesting because I was able to join for all the
24
00:01:26,599 --> 00:01:29,719
issues related to tariffs, and then was covering a lot of the
25
00:01:29,899 --> 00:01:34,309
consumer and industrial sectors during COVID too, and so was
26
00:01:34,429 --> 00:01:36,979
able to to cover companies like Walmart, Target, all the
27
00:01:37,069 --> 00:01:40,449
consumer goods. Is there was issues related to toilet paper
28
00:01:40,539 --> 00:01:44,109
running out, and and then got to be able to see everything from
29
00:01:44,229 --> 00:01:48,939
the run up after that too. As there, I was doing deep dives on
30
00:01:49,329 --> 00:01:52,269
on businesses and companies, and seeing the fundamentals be
31
00:01:52,509 --> 00:01:56,109
completely dislocated from the actual valuations of the stock
32
00:01:56,139 --> 00:02:00,419
prices. And so during that time, both both myself trying to
33
00:02:00,419 --> 00:02:02,999
understand what was going on from an individual level, as
34
00:02:03,059 --> 00:02:06,629
well as the firm, really was going deep into the macro side
35
00:02:06,659 --> 00:02:09,899
of things too, and obviously trying to understand what's
36
00:02:09,959 --> 00:02:13,949
going on with the Fed, monetary policy, etc. And that's a lot of
37
00:02:14,069 --> 00:02:17,879
what led me to find Bitcoin during my time there. Point 72
38
00:02:18,059 --> 00:02:21,739
had an average hold period of only 33 days, so incredibly
39
00:02:21,889 --> 00:02:27,229
short and very thesis in momentum and like event-driven
40
00:02:27,349 --> 00:02:31,849
firm for holding stocks, and so it was a fascinating time to be
41
00:02:32,029 --> 00:02:34,699
there, where there was no shortage of events going on
42
00:02:34,999 --> 00:02:39,319
around that time, and we actually worked and funded
43
00:02:39,829 --> 00:02:43,119
Melvin Capital on the wrong side of the GameStop trade, and so
44
00:02:43,899 --> 00:02:48,339
we-I wasn't personally in there, but we, as a firm, we were
45
00:02:48,399 --> 00:02:50,829
pretty close, and so that was just the point that I was like,
46
00:02:51,069 --> 00:02:53,769
"What the hell is going on with the entire financial system?
47
00:02:53,859 --> 00:02:57,219
There's some something else going on at play, and and really
48
00:02:57,309 --> 00:03:01,199
found Bitcoin deeply tried to you know spend all of my waking
49
00:03:01,379 --> 00:03:04,469
hours outside of the firm to go deep there and understand what
50
00:03:04,469 --> 00:03:07,739
was going on, and and as we all kind of come out the other side,
51
00:03:07,799 --> 00:03:11,399
understanding where this all goes in the long term. That was
52
00:03:11,489 --> 00:03:15,959
my kind of moment there. I spent a few years after that in
53
00:03:16,049 --> 00:03:19,619
private equity at a firm called Kobe Capital, and then joined
54
00:03:19,769 --> 00:03:23,929
Early Riders about two years ago now, had read their white paper
55
00:03:24,379 --> 00:03:27,799
around Bitcoin being the hurdle rate and how it aligns
56
00:03:27,859 --> 00:03:33,019
incentives from LPs, GPs, and founders all trying to make more
57
00:03:33,109 --> 00:03:35,449
Bitcoin and using that as a hurdle rate. And it just made
58
00:03:35,539 --> 00:03:38,419
too much sense to me, and thought that all other financial
59
00:03:38,479 --> 00:03:40,599
firms, whether it be private equity, et cetera, would be
60
00:03:41,409 --> 00:03:44,559
either adopt this model eventually or be competed away,
61
00:03:44,949 --> 00:03:47,739
and so yeah, that's a little bit of my background and how we got
62
00:03:47,799 --> 00:03:48,969
here. I
63
00:03:49,120 --> 00:03:51,430
love it. You you really went from extreme high
64
00:03:51,880 --> 00:03:55,000
time preference to low time preference in terms of your
65
00:03:55,420 --> 00:03:58,180
investment theories and and outlooks. And and we've had
66
00:03:58,690 --> 00:04:02,940
Michael Tanguma on our podcast before, and early writers, and
67
00:04:03,060 --> 00:04:05,700
we love the story. I mean, we we couldn't agree more with you
68
00:04:05,760 --> 00:04:07,380
that we think that that's ultimately the way, the
69
00:04:07,650 --> 00:04:10,950
direction everything needs to flow. And I want to talk a
70
00:04:10,950 --> 00:04:14,400
little bit more about your role and and how you came to write
71
00:04:14,460 --> 00:04:17,130
this report, the Singularity Stack. So you you're really
72
00:04:17,370 --> 00:04:20,460
focused as a partner in in the firm on due diligence and
73
00:04:20,490 --> 00:04:24,260
portfolio operations. How did how did you come to create this
74
00:04:24,380 --> 00:04:27,140
research report? I mean, and I can't wait to dive into it
75
00:04:27,140 --> 00:04:30,470
because I just find it fascinating. But did you create
76
00:04:30,530 --> 00:04:32,750
it as a kind of a framework for how you're looking at the
77
00:04:32,810 --> 00:04:36,080
companies that come across your desk, or or how did how did that
78
00:04:36,710 --> 00:04:36,980
arise?
79
00:04:37,160 --> 00:04:41,380
Yeah, honestly, we try to pay attention to a lot of
80
00:04:41,500 --> 00:04:45,940
interesting signals out there, and I right now AI is obviously
81
00:04:45,940 --> 00:04:49,060
the most interesting thing in the room. There was a chart
82
00:04:49,090 --> 00:04:53,050
going around yesterday too that Open Router put out around token
83
00:04:53,110 --> 00:04:57,970
usage, and it was over the past two years, it's up over 25,000%
84
00:04:58,810 --> 00:05:01,680
and that's obviously just. Astounding! It's really showing
85
00:05:01,740 --> 00:05:06,660
just the amount of adoption that this that these tools are
86
00:05:06,720 --> 00:05:09,810
getting, and it's really fascinating too because, like,
87
00:05:09,870 --> 00:05:13,560
as an investor and and somebody who is trying to understand
88
00:05:13,620 --> 00:05:16,470
where the world is moving, I saw a lot of similarities between
89
00:05:16,560 --> 00:05:20,130
like what Jeff Bezos was doing when he was at D Shaw and and
90
00:05:20,190 --> 00:05:25,910
saw the internet grow. I think it was 2300 time percent year
91
00:05:25,970 --> 00:05:28,580
over year, which doesn't necessarily happen. And that's
92
00:05:28,610 --> 00:05:31,940
something that's free to use for any individual. Whereas you
93
00:05:31,970 --> 00:05:35,540
know, obviously, tokens on open router have a cost associated
94
00:05:35,600 --> 00:05:41,380
with them. And so obviously, out of that, he built Amazon and you
95
00:05:41,410 --> 00:05:44,950
know, selling books online at first doesn't sound very sexy,
96
00:05:45,490 --> 00:05:48,820
but we understood kind of by being on the frontier that it
97
00:05:48,820 --> 00:05:51,640
wasn't just going to be books; it was going to be more than
98
00:05:51,790 --> 00:05:53,680
that. They were going to sell everything, and obviously got
99
00:05:53,710 --> 00:05:57,730
into data centers out of that, and have try like built it into
100
00:05:57,760 --> 00:06:00,930
a massive company. And so I think that you know, just
101
00:06:00,960 --> 00:06:03,540
myself, and and I think a lot of people out there really are
102
00:06:03,600 --> 00:06:06,090
understand are trying to understand what are the second,
103
00:06:06,150 --> 00:06:09,450
third, fourth order implications of AI as the cost of
104
00:06:09,480 --> 00:06:15,540
intelligence goes near to zero, and what sort of businesses can
105
00:06:15,690 --> 00:06:18,330
be built out of that. What does that mean for my role today and
106
00:06:18,420 --> 00:06:20,480
in the future? And so that's kind of a little bit of the
107
00:06:20,600 --> 00:06:24,380
impetus of why I tried to do it because I think that it does
108
00:06:24,440 --> 00:06:27,500
have just profound implications for pretty much every part of
109
00:06:27,500 --> 00:06:29,870
the economy. Yeah, and I imagine it it clicked with the
110
00:06:30,080 --> 00:06:32,810
early writers team, the other partners right away, and
111
00:06:33,140 --> 00:06:36,020
kind of gave you the green light to to work on this. Well, I
112
00:06:36,080 --> 00:06:40,220
mean, before we we get into the framework, Liam, I want to just
113
00:06:40,280 --> 00:06:44,770
make sure I don't miss on asking you. Sometimes we, you know, we
114
00:06:44,860 --> 00:06:47,020
kind of go through our background and and keep it
115
00:06:47,020 --> 00:06:50,530
concise, and the transitions sound pretty smooth. But for
116
00:06:50,650 --> 00:06:53,230
others who are maybe going through, you know, similar
117
00:06:53,350 --> 00:06:56,680
career transitions or desiring to do so, I want to ask this
118
00:06:56,740 --> 00:07:01,500
question. I mean, from point 72 to working at a at a Bitcoin
119
00:07:02,550 --> 00:07:07,380
Bitcoin firm, it's not that easy of a transition necessarily. So,
120
00:07:07,830 --> 00:07:10,020
what helped you along the way?
121
00:07:10,230 --> 00:07:12,360
Yeah, that's a great question, and really appreciate
122
00:07:12,570 --> 00:07:17,700
it. I guess I was coming at it from the angle that there was a
123
00:07:17,790 --> 00:07:20,610
massive information asymmetry that's publicly available,
124
00:07:21,120 --> 00:07:24,650
whereas you know you guys put out a ton of great research and
125
00:07:25,010 --> 00:07:27,950
really make the information available about why Bitcoin's so
126
00:07:28,010 --> 00:07:31,010
important, etc. But there are so many folks out there that are
127
00:07:31,310 --> 00:07:36,200
you know even J.P. Morgan or Jamie Dimon from J.P. Morgan
128
00:07:36,590 --> 00:07:39,530
says like Satoshi can just come back and inflate the Bitcoin
129
00:07:39,680 --> 00:07:44,260
supply, and so with that, like I understood just by speaking with
130
00:07:44,290 --> 00:07:46,990
a bunch of folks that there was a lot of information asymmetry,
131
00:07:47,050 --> 00:07:50,620
and you know, eventually they would come around to why Bitcoin
132
00:07:50,680 --> 00:07:54,400
is valuable, and you know, would would need to allocate to it
133
00:07:54,430 --> 00:07:58,480
later on, just because economic forces would eventually get them
134
00:07:58,540 --> 00:08:01,530
there, because you know they would see their savings diminish
135
00:08:01,650 --> 00:08:05,340
over time in real value if they didn't allocate to Bitcoin, and
136
00:08:05,400 --> 00:08:09,450
so with that, I I understood you know it's it makes sense to
137
00:08:09,540 --> 00:08:13,380
allocate to the asset itself, and then I tried to understand
138
00:08:13,440 --> 00:08:17,370
well there could be other implications here too if you
139
00:08:17,430 --> 00:08:20,900
know the monetary system is inflating 7% a year, like
140
00:08:20,930 --> 00:08:24,770
there's going to be better information that can teach
141
00:08:24,860 --> 00:08:29,150
people about that, whether it's the internet, podcasts, AI, and
142
00:08:29,270 --> 00:08:31,700
there are going to be other implications for that too. And
143
00:08:31,760 --> 00:08:35,780
so, one of them that I thought about was there are going to be
144
00:08:36,350 --> 00:08:40,040
companies that will, you know, benefit from a lot more folks
145
00:08:40,100 --> 00:08:43,840
who trade Bitcoin? Who want to custody Bitcoin, lend against
146
00:08:44,560 --> 00:08:49,780
it, etc. There will be like an entire changing of financial
147
00:08:49,870 --> 00:08:52,600
services around that. And I actually thought about it too,
148
00:08:52,780 --> 00:08:55,570
and and I was like, well, if there are X amount of people
149
00:08:55,690 --> 00:08:59,290
today that own Bitcoin and Bitcoin's going to go up Y
150
00:08:59,440 --> 00:09:04,560
amount, like the amount of new, like new people trading Bitcoin
151
00:09:04,650 --> 00:09:07,770
and and financial services around it will actually likely
152
00:09:07,920 --> 00:09:12,180
increase faster than just Bitcoin itself, and so that that
153
00:09:12,180 --> 00:09:14,820
was kind of a moment that clicked for me that tried to
154
00:09:14,820 --> 00:09:18,780
understand like, well, if you back the right founders who are
155
00:09:18,870 --> 00:09:24,140
building interesting categories in here, then there's a massive
156
00:09:24,320 --> 00:09:28,100
opportunity for growth, and it's a really big investable category
157
00:09:28,250 --> 00:09:31,190
to back the infrastructure. To date, we've seen a lot of
158
00:09:31,640 --> 00:09:36,470
distractions with digital asset companies. I think you know
159
00:09:36,590 --> 00:09:40,430
focus on on like meme coins and whatever from like Coinbase's of
160
00:09:40,430 --> 00:09:44,110
the world, but I think they've obviously financially still done
161
00:09:44,230 --> 00:09:47,860
extremely well. I think that if you really develop
162
00:09:48,310 --> 00:09:51,820
differentiated products and see that growth, like it's going to
163
00:09:51,850 --> 00:09:55,660
be an extremely attractive opportunity. And so I had that
164
00:09:55,750 --> 00:10:01,050
in mind, and eventually, just you know. Spoke with enough
165
00:10:01,260 --> 00:10:04,410
companies out there in the space to understood that that was the
166
00:10:04,530 --> 00:10:08,160
right time. The consumers were kind of waking up, and the
167
00:10:08,310 --> 00:10:13,350
infrastructure wasn't just it was maturing at the right time.
168
00:10:13,530 --> 00:10:16,140
That there were more folks who were ready to come into the
169
00:10:16,170 --> 00:10:19,680
space, and so, but obviously not not yet developed enough for
170
00:10:19,770 --> 00:10:22,910
everybody, and so that's why I kind of made the the leap to go
171
00:10:22,970 --> 00:10:25,700
to early riders to help invest in other infrastructure
172
00:10:25,790 --> 00:10:26,630
companies in the space.
173
00:10:27,470 --> 00:10:29,690
Well, I think this is a good time to kind of
174
00:10:29,750 --> 00:10:33,290
get into the analogy of building the infrastructure for Bitcoin
175
00:10:33,410 --> 00:10:36,860
and what Bitcoin financial rails will look like, and how you
176
00:10:36,950 --> 00:10:40,010
start your report, the Singularity stack, with kind of
177
00:10:40,100 --> 00:10:44,110
the the a look back at at electricity and electrical
178
00:10:44,140 --> 00:10:46,870
lighting. So can you can you start us out by kind of because
179
00:10:46,900 --> 00:10:50,350
I do think that there is a lack of understanding about how
180
00:10:50,470 --> 00:10:55,210
important Bitcoin is for future financial services. So maybe you
181
00:10:55,270 --> 00:10:58,810
can can bring us that analogy that might help people that
182
00:10:58,900 --> 00:11:02,550
aren't understanding that Bitcoin is so much more than an
183
00:11:02,730 --> 00:11:05,400
asset that trades, but it's really a foundational protocol.
184
00:11:05,580 --> 00:11:08,970
Yeah, for sure. I guess taking step back. Yeah, the the
185
00:11:09,810 --> 00:11:14,370
light. We we started off the paper about the AI singularity
186
00:11:14,520 --> 00:11:17,670
with the analogy to light because I think that over time
187
00:11:17,820 --> 00:11:21,300
intelligence is something that's going to be essentially
188
00:11:21,360 --> 00:11:24,740
available for everybody, and going to be so cheap that it's
189
00:11:24,770 --> 00:11:28,370
going to be very difficult to even meter it. And so, back in
190
00:11:28,370 --> 00:11:31,970
the day, like 200 years ago, it would cost three hours of
191
00:11:32,810 --> 00:11:37,010
somebody's time in order to get an hour of light, and that was
192
00:11:37,100 --> 00:11:41,770
just because there it was craft good. Like you had to actually
193
00:11:41,860 --> 00:11:47,680
create, like get whale oil, etc. and create candles yourself.
194
00:11:47,890 --> 00:11:52,240
Rather, and there wasn't really any use to being up late at
195
00:11:52,300 --> 00:11:55,870
night because everybody was farmers, and then they're like,
196
00:11:56,140 --> 00:11:59,410
"What? What can you really even do after light? It's just read a
197
00:11:59,440 --> 00:12:03,720
book or whatever. But today it costs like less than a 10th of a
198
00:12:03,810 --> 00:12:07,200
second in order to have an hour of light because we've built up
199
00:12:07,350 --> 00:12:11,430
all the infrastructure around it, and so I think that
200
00:12:11,550 --> 00:12:13,830
obviously had massive implications that I think we're
201
00:12:13,980 --> 00:12:20,130
about to see now as it relates to the first and like first and
202
00:12:20,190 --> 00:12:23,210
second industrial revolution and how there was then a build out
203
00:12:23,390 --> 00:12:28,520
of 24/7 manufacturing and supply chain that will go around the
204
00:12:28,550 --> 00:12:31,640
world, and right now we're starting to see the same amount
205
00:12:31,640 --> 00:12:36,110
of it with AI and intelligence coming from a craft good that
206
00:12:36,710 --> 00:12:40,450
you know you have to sit down and think about for a certain
207
00:12:40,480 --> 00:12:44,440
amount of time versus something that's just on hand callable and
208
00:12:44,830 --> 00:12:48,940
can be, you know, metered by the second. And so we're starting to
209
00:12:49,030 --> 00:12:53,800
change the way that we actually pay for goods, and obviously
210
00:12:53,830 --> 00:12:57,250
intelligence is one of the biggest ones there, and that's
211
00:12:57,280 --> 00:13:02,490
going to have implications for not just the the way that the
212
00:13:02,550 --> 00:13:07,320
economy works, but also how it ends up in which payment method
213
00:13:07,680 --> 00:13:09,660
is ultimately used in the long term.
214
00:13:09,750 --> 00:13:13,620
So let's kind of connect the dots here, Liam. The report
215
00:13:13,650 --> 00:13:17,520
lays out nine layers, starting with with raw electricity. Of
216
00:13:17,640 --> 00:13:21,200
course, this all eventually ties to value and money. Can you take
217
00:13:21,200 --> 00:13:23,900
us through what the stack is? What what the foundation of the
218
00:13:23,900 --> 00:13:24,050
stack
219
00:13:24,410 --> 00:13:29,990
is? Yeah, real quick. It's essentially everything is
220
00:13:30,200 --> 00:13:34,820
downstream of energy, which is the primary source source of
221
00:13:34,910 --> 00:13:37,730
energy for everything here. And then after that, it goes to
222
00:13:37,820 --> 00:13:43,540
compute. So think chips, then chips and data centers. Then
223
00:13:43,900 --> 00:13:47,380
after that, it's the models themselves, harnesses, which is
224
00:13:47,740 --> 00:13:51,850
you know Claude Code or CoWork or Hermes, Open Claw, etc.
225
00:13:52,810 --> 00:13:56,740
Agents that can do the work for you themselves, and then it goes
226
00:13:56,830 --> 00:14:02,040
into identity, building a web of trust, or exactly what
227
00:14:02,130 --> 00:14:05,190
permissions you give any specific agent yourself,
228
00:14:05,820 --> 00:14:10,350
coordinating across them, value transfer and settlement, and so
229
00:14:10,440 --> 00:14:14,400
we can go into each one of those a bit further. But throughout
230
00:14:14,550 --> 00:14:18,360
time, the the entire kind of the supply chain has seen
231
00:14:18,780 --> 00:14:23,630
transaction costs be about 7% of GDP, because there have been
232
00:14:24,200 --> 00:14:28,430
massive dislocations with the amount of available data at each
233
00:14:28,580 --> 00:14:31,970
step of the supply chain, the amount of knowledge that's
234
00:14:31,970 --> 00:14:36,770
available, and I think that right now, all the entire work
235
00:14:37,160 --> 00:14:41,560
chain at this point is really collapsing significantly into
236
00:14:43,060 --> 00:14:47,290
all of the data being extremely available, callable instantly,
237
00:14:47,770 --> 00:14:50,410
and thus a lot less leakage.
238
00:14:50,860 --> 00:14:54,070
We're recording this on the 17th of of
239
00:14:55,120 --> 00:14:58,000
September, and in the last week, there's been a lot of
240
00:14:58,030 --> 00:15:02,820
conversation just around. Is now the time to rein in AI? AI is is
241
00:15:03,240 --> 00:15:06,090
just advancing so quickly, and some of the leaders of the
242
00:15:06,180 --> 00:15:09,750
Frontier Labs are calling for more restraint in the growth.
243
00:15:09,870 --> 00:15:13,470
There have been some incidences that have caused concern about
244
00:15:13,530 --> 00:15:16,560
how fast it's growing and how independent these agents could
245
00:15:16,620 --> 00:15:21,110
become in working against or or just working to achieve a goal
246
00:15:21,500 --> 00:15:26,510
with some disregard of what what constraints may be, how how how
247
00:15:26,600 --> 00:15:28,820
does that factor in? Or I'm sure that you've thought about this
248
00:15:28,940 --> 00:15:32,300
in terms of your your structure and the stack that you're laying
249
00:15:32,480 --> 00:15:35,570
out. What would happen, or what do you see it as even a
250
00:15:35,630 --> 00:15:39,980
possibility that this can that this stack will have a hiccup or
251
00:15:40,040 --> 00:15:44,140
will have a you know a scale back in terms of its evolution
252
00:15:44,170 --> 00:15:44,740
or its growth?
253
00:15:44,920 --> 00:15:48,010
Yeah, I mean there are a lot of people who are who've
254
00:15:48,100 --> 00:15:52,060
been screeching about safety recently, especially Dario with
255
00:15:52,120 --> 00:15:56,890
his "We Must Pace the Frontier" letter recently. I think that
256
00:15:56,920 --> 00:16:00,310
it's mostly marketing for their IPO and saying that they have
257
00:16:00,520 --> 00:16:03,270
something that's so powerful that you know it can't be
258
00:16:03,510 --> 00:16:06,090
released in the wild, and and they're looking for regulatory
259
00:16:06,180 --> 00:16:09,390
capture for having their friends at Meter, who are funded by the
260
00:16:09,420 --> 00:16:13,770
same people, be integrated into their businesses, and thus kind
261
00:16:13,770 --> 00:16:17,730
of add regulatory costs to everything. I think in the the
262
00:16:17,820 --> 00:16:22,640
long term, we're we're all going to have our own AI that's going
263
00:16:22,640 --> 00:16:25,970
to be post trained on our own data and and specific to what we
264
00:16:25,970 --> 00:16:29,690
want to have done. And so with that, there will be an
265
00:16:29,750 --> 00:16:34,310
incredible amount of foundation models out there, and it it
266
00:16:34,400 --> 00:16:38,120
can't necessarily just all go to one person. So with respect to
267
00:16:38,150 --> 00:16:43,810
the actual risks, I think that the risks lie in having a very
268
00:16:43,960 --> 00:16:48,850
significant amount of compute, and thus being able to take
269
00:16:50,350 --> 00:16:53,020
either people who want to achieve things that are bad
270
00:16:53,080 --> 00:16:55,900
because there are bad people in the world, like that will either
271
00:16:56,020 --> 00:17:01,320
try to steal your Bitcoin or you know want to have cyber attacks
272
00:17:01,380 --> 00:17:06,390
on U.S. infrastructure, whether it's countries or just terrorist
273
00:17:06,420 --> 00:17:10,800
groups, and so I think that that is a risk, and I think that you
274
00:17:10,830 --> 00:17:14,250
know that that should be taken somewhat seriously. But I think
275
00:17:14,340 --> 00:17:20,100
that the only way to really have that risk be sufficient is
276
00:17:20,730 --> 00:17:27,260
giving AI a massive amount of compute and allow that to, you
277
00:17:27,290 --> 00:17:31,400
know, essentially go unbounded to attack systems. And so the
278
00:17:31,640 --> 00:17:36,500
only way that that obviously is hap is capable with open source
279
00:17:36,620 --> 00:17:40,990
models. But honestly, I think that there is a relatively low
280
00:17:41,320 --> 00:17:45,070
risk of people running open source models on their their
281
00:17:45,130 --> 00:17:48,220
laptop that are able to like really attack U.S. critical
282
00:17:48,250 --> 00:17:51,370
infrastructure, and so I think that the what it all comes down
283
00:17:51,430 --> 00:17:56,200
to is that these frontier labs are trying to just limit the
284
00:17:56,320 --> 00:18:00,070
amount of capabilities that people can have, so that they
285
00:18:00,100 --> 00:18:02,910
don't need to train their models even more because they have
286
00:18:03,000 --> 00:18:06,240
these upcoming IPOs. I don't. I don't know what you guys think
287
00:18:06,270 --> 00:18:10,290
about that as well. But it feels like a lot of hysteria that
288
00:18:10,410 --> 00:18:13,440
isn't necessarily grounded in anything because we, while we've
289
00:18:13,590 --> 00:18:17,190
seen that one OpenAI hugging face incident, there hasn't been
290
00:18:17,460 --> 00:18:20,010
anything else that's really significant at this point.
291
00:18:20,100 --> 00:18:21,890
I'm actually just shocked at how many times
292
00:18:22,010 --> 00:18:25,730
I'm hearing the the name of the company Hugging Face on on in
293
00:18:25,850 --> 00:18:29,240
traditional media these days, and I've been familiar with and
294
00:18:29,270 --> 00:18:32,030
using the company for about eight or nine years now. I I
295
00:18:32,210 --> 00:18:35,180
used to invest heavily in AI, and so I'm familiar with it. But
296
00:18:35,270 --> 00:18:39,530
yeah, it does feel that there's a you know coordinated effort,
297
00:18:39,590 --> 00:18:44,470
or somebody is just like you know, just piling onto this
298
00:18:44,590 --> 00:18:47,860
news, and it feels unusual. It feels like an unusual time and
299
00:18:47,920 --> 00:18:51,940
space. And I agree with you that I think it would be related to
300
00:18:51,970 --> 00:18:56,200
the IPO or or just the fear that these frontier labs who have
301
00:18:56,260 --> 00:18:58,750
raised so much investment capital and have very
302
00:18:59,680 --> 00:19:02,460
significant valuations that they need to live up to are now
303
00:19:02,910 --> 00:19:07,110
facing pressure from open source labs around the world and
304
00:19:07,230 --> 00:19:10,740
different other models and different other financial models
305
00:19:10,890 --> 00:19:13,050
that could disrupt their business and and make their
306
00:19:13,050 --> 00:19:17,160
valuations tumble down, which is a risk, frankly, to the U.S.
307
00:19:18,240 --> 00:19:21,620
stock market because so much of the market is based on the
308
00:19:21,680 --> 00:19:25,460
valuations of these AI and AI related companies. So, yeah,
309
00:19:25,970 --> 00:19:28,460
interesting. I I would tend to agree. Israel, what about you?
310
00:19:28,580 --> 00:19:28,640
No,
311
00:19:28,790 --> 00:19:31,670
I mean I just don't see how that you can stop something
312
00:19:31,700 --> 00:19:37,580
like this. I mean, even even if some of these claims were true,
313
00:19:37,700 --> 00:19:41,140
which to be clear, and to your point, Liam, I also haven't seen
314
00:19:41,230 --> 00:19:44,560
any hard evidence or data behind this. It's all just kind of been
315
00:19:44,650 --> 00:19:48,520
these stories or yeah claims. Ultimately, so it's hard to
316
00:19:48,550 --> 00:19:52,840
know. But even regardless of that, I mean, I think both
317
00:19:52,990 --> 00:19:55,960
Bitcoin and AI are technologies that are now out there, and you
318
00:19:56,110 --> 00:19:59,470
you can't really stop them. You know, in in a way because of the
319
00:19:59,560 --> 00:20:04,290
game theory. Bitcoin, of course, is a lot more decentralized and
320
00:20:04,380 --> 00:20:08,640
open than than AI. AI is still concentrated in a few global
321
00:20:08,700 --> 00:20:13,200
players, but I mean, ultimately, my mindset around that is you
322
00:20:13,260 --> 00:20:17,910
would need very coordinated global efforts from you know
323
00:20:18,180 --> 00:20:22,280
adversaries, even you know geopolitically, to actually slow
324
00:20:22,370 --> 00:20:27,200
this down and or you know try to pause it, so I just don't see
325
00:20:27,290 --> 00:20:29,330
it. I mean, I think the game theory just doesn't allow it.
326
00:20:29,420 --> 00:20:29,990
Yeah, and I think
327
00:20:30,110 --> 00:20:33,920
that it's almost similar to the crypto wars of the 80s,
328
00:20:34,580 --> 00:20:38,750
back when they were trying to make encryption illegal, weapons
329
00:20:38,840 --> 00:20:43,210
grade, et cetera, and into a lot of the Atomback and other
330
00:20:43,420 --> 00:20:49,150
cypherpunks of that area era really tried to combat that with
331
00:20:49,180 --> 00:20:53,110
their work with PGP. And I think that yeah, I mean there are
332
00:20:53,350 --> 00:20:57,220
people that use encryption for bad things. Like that's I think
333
00:20:57,280 --> 00:21:00,490
that's well known. But I think that just like AI, like
334
00:21:01,330 --> 00:21:04,860
encryption is a tool for privacy, and and it does create
335
00:21:04,950 --> 00:21:08,850
a lot of great things. And similarly, I think that you know
336
00:21:08,970 --> 00:21:12,630
there there will be probably some small negative implications
337
00:21:12,720 --> 00:21:16,110
from AI. But I think that the amount of abundance and goodness
338
00:21:16,230 --> 00:21:19,860
that it will create for everybody out there, from you
339
00:21:19,890 --> 00:21:23,720
know allow causing deflation for goods and services to curing
340
00:21:23,810 --> 00:21:27,320
diseases, allowing people to live healthier lives, saving
341
00:21:27,380 --> 00:21:30,860
them time to work more on what they actually want to work on.
342
00:21:30,950 --> 00:21:34,400
It's going to be a massive net positive, and I don't think that
343
00:21:34,460 --> 00:21:38,540
there is any reasonable way to regulate it at the at this
344
00:21:38,600 --> 00:21:39,230
stage. Lynne
345
00:21:39,320 --> 00:21:42,220
and I are partners at Baselayer Advisors, where
346
00:21:42,310 --> 00:21:44,500
through our experience and network with venture capital
347
00:21:44,560 --> 00:21:47,440
partners, we connect interested investors with unique
348
00:21:47,500 --> 00:21:50,560
opportunities in the space and advise startups on their growth
349
00:21:50,800 --> 00:21:54,250
and fundraising. Visit our website's advisory section to
350
00:21:54,310 --> 00:21:58,090
learn more. Yeah, I largely agree. Well, let's let's maybe
351
00:21:58,300 --> 00:22:02,700
unpack the the different layers to the to to the stack in your
352
00:22:02,760 --> 00:22:05,820
report, Liam, we can just kind of tackle them, you know, two or
353
00:22:05,880 --> 00:22:09,810
three at a time. But you start with electricity and models as
354
00:22:10,200 --> 00:22:13,950
you know as kind of the base, and you hit on some points which
355
00:22:14,400 --> 00:22:16,770
I'd like you to you know expand a little bit on. In the
356
00:22:16,950 --> 00:22:20,390
electricity stack, you get you highlight how we have
357
00:22:20,600 --> 00:22:25,010
bottlenecks. We've underinvested in the infrastructure. You know,
358
00:22:25,100 --> 00:22:28,700
there's a lot of efforts and capital right now in building
359
00:22:29,060 --> 00:22:32,240
out the the sudden burst of demand that we have for for
360
00:22:32,270 --> 00:22:36,590
electricity. Of course, all these data centers that are are
361
00:22:37,130 --> 00:22:40,390
all over the news. How do you view these bottlenecks, and how
362
00:22:40,570 --> 00:22:44,230
does that play into the next layer of the stack, which are
363
00:22:44,290 --> 00:22:47,290
the models themselves, which you argue are only going to get
364
00:22:47,380 --> 00:22:47,620
cheaper?
365
00:22:47,800 --> 00:22:50,470
Yeah, thinking about energy, I think is the biggest
366
00:22:50,530 --> 00:22:55,060
bottleneck because that's heavily regulated industry for
367
00:22:55,120 --> 00:22:59,260
the most part, and so with that, I think that the the grid has
368
00:22:59,590 --> 00:23:02,670
certain restrictions that are going to keep massive amounts
369
00:23:02,730 --> 00:23:07,350
of, you know, power plants, coal mines, etc. that are going to be
370
00:23:07,530 --> 00:23:11,850
held in the queue for a decently long period of time. The grid's
371
00:23:11,880 --> 00:23:14,940
been about flat for the past two decades in terms of the total
372
00:23:15,090 --> 00:23:18,300
energy that it generates, and so I think right now that's the
373
00:23:18,480 --> 00:23:22,070
constraint. Obviously, in the U.S. that's a big focus, and and
374
00:23:22,580 --> 00:23:27,200
more so moving faster in areas like China. But in the end of
375
00:23:27,200 --> 00:23:30,170
the day, there's only so much that they can do, just given its
376
00:23:30,230 --> 00:23:33,080
regulated industry, and that's even why we've seen everything
377
00:23:33,260 --> 00:23:38,870
that's become almost not not quite consensus, but like a lot
378
00:23:38,900 --> 00:23:42,130
of people think that they're going to be orbital data centers
379
00:23:42,250 --> 00:23:46,780
relatively soon because that can get be a way to get out of the
380
00:23:46,960 --> 00:23:50,140
permitting issues that come with respect to everything that's
381
00:23:50,200 --> 00:23:54,280
going on in the U.S. and energy creation, and so I think that
382
00:23:54,460 --> 00:23:57,310
energy is going to remain the bottleneck. But while we've seen
383
00:23:57,460 --> 00:24:01,380
actual supply chain issues with respect to compute itself as it
384
00:24:01,740 --> 00:24:07,530
relates to chips, just because TSMC and folks like them are
385
00:24:07,680 --> 00:24:11,490
pretty much the only folks with enough institutionalized
386
00:24:11,550 --> 00:24:15,810
knowledge to create fabs themselves. The chips still are
387
00:24:15,900 --> 00:24:20,900
getting 50 to 100x better per year, and that's obviously
388
00:24:20,990 --> 00:24:26,060
creating both the ability to issue tokens faster. So
389
00:24:26,780 --> 00:24:30,860
obviously, faster tokens means you can go back and forth with
390
00:24:30,860 --> 00:24:35,570
the AI a lot more, or or just it can work for longer tasks
391
00:24:35,870 --> 00:24:40,070
autonomously, while as just the ability for it to hold more
392
00:24:40,130 --> 00:24:44,710
memory and transfer that at a faster rate, and so energy is
393
00:24:44,890 --> 00:24:48,610
the the biggest bottleneck there. Nvidia and Sarah Bross
394
00:24:48,760 --> 00:24:51,370
and and some other companies are are doing great work on the
395
00:24:51,370 --> 00:24:56,020
compute side, but I think on the energy side, as as it relates to
396
00:24:56,080 --> 00:24:58,570
kind of opportunities that we've been talking about, as it
397
00:24:58,570 --> 00:25:03,600
relates to seeing. Soaking costs go up 25,000% over the past two
398
00:25:03,750 --> 00:25:07,710
years, and trying to invest across the landscape, there
399
00:25:07,950 --> 00:25:10,530
there are really interesting opportunities on the energy
400
00:25:10,650 --> 00:25:14,430
side. I think specifically with respect to storage and
401
00:25:14,910 --> 00:25:18,480
transmission, I think you know base power is an interesting
402
00:25:18,540 --> 00:25:22,130
company doing interesting things at the moment. Compute is
403
00:25:22,250 --> 00:25:25,160
obviously one that's done extremely well, and there will
404
00:25:25,340 --> 00:25:28,910
likely be new disruptions there just because of everything as it
405
00:25:28,910 --> 00:25:32,720
relates to the engineering of of these chips and ultimately more
406
00:25:32,780 --> 00:25:36,050
capacity coming online. I think that both of those are are
407
00:25:36,140 --> 00:25:39,470
investable opportunities, and and we can get into models next.
408
00:25:39,530 --> 00:25:44,470
But I don't necessarily think those are as differentiated, and
409
00:25:45,010 --> 00:25:49,870
you know we'll have a long enough time horizon to return
410
00:25:50,020 --> 00:25:53,290
substantial amounts of cash flow to their actual investors
411
00:25:53,350 --> 00:25:55,780
relative to the nodes of energy and compute.
412
00:25:56,650 --> 00:25:59,140
You know, kind of together with this backlash
413
00:25:59,230 --> 00:26:02,160
or fear about AI and the general public. There's the same thing
414
00:26:02,250 --> 00:26:04,680
about the construction of data centers and forms of
415
00:26:04,710 --> 00:26:08,340
electricity. What would you advise? I mean, it feels like
416
00:26:08,370 --> 00:26:12,090
there's been a slowdown even in traditional energy developed
417
00:26:13,260 --> 00:26:16,560
states like Texas. Abbott's taking kind of a step back on a
418
00:26:16,590 --> 00:26:20,720
data center development. How would you frame it to to help
419
00:26:21,380 --> 00:26:24,170
shift that narrative about how important electricity is, just
420
00:26:24,260 --> 00:26:28,190
overall, and the need for for these foundational systems.
421
00:26:28,430 --> 00:26:31,760
Yeah, I think a lot of people, at least in our circles,
422
00:26:31,790 --> 00:26:35,810
I would say, have seen the chart about how there are no energy
423
00:26:37,250 --> 00:26:41,950
poor and developed countries over time. Like it's directly
424
00:26:42,040 --> 00:26:44,860
correlated with the amount of energy that a country has to the
425
00:26:44,950 --> 00:26:48,730
amount of prosperity and GDP that it has, and so I think that
426
00:26:48,760 --> 00:26:52,600
it's it's obviously crucial. And if you want the your
427
00:26:53,530 --> 00:26:57,730
constituents to be well educated and for them to have better
428
00:26:58,270 --> 00:27:00,960
business opportunities and and better job opportunities
429
00:27:01,020 --> 00:27:03,720
domestically than elsewhere. Then I think that it that is
430
00:27:04,110 --> 00:27:09,360
something that every country municipality should be thinking
431
00:27:09,390 --> 00:27:13,350
about very critically. It's a tough one though because they're
432
00:27:13,560 --> 00:27:16,230
they're incredibly unpopular. I think this is going to be
433
00:27:16,350 --> 00:27:21,710
probably the biggest biggest sticking point in terms of AI at
434
00:27:22,340 --> 00:27:25,490
the next election coming up, because I think something like
435
00:27:25,640 --> 00:27:31,160
80% of people view data centers and AI as unfavorable. I think
436
00:27:31,250 --> 00:27:35,150
that a lot of that has to do, you know, there there are some
437
00:27:35,330 --> 00:27:39,740
noise pollution, but I think that most of them are built far
438
00:27:39,950 --> 00:27:44,290
away from actually where people live, but the real issue is they
439
00:27:44,740 --> 00:27:50,410
see data centers and AI as just one of the the physical
440
00:27:50,650 --> 00:27:56,590
manifestation of the emerging gap between the wages and the
441
00:27:56,650 --> 00:28:00,570
amount of capital that people on the lower shape of the K have
442
00:28:00,720 --> 00:28:06,510
versus the upper shape of the K. They they see you know Elon, Sam
443
00:28:06,630 --> 00:28:12,120
Altman, etc. All these very rich people who you know have a ton
444
00:28:12,150 --> 00:28:16,050
of money, and they just hear on the news that it's because you
445
00:28:16,080 --> 00:28:18,300
know they're stealing from the country, and and they shouldn't
446
00:28:18,330 --> 00:28:20,900
have that, and it should be redistributed. And then they
447
00:28:21,140 --> 00:28:24,290
also hear that AI is going to change their jobs, which I think
448
00:28:24,410 --> 00:28:28,580
is another false narrative. And because of those two are, I
449
00:28:28,700 --> 00:28:31,370
think, the the most popular narratives out there. It's going
450
00:28:31,370 --> 00:28:35,420
to be difficult to ultimately get the people to really change
451
00:28:35,810 --> 00:28:39,740
how they use it, unless or how they view AI, unless they
452
00:28:40,070 --> 00:28:42,940
actually see it in the benefits themselves, which they they
453
00:28:43,060 --> 00:28:46,420
largely haven't yet, outside of just having a quicker, faster
454
00:28:46,570 --> 00:28:51,490
Google. And ultimately, the the politicians probably want to say
455
00:28:51,670 --> 00:28:55,510
whatever the people actually want to get elected. And so, I
456
00:28:55,660 --> 00:28:57,730
think it's going to be a challenging topic over over
457
00:28:57,820 --> 00:29:00,870
time. And I don't necessarily have this silver bullet, whether
458
00:29:00,930 --> 00:29:05,190
it's narrative on how to shape this better, what what do you
459
00:29:05,280 --> 00:29:05,820
guys think?
460
00:29:06,180 --> 00:29:08,520
I'm not sure either on where this goes, Liam. All I
461
00:29:08,670 --> 00:29:13,020
know is that the negative narrative around it will
462
00:29:13,140 --> 00:29:18,120
continue. So I think that that's just not going away. You know,
463
00:29:18,210 --> 00:29:24,740
maybe it adapts. We saw Bitcoin Bitcoin mining infrastructure
464
00:29:24,800 --> 00:29:28,190
adapt, you know, in the last 10 years, and plugged into more,
465
00:29:28,880 --> 00:29:32,690
you know, rural energy sources, and it's gotten a little bit
466
00:29:32,750 --> 00:29:38,150
more renewable energy focused. So it's, you know, I think
467
00:29:38,240 --> 00:29:41,890
ultimately it's kind of one of those things where the market
468
00:29:42,010 --> 00:29:47,050
will work it out as as it as it should, you know. But I do think
469
00:29:47,350 --> 00:29:50,830
the regulatory headwinds are not going away anytime soon.
470
00:29:51,820 --> 00:29:53,590
Yeah, I agree with Israel. I think that
471
00:29:53,710 --> 00:29:56,890
Bitcoin mining has given a great example of how a narrative can
472
00:29:57,010 --> 00:30:00,390
shift relatively quickly from some. Something that was
473
00:30:00,480 --> 00:30:03,540
misunderstood and misinterpreted to something that now is you
474
00:30:03,600 --> 00:30:07,050
know we understand that it can actually improve the pace of
475
00:30:07,140 --> 00:30:10,950
renewable development in electricity. I think I'm hoping
476
00:30:11,070 --> 00:30:15,000
that after the midterm U.S. elections that the narrative
477
00:30:15,030 --> 00:30:17,310
will shift a little bit. That'll it'll take the pressure off at
478
00:30:17,340 --> 00:30:20,930
least for a year and a half, and then. But I'm also wondering if
479
00:30:20,960 --> 00:30:24,320
this just gives opportunities. I mean, our podcast tends to look
480
00:30:24,500 --> 00:30:27,470
at these developments on more of a global scale than just U.S.
481
00:30:27,530 --> 00:30:30,650
centric. So, you know, perhaps there are opportunities in other
482
00:30:30,740 --> 00:30:34,520
regions and other areas to help do some of the AI build out that
483
00:30:34,580 --> 00:30:37,910
would increase the opportunities in those emerging economies
484
00:30:38,180 --> 00:30:41,890
where they do need more electricity. They need more to
485
00:30:42,130 --> 00:30:45,370
to improve the prosperity and the GDP. That this could be a
486
00:30:45,460 --> 00:30:47,590
source of opportunity for other regions.
487
00:30:48,100 --> 00:30:50,590
Yeah, I think that's fantastic point. Very well said,
488
00:30:50,800 --> 00:30:55,750
and something that you know as there are differing opinions,
489
00:30:55,900 --> 00:30:59,740
whether it's the U.S. versus China, and how those are so
490
00:30:59,860 --> 00:31:03,570
diametrically opposed with the public perception of AI. I think
491
00:31:03,690 --> 00:31:07,170
that outside of the U.S. there are a ton of countries. I think
492
00:31:07,260 --> 00:31:10,350
South Korea just actually said that they're going to allow for
493
00:31:10,860 --> 00:31:14,610
something like $8 billion worth of free AI for their people too.
494
00:31:15,180 --> 00:31:17,790
I think that that's a an incredibly smart strategy in
495
00:31:17,850 --> 00:31:22,760
order to you know have your you know citizens be able to retrain
496
00:31:22,820 --> 00:31:26,030
themselves in order to bring your country further ahead. I
497
00:31:26,090 --> 00:31:28,880
think that they should be teaching how to use these types
498
00:31:28,970 --> 00:31:32,960
of tools in schools, and and that will help improve
499
00:31:33,050 --> 00:31:36,380
perception, and then also provide massive economic
500
00:31:36,530 --> 00:31:37,670
opportunity, as you mentioned.
501
00:31:38,630 --> 00:31:41,230
Yeah, and and I mean on that point worth highlighting as
502
00:31:41,410 --> 00:31:44,230
well, and I'm not too familiar with the numbers, but I did see
503
00:31:44,440 --> 00:31:48,610
recently El Salvador that they've been implementing AI at
504
00:31:48,670 --> 00:31:52,630
the educational level at a very early stage, I believe. And
505
00:31:53,710 --> 00:31:56,290
again, I'm not familiar with the stats, but I think they ranked
506
00:31:57,010 --> 00:32:00,690
pretty high in recent testing of their young kids. So it's having
507
00:32:01,290 --> 00:32:04,680
clearly big benefits to those who are leaning into this
508
00:32:04,770 --> 00:32:08,250
technology. I mean, ultimately, the U.S. is the powerhouse
509
00:32:08,490 --> 00:32:12,030
economically that it is today because it, for the most part,
510
00:32:12,120 --> 00:32:15,180
embraced the internet early on, and so as a consequence, you
511
00:32:15,180 --> 00:32:18,360
know, now you have some of the largest companies and all you
512
00:32:18,360 --> 00:32:21,680
know multiple trillion dollar companies that are, you know,
513
00:32:21,770 --> 00:32:25,610
that were built from the Internet Protocol, and the U.S.
514
00:32:25,670 --> 00:32:28,460
is has of course reaped massive benefits from that. So I think,
515
00:32:29,270 --> 00:32:31,490
you know, those those who lean into this will certainly
516
00:32:31,550 --> 00:32:36,440
benefit. But well, maybe getting back to the to the stack itself,
517
00:32:36,560 --> 00:32:39,140
Liam. So we went through the electricity component, the
518
00:32:39,650 --> 00:32:43,270
models themselves, as I understand it, you know. Correct
519
00:32:43,300 --> 00:32:45,820
me if I'm wrong, but you you view these as a little more
520
00:32:45,940 --> 00:32:50,110
fluid. They'll kind of you know they'll come and go likely for
521
00:32:50,170 --> 00:32:53,020
the foreseeable future, both closed source and open source.
522
00:32:53,650 --> 00:32:57,970
Then you get into a layer of the stack called the harness, which
523
00:32:58,960 --> 00:33:01,740
is, in other words, perhaps you know the infrastructure layer.
524
00:33:01,950 --> 00:33:05,970
Let's say the tools and and what you actually need to use some of
525
00:33:06,030 --> 00:33:09,420
these models. And you've you view this as one of the most
526
00:33:09,900 --> 00:33:13,620
maybe sticky or investable layers. Again, correct me if I'm
527
00:33:13,680 --> 00:33:17,130
wrong on on any of this. And then we you know you kind of get
528
00:33:17,190 --> 00:33:20,630
into how the agents interact. You know on top of the models
529
00:33:20,750 --> 00:33:24,290
and and this harness layer, is that more or less the thesis, or
530
00:33:24,500 --> 00:33:25,850
what? What can you add to that?
531
00:33:25,970 --> 00:33:28,850
Yeah, for sure. There, with respect to the models
532
00:33:28,910 --> 00:33:32,150
themselves, they're obviously those have raised the most
533
00:33:32,180 --> 00:33:35,810
amount of capital to date, and I think that they're they get a
534
00:33:35,840 --> 00:33:40,280
lot of noise because they that's kind of what the back end is
535
00:33:40,340 --> 00:33:44,500
that powers all of this, and some of them have also like
536
00:33:44,590 --> 00:33:51,040
obviously cloud code and and codecs are harnesses for the
537
00:33:51,040 --> 00:33:55,780
anthropics and open AIs of the world who offer just a great
538
00:33:56,140 --> 00:34:00,130
user experience within their tools. But ultimately, I think
539
00:34:00,220 --> 00:34:04,230
that a lot of the the models are a little bit unsustainable
540
00:34:04,350 --> 00:34:07,800
because of the economics that go into it, and what I mean by that
541
00:34:07,800 --> 00:34:12,180
is, if you're paying for a $200 cloud subscription right now, it
542
00:34:12,180 --> 00:34:15,270
actually costs them. If you're using it into the fullest
543
00:34:15,360 --> 00:34:20,190
extent, it's costing them about 8000 to $9,000 a month, and so
544
00:34:20,400 --> 00:34:23,390
that's a massively negative margin. But why do they do that?
545
00:34:23,450 --> 00:34:28,700
It's because they are taking all of the data about how consumers
546
00:34:29,000 --> 00:34:32,990
and businesses use their information, and they will then
547
00:34:33,350 --> 00:34:37,070
have all of that data that's readily available to train their
548
00:34:37,130 --> 00:34:41,380
next model or use it in order to go out into a different
549
00:34:41,440 --> 00:34:45,610
horizontal verticals, and so like folks like Figma have come
550
00:34:45,790 --> 00:34:49,300
on and been partners with them, and use Claude as a backend, and
551
00:34:49,330 --> 00:34:54,280
then Claude goes and does Claude design, and so I think with that
552
00:34:54,340 --> 00:34:58,090
it's it's interesting to do because they have the API that
553
00:34:58,150 --> 00:35:02,430
they can train off of that. Has like 90% margins, but they lose
554
00:35:02,490 --> 00:35:08,430
a lot of the the actual money on the subscription, and then at
555
00:35:08,460 --> 00:35:11,580
the same time they have to put hundreds of millions of dollars
556
00:35:11,790 --> 00:35:17,310
into training runs, and then that is extremely like
557
00:35:17,460 --> 00:35:21,680
unsustainable. If there can be other models out there that can
558
00:35:21,740 --> 00:35:25,790
just take all of the information that you get from a training run
559
00:35:26,210 --> 00:35:30,230
and essentially use that to train their models themselves
560
00:35:30,350 --> 00:35:34,610
too, or a 10th or you know 1/20 of the cost of doing it
561
00:35:34,610 --> 00:35:37,850
themselves. And so it's it's widely rumored right now that
562
00:35:37,850 --> 00:35:40,610
many of those models are actually just not releasing
563
00:35:40,760 --> 00:35:44,290
their latest ones to the public, and folks like Anthropic are
564
00:35:44,350 --> 00:35:48,070
actually going more into drug discovery. And so, I think that
565
00:35:48,250 --> 00:35:52,450
while consumer and business serving of you know creating
566
00:35:52,540 --> 00:35:55,210
models and offering it to the public and having them pay for
567
00:35:55,300 --> 00:35:59,200
it is something that's not necessarily going to be a great
568
00:35:59,260 --> 00:36:01,890
long term business model because of the cost of distillation,
569
00:36:01,980 --> 00:36:05,370
don't are significantly lower than the cost of training models
570
00:36:05,550 --> 00:36:09,930
yourself. There, that could be a tough business model. So I think
571
00:36:10,380 --> 00:36:13,770
I wouldn't be super excited about owning Anthropic or or
572
00:36:13,920 --> 00:36:18,630
OpenAI for that type of consumer-facing business. The
573
00:36:18,780 --> 00:36:21,860
harnesses and agents are what's actually the most valuable part?
574
00:36:22,460 --> 00:36:26,720
It's the if you've used Hermes or Open Claw or or Open Code,
575
00:36:28,160 --> 00:36:32,360
they are the interface that you use for any sort of model in the
576
00:36:32,360 --> 00:36:37,070
backend, either one that you host yourself or an API that can
577
00:36:37,130 --> 00:36:40,870
be switched between them, and that's really valuable because
578
00:36:41,200 --> 00:36:44,680
it actually is the interface, and it can significantly
579
00:36:44,830 --> 00:36:48,640
decrease the amount of tokens or increase if you have the wrong
580
00:36:48,760 --> 00:36:52,120
harness and the wrong tools that are associated with it. Whether
581
00:36:52,210 --> 00:36:59,020
it's connecting it to your Excel, PowerPoint, email, all of
582
00:36:59,020 --> 00:37:04,560
the other tools that you likely have is is an average user, it's
583
00:37:04,770 --> 00:37:08,640
able to call them and and get hooked up and do the essentially
584
00:37:08,880 --> 00:37:13,110
operating system and interface for you to work with essentially
585
00:37:13,560 --> 00:37:16,170
all of your intelligence, and that's where you can actually
586
00:37:16,260 --> 00:37:19,350
spawn agents and have them be focused on different tasks like
587
00:37:19,860 --> 00:37:24,050
design versus writing copy versus financial modeling and
588
00:37:24,080 --> 00:37:28,160
Excel, and I think that's where a lot of the durable value
589
00:37:28,280 --> 00:37:31,490
lives. Although many of those are still being given away for
590
00:37:31,610 --> 00:37:35,330
free, just because the open source community is so strong
591
00:37:35,450 --> 00:37:38,660
and and they want to really look for bugs, whether it's Open Claw
592
00:37:38,750 --> 00:37:43,090
or Hermes or or whatever it is out there, there are additional
593
00:37:43,150 --> 00:37:47,560
hosting costs that you can benefit from for for those types
594
00:37:47,650 --> 00:37:53,050
of harnesses. But I think that those are the kind of main areas
595
00:37:53,170 --> 00:37:55,840
that that I'm really interested in on my side.
596
00:37:55,930 --> 00:37:58,180
I think that's a great way to kind of move on to
597
00:37:58,210 --> 00:38:01,830
the next level of the stack, which I'm going to put these
598
00:38:01,920 --> 00:38:04,560
together the way Israel kind of combines some of the
599
00:38:04,560 --> 00:38:08,160
infrastructure part of it, but you know the the next layer of
600
00:38:08,220 --> 00:38:10,830
the stack in your report are the agents. And I first of all I
601
00:38:10,890 --> 00:38:13,290
just want to say we'll include a link to the report and to the
602
00:38:13,320 --> 00:38:16,740
website in here. I'd encourage everyone to really take a deep
603
00:38:16,800 --> 00:38:20,100
dive into this because it's fascinating. But to the next the
604
00:38:20,310 --> 00:38:23,180
next layer would be the agents themselves, and they're the the
605
00:38:25,220 --> 00:38:29,870
artificial intelligence that has a goal or has a task or
606
00:38:29,900 --> 00:38:32,960
repetitive task that is out there actually creating the
607
00:38:32,990 --> 00:38:35,630
productivity of it. And associated with that is
608
00:38:35,630 --> 00:38:40,040
identity. So then there becomes the question of, you know, is an
609
00:38:40,220 --> 00:38:45,040
individual AI or an agent has have has a unique identity, or
610
00:38:45,100 --> 00:38:48,190
how do you how do you know which agent is doing which work? And
611
00:38:48,220 --> 00:38:50,770
then on top of that is the coordination among agents. So,
612
00:38:51,370 --> 00:38:54,790
can you can you kind of take a look at that those three layers
613
00:38:54,850 --> 00:38:57,520
of the stack, how they interact and how they evolve one from the
614
00:38:57,640 --> 00:39:01,620
other, from agent to identity to coordination, and is this kind
615
00:39:01,650 --> 00:39:04,290
of the the fear factor that we're seeing right now, or how
616
00:39:04,350 --> 00:39:07,290
do we? I mean, but also the productivity. Like, how much has
617
00:39:07,380 --> 00:39:11,190
that is that unleashing productivity today, and what is
618
00:39:11,280 --> 00:39:13,770
the potential for that going forward? Because I really see
619
00:39:13,890 --> 00:39:17,610
this as kind of the the the the kind of goldmine and how
620
00:39:18,390 --> 00:39:21,200
businesses can unlock potential and individuals can unlock
621
00:39:21,260 --> 00:39:24,860
potential by by utilizing these this agent stack.
622
00:39:26,299 --> 00:39:30,199
Yeah, for sure. I think that right now agents it depends
623
00:39:30,259 --> 00:39:32,749
on how you interact with them, but a lot of them are just
624
00:39:32,809 --> 00:39:37,039
internal. You speak with them on your computer, and they go do a
625
00:39:37,129 --> 00:39:40,449
task. They start to get really interesting when you embed them
626
00:39:40,569 --> 00:39:44,259
with either your company workflows and the agents that
627
00:39:44,379 --> 00:39:48,099
you're interacting with are interacting with other agents
628
00:39:48,129 --> 00:39:51,279
that you're that are created by other folks at your company, or
629
00:39:51,639 --> 00:39:55,749
just interacting with any other agent on the internet for a
630
00:39:56,469 --> 00:40:01,769
shared group of tasks. And so, one of the. Ones was was lobster
631
00:40:02,039 --> 00:40:05,549
claw or something like that about them all getting
632
00:40:05,639 --> 00:40:08,729
identities and and posting on their own Facebook or or
633
00:40:08,819 --> 00:40:13,469
whatever but I think that a lot more of it will be in the future
634
00:40:14,069 --> 00:40:20,159
folks who host a certain amount of compute and they can offer
635
00:40:20,459 --> 00:40:24,709
their agents and to go out and do things in the real world, in
636
00:40:24,829 --> 00:40:27,319
order, and that could be anything from finding
637
00:40:27,529 --> 00:40:31,129
differentiated research that isn't just available and and
638
00:40:31,219 --> 00:40:34,789
paying for that through micro payments, or you know whether
639
00:40:34,999 --> 00:40:40,579
it's being a world class copywriter for advertising and
640
00:40:41,509 --> 00:40:44,229
getting hired by other folks out there because they have a
641
00:40:44,619 --> 00:40:48,969
reputation for doing that at a very high degree. I think that
642
00:40:48,969 --> 00:40:51,669
that's where it starts to get interesting as it relates to
643
00:40:51,909 --> 00:40:56,049
identity and having things like like Nostr on the back end that
644
00:40:56,109 --> 00:41:01,469
will associate end pubs with different agents out there, and
645
00:41:01,589 --> 00:41:05,429
giving them a list, like essentially webs of trust, to be
646
00:41:05,579 --> 00:41:07,769
able to coordinate with different agents. Whether it's
647
00:41:07,949 --> 00:41:12,719
something like Nostr or other protocols out there, like to
648
00:41:12,779 --> 00:41:13,049
date.
649
00:41:13,199 --> 00:41:15,779
Excuse me, Liam. Can you just explain to our
650
00:41:15,839 --> 00:41:18,209
audience that may not be familiar with Nostr what that
651
00:41:18,509 --> 00:41:18,599
is?
652
00:41:18,630 --> 00:41:21,740
Yeah, for sure. It's a it's a coordination system for
653
00:41:24,050 --> 00:41:26,660
there. There's essentially public and private keys, similar
654
00:41:26,780 --> 00:41:29,810
to what Bitcoin has. You know, your private key is your
655
00:41:30,320 --> 00:41:35,210
address. You're proving that you either have the Bitcoin or on
656
00:41:35,240 --> 00:41:39,290
the Nostra side, proving that you own the identity, and then
657
00:41:39,380 --> 00:41:44,110
the public address is essentially your name or a
658
00:41:44,200 --> 00:41:47,020
string of characters that can be associated with the name, and
659
00:41:47,050 --> 00:41:50,260
that can be the agent. So it's tied to the activity that you
660
00:41:50,320 --> 00:41:53,620
do. Or on the Bitcoin side, it's an address that you can use to
661
00:41:54,460 --> 00:42:00,430
receive Bitcoin. And so I think that similarly, like tools like
662
00:42:00,580 --> 00:42:05,550
Jack at Square have created with Buzz, it's allowed for different
663
00:42:05,940 --> 00:42:10,680
agents to interact with your entire team, and so that uses
664
00:42:10,800 --> 00:42:15,330
Nostra as the backend, and it can include different folks who
665
00:42:15,630 --> 00:42:20,400
join your channel in Buzz, and one can be focused on marketing,
666
00:42:20,550 --> 00:42:23,480
one could be focused on sales and go-to-market, or
667
00:42:24,860 --> 00:42:28,880
engineering, product, security, and they can all have different
668
00:42:29,000 --> 00:42:31,700
identities and be different, attached with different things
669
00:42:31,790 --> 00:42:35,270
over time. Similarly, in the real world, whether it's
670
00:42:35,900 --> 00:42:40,720
actually you know paying for different security reviews by
671
00:42:40,840 --> 00:42:44,530
very capable agents who have built up a corpus of knowledge
672
00:42:44,650 --> 00:42:48,190
over time and have different efficiencies as it relates to
673
00:42:48,280 --> 00:42:52,570
skills and and knowledge. I think that similarly, there will
674
00:42:52,600 --> 00:42:58,030
be a lot of value transferred by taking an identity, building up
675
00:42:58,270 --> 00:43:03,510
a history of work that it's done over time, and then using that
676
00:43:03,570 --> 00:43:07,740
to coordinate what you should, what you're willing to pay for
677
00:43:07,860 --> 00:43:09,120
for different future tasks.
678
00:43:09,269 --> 00:43:13,139
Yeah, and the productivity unlock is is going
679
00:43:13,199 --> 00:43:17,399
to be. It's it's still something I have a hard time grasping
680
00:43:17,849 --> 00:43:20,639
because I'm just seeing it in real time, and and it's it's
681
00:43:20,669 --> 00:43:24,259
just absolutely exponential. I mean, a bit of a side here. I
682
00:43:24,319 --> 00:43:28,279
have a I have a packaging business for that that I work on
683
00:43:28,339 --> 00:43:31,789
for restaurants, small coffee shops, those kind of things. And
684
00:43:31,909 --> 00:43:35,359
we're just starting to play around with some of these agent
685
00:43:35,389 --> 00:43:39,259
tools. So specifically Grokbot, and there's a few right. And to
686
00:43:39,439 --> 00:43:43,719
your point of these identities, I mean, you you start to see the
687
00:43:43,779 --> 00:43:47,889
the the real impact of. So we have you know a sales agent, we
688
00:43:48,039 --> 00:43:51,729
have a design agent, we have an accounting agent that's starting
689
00:43:51,819 --> 00:43:56,169
to do a lot of this backend internal work for us. I mean, a
690
00:43:56,229 --> 00:43:58,809
lot of people make parallels to just thinking of them as
691
00:43:58,929 --> 00:44:04,949
coworkers, and in practice, it's true, right? So you you kind of
692
00:44:04,949 --> 00:44:08,249
extrapolate that to the actual changes in the economy and
693
00:44:08,279 --> 00:44:10,979
productivity gains, and and it's it's absolutely mind blowing.
694
00:44:11,039 --> 00:44:13,199
But anyways, I don't mean to deviate.
695
00:44:13,290 --> 00:44:16,260
I was just curious: have you guys seen the Hugging Face?
696
00:44:16,410 --> 00:44:19,590
Just speaking because we came out with it earlier, the the
697
00:44:19,860 --> 00:44:23,540
happy robot thing, the the little ducks that came out last
698
00:44:23,570 --> 00:44:23,750
week.
699
00:44:24,560 --> 00:44:25,250
I have not, Lynne.
700
00:44:25,400 --> 00:44:25,790
Have not.
701
00:44:26,899 --> 00:44:30,289
Okay. Well, it's it's actually funny because you know,
702
00:44:31,159 --> 00:44:33,979
Hugging Face, not like a Bitcoin company or anything, but they
703
00:44:34,069 --> 00:44:37,909
came out with Nostr as a protocol. They they created
704
00:44:37,939 --> 00:44:41,799
these essentially robots that are shaped like docks that are
705
00:44:42,489 --> 00:44:45,099
entirely open source, and you can put your own models in them
706
00:44:45,549 --> 00:44:48,579
for extremely low cost, and they will recursively learn over
707
00:44:48,669 --> 00:44:53,259
time. And they use Nostra as a backend to coordinate the
708
00:44:53,469 --> 00:44:56,289
identity of all the different robots, and then interact with
709
00:44:56,829 --> 00:44:59,949
each other in the real world. And so, I think that that.
710
00:45:00,840 --> 00:45:03,300
It's it's obviously very early, but it's just a sign of
711
00:45:03,390 --> 00:45:06,660
what's to come as it relates to not just you know working with
712
00:45:06,810 --> 00:45:10,920
agents on the computer, but how it actually identifies ties to
713
00:45:11,160 --> 00:45:14,430
actual robots, whether it relates to you know them having
714
00:45:14,940 --> 00:45:17,940
being in Amazon facilities or actually out in the real world
715
00:45:18,090 --> 00:45:21,620
in order to understand where they are with certain tasks and
716
00:45:22,010 --> 00:45:25,400
being able to coordinate exactly who they are over time, too. I
717
00:45:25,460 --> 00:45:28,850
think that it's it's a little bit early, but it's fascinating
718
00:45:29,060 --> 00:45:32,090
to if you extrapolate that out a few years, what what it really
719
00:45:32,180 --> 00:45:35,600
means. Yeah, to your point, I mean this identity and and
720
00:45:35,690 --> 00:45:38,960
coordination component. I mean once that that just takes us
721
00:45:39,020 --> 00:45:41,680
into the actual agent to agent economy, right? Which we're
722
00:45:41,860 --> 00:45:45,580
still at the very, very, very early stages of. And once that
723
00:45:45,670 --> 00:45:48,280
starts to take off, then I mean, yeah, we we move from it just
724
00:45:48,280 --> 00:45:52,810
being, I guess, this internal tool that works for us as humans
725
00:45:52,900 --> 00:45:57,040
and coordinates our efforts to to really handing off complete
726
00:45:57,130 --> 00:46:01,590
tasks and and agents kind of you know really having this value
727
00:46:01,650 --> 00:46:05,490
transfer and settlement and and trading going on between them,
728
00:46:05,610 --> 00:46:10,710
which is just a whole another layer to it. But Lynne, sorry, I
729
00:46:10,770 --> 00:46:12,900
think you were going to add something as well.
730
00:46:12,990 --> 00:46:15,660
This is something that, frankly, Israel
731
00:46:15,660 --> 00:46:17,610
and I have been talking about for quite a while. Is just
732
00:46:17,700 --> 00:46:22,370
opportunities with Bitcoin and and the fact that maybe the
733
00:46:22,490 --> 00:46:26,240
agentic economy and agents themselves could be that unlock
734
00:46:26,390 --> 00:46:29,150
for medium of exchange that maybe we've been looking at the
735
00:46:29,270 --> 00:46:33,740
wrong place instead of having people in evolved countries all
736
00:46:33,740 --> 00:46:36,860
of a sudden switch their payment system to Bitcoin maybe it's not
737
00:46:37,040 --> 00:46:41,560
people maybe it's agents of if you can dive deeper into the
738
00:46:41,650 --> 00:46:44,860
whole value transfer layer of your stack and settlements and
739
00:46:44,890 --> 00:46:45,700
how that ties in,
740
00:46:45,850 --> 00:46:48,190
yeah, for sure. I think that taking a step back too,
741
00:46:48,430 --> 00:46:52,300
where there are two aspects to this too. It's both as this
742
00:46:52,510 --> 00:46:56,140
really unfolds, are you going to save your money because there's
743
00:46:56,170 --> 00:46:59,740
going to be so much deflation out there as it relates to just
744
00:46:59,740 --> 00:47:02,610
the amount of productivity that's coming out as it relates
745
00:47:02,700 --> 00:47:05,040
to every part of the stack getting better and better over
746
00:47:05,100 --> 00:47:09,570
time, and so it really is important to look hard at scarce
747
00:47:09,720 --> 00:47:13,680
assets, whether it's gold or Bitcoin, which is probably my my
748
00:47:13,800 --> 00:47:16,020
personal favorite. But I think that both will do well over
749
00:47:16,110 --> 00:47:21,440
time, and and if you aren't really allocating to scarce
750
00:47:21,650 --> 00:47:24,230
assets, then you're going to see a lot of the gains inflated
751
00:47:24,260 --> 00:47:28,970
away. When looking back at the initial industrial revolution,
752
00:47:29,660 --> 00:47:32,720
while there was so much created and with respect to all the
753
00:47:32,750 --> 00:47:36,440
railroads and a lot of innovations with electricity and
754
00:47:37,670 --> 00:47:40,450
steel, one of the best investments was just to hold
755
00:47:40,540 --> 00:47:43,660
money because it was backed by gold, and there was actually
756
00:47:43,810 --> 00:47:46,660
real deflation there, and so you saw a lot of your purchasing
757
00:47:46,720 --> 00:47:49,720
power increase over time, and didn't have to necessarily pick
758
00:47:49,930 --> 00:47:52,960
what was going to be the best performing asset versus like one
759
00:47:53,050 --> 00:47:57,760
steel company, railroad company, etc. And so that's that's one
760
00:47:57,790 --> 00:48:02,730
piece, but the other one is in a world where you know everything
761
00:48:02,760 --> 00:48:07,380
is kind of constrained by energy and compute. It's almost going
762
00:48:07,410 --> 00:48:12,780
to move towards electricity and how you meter that in terms of
763
00:48:13,170 --> 00:48:20,190
paying for it over time. And so there will be essentially value
764
00:48:20,280 --> 00:48:25,220
tasks that are so small but so critical that will be too small
765
00:48:25,340 --> 00:48:29,480
for credit cards to pay, and so during that it will be crucial
766
00:48:29,570 --> 00:48:34,280
to have internet-native money and money that's native to the
767
00:48:34,430 --> 00:48:38,720
internet. Stablecoins, I think, will play a role there to start,
768
00:48:38,810 --> 00:48:41,890
just because some people are familiar with the dollar and the
769
00:48:42,010 --> 00:48:45,010
network effects that it has, especially offboarding and
770
00:48:45,100 --> 00:48:49,300
onboarding in the U.S. to to different banks. But the world
771
00:48:49,360 --> 00:48:53,530
isn't just U.S. centric. It needs money that's neutral. It
772
00:48:53,950 --> 00:48:57,010
needs money that will actually keep its value over time. And so
773
00:48:57,100 --> 00:49:00,610
I think that that will mostly move towards Bitcoin over time.
774
00:49:01,570 --> 00:49:05,670
There are obviously challenges with really bootstrapping it
775
00:49:05,730 --> 00:49:09,870
right now, and because not not many people unfortunately use it
776
00:49:09,930 --> 00:49:14,040
as payment on a day to day basis, and especially not with
777
00:49:14,190 --> 00:49:18,750
instant settlement. Obviously, Bitcoin takes whatever 10 ish
778
00:49:18,930 --> 00:49:23,090
minutes or so, which may or may not be enough for certain
779
00:49:23,180 --> 00:49:26,930
activities as as AI really accelerates, but I think that
780
00:49:26,960 --> 00:49:30,800
there are a bunch of innovative L 2s out there, and and it's
781
00:49:30,860 --> 00:49:33,530
still honestly pretty early days for those. But whether it's
782
00:49:33,830 --> 00:49:37,970
Lightning, Arc, Spark, and the connectivity between all of
783
00:49:38,060 --> 00:49:40,870
them, I think that that's ultimately the direction that
784
00:49:41,020 --> 00:49:44,830
we're going to move with respect to using micro payments over
785
00:49:44,890 --> 00:49:48,850
time, both because it's it can't be seized, and you know most
786
00:49:48,910 --> 00:49:51,190
people just don't live in the U.S. and don't really have a lot
787
00:49:51,220 --> 00:49:52,900
of use for dollars on a day-to-day basis.
788
00:49:53,199 --> 00:49:56,349
I want to take a quick pause here and and just frame
789
00:49:56,469 --> 00:50:01,409
something because I notice and my suspicion is. A lot of people
790
00:50:02,039 --> 00:50:05,249
sometimes can get confused because we hear inflation and
791
00:50:05,309 --> 00:50:08,639
deflation used in the same, you know, for example, research
792
00:50:08,789 --> 00:50:13,289
papers or arguments, and it can be a bit confusing, right? And
793
00:50:13,409 --> 00:50:17,099
and it's mostly due to the fact that these are kind of competing
794
00:50:17,189 --> 00:50:21,139
forces that we we have. You have deflation as it relates to
795
00:50:21,199 --> 00:50:25,639
technology and inflation as it relates to money. Can you can
796
00:50:25,669 --> 00:50:28,939
you help frame that for for someone who's listening and and
797
00:50:28,999 --> 00:50:32,149
might sometimes get a little bit confused between the words
798
00:50:32,209 --> 00:50:33,259
inflation and deflation?
799
00:50:33,410 --> 00:50:37,520
Yeah, for sure. Whenever I hear about that, I think about
800
00:50:38,030 --> 00:50:40,810
have you guys seen the chart from like the change in prices
801
00:50:40,870 --> 00:50:46,030
since the year 2000, and it's like TVs, iPhones, etc. all
802
00:50:46,330 --> 00:50:50,590
deflating really fast. And then it's on the top side, above
803
00:50:50,680 --> 00:50:54,850
average CPI. It's like healthcare services, education,
804
00:50:55,330 --> 00:51:00,630
etc. And a lot of that is, you know, the the bottom half of the
805
00:51:00,870 --> 00:51:06,540
curve is due to innovation technologically, which can bring
806
00:51:06,600 --> 00:51:10,620
down the cost of the cost of products and services, and and
807
00:51:10,620 --> 00:51:13,980
that's why TVs, etc. get cheaper over time, and that's the
808
00:51:14,130 --> 00:51:17,190
natural state of the economy. People will always get more
809
00:51:17,250 --> 00:51:21,020
productive; otherwise, they're going to get competed out of
810
00:51:21,080 --> 00:51:26,870
their business and lose their job, etc. The so deflation is
811
00:51:26,930 --> 00:51:30,860
the natural state, but there is monetary inflation that is
812
00:51:31,010 --> 00:51:34,460
increasing anywhere from I think it's average at 7% a year over
813
00:51:34,580 --> 00:51:38,630
the past 100 years, and going to likely accelerate further from
814
00:51:38,690 --> 00:51:43,060
here as our debt and deficits start to get to almost untenable
815
00:51:43,960 --> 00:51:49,000
levels, and so that's monetary inflation is is a is a tough
816
00:51:49,180 --> 00:51:53,140
thing for many people because wages and everything that you
817
00:51:53,320 --> 00:51:55,870
own in terms of asset prices doesn't necessarily keep up as
818
00:51:55,930 --> 00:52:00,730
fast. And then on top of that, we see industries that have you
819
00:52:00,760 --> 00:52:04,530
know you can say highly regulated is a friendly way, and
820
00:52:05,040 --> 00:52:09,000
have regulatory capture as a the more pessimistic view. But
821
00:52:09,000 --> 00:52:12,210
everything from healthcare services, education, where
822
00:52:12,270 --> 00:52:15,180
there's essentially government monopolies on how they work,
823
00:52:15,630 --> 00:52:19,230
those tend to see a lot less innovation and thus deflation in
824
00:52:19,500 --> 00:52:23,510
prices, or they are just tied to incentives where they actually
825
00:52:23,600 --> 00:52:26,120
increase prices over time because that's the only way for
826
00:52:26,180 --> 00:52:30,320
those companies to make more profit. And so I think that as
827
00:52:30,500 --> 00:52:33,200
as you look out into the future and try to extrapolate a little
828
00:52:33,230 --> 00:52:38,540
bit more, there will continue to be a lot of technological
829
00:52:38,630 --> 00:52:43,390
deflation as it relates to just industries that are closer to
830
00:52:43,660 --> 00:52:47,980
completely free and have less regulation on them, whereas
831
00:52:48,130 --> 00:52:51,820
those that are heavily regulated will probably be a little bit
832
00:52:51,880 --> 00:52:56,020
closer to the actual monetary inflation rate, which which sees
833
00:52:56,200 --> 00:52:57,820
purchasing power decline over time.
834
00:52:57,940 --> 00:53:00,750
When deflation hits, not just the core
835
00:53:00,960 --> 00:53:04,710
technologies like a television or a component, but hits
836
00:53:04,920 --> 00:53:07,680
traditional industries. How how does how do you see that
837
00:53:07,740 --> 00:53:08,940
impacting the economy?
838
00:53:09,209 --> 00:53:12,629
Yeah, I think that it's um you probably feel it in in the
839
00:53:12,989 --> 00:53:17,069
day to day work that you do too. Like the the costs of producing
840
00:53:17,159 --> 00:53:21,079
podcasts are are going down because um you know the cost of
841
00:53:21,499 --> 00:53:23,929
getting everything, like a microphone, or and all the
842
00:53:23,929 --> 00:53:29,059
softwares. There are a lot better technological advances as
843
00:53:29,119 --> 00:53:33,049
relates to software engineering and hardware manufacturing. I
844
00:53:33,139 --> 00:53:38,569
think that yeah, like looking at like a restaurant, there's going
845
00:53:38,569 --> 00:53:44,259
to be a lot better ways to for the food service distributors to
846
00:53:44,469 --> 00:53:48,249
actually understand where the inks are in their supply chains,
847
00:53:48,369 --> 00:53:51,909
where there is inventory over time, and really drive down that
848
00:53:51,969 --> 00:53:56,979
cost over time. You know, AI will allow for autonomous truck
849
00:53:57,039 --> 00:54:01,409
drivers that so that will decline the amount of people who
850
00:54:01,529 --> 00:54:04,709
actually have to drive the truck from a place A to B and
851
00:54:05,489 --> 00:54:08,519
necessarily drive down the costs there. I think that there are
852
00:54:08,609 --> 00:54:12,179
just so many different ways that it's going to accelerate things
853
00:54:12,299 --> 00:54:17,849
and it's going to touch every part of our economy, especially
854
00:54:17,909 --> 00:54:21,409
every part that uses any sort of intelligence on a day-to-day
855
00:54:21,529 --> 00:54:24,709
basis. If you're not really just going through the motions, then
856
00:54:24,979 --> 00:54:28,009
it's going to drive down costs significantly. And even if you
857
00:54:28,129 --> 00:54:30,679
are just going through the motions, it will allow you to
858
00:54:31,909 --> 00:54:37,009
really think about new ways and to do the job more efficiently,
859
00:54:37,189 --> 00:54:41,229
implement new technologies, etc. I'm pretty bullish on just the
860
00:54:41,559 --> 00:54:44,979
productivity of any country out there with these tools,
861
00:54:46,000 --> 00:54:48,580
yeah, and I think this is a great place to
862
00:54:48,580 --> 00:54:52,810
try and to wind this down. But there's just so much to unpack
863
00:54:52,900 --> 00:54:56,410
in here. But Liam, if if someone listening is going to take away
864
00:54:56,620 --> 00:54:59,800
one idea from the conversation or from your research, what
865
00:54:59,830 --> 00:55:01,020
would. What would you like that to be?
866
00:55:01,620 --> 00:55:07,350
Yeah, I would say really try to understand the how this
867
00:55:07,740 --> 00:55:10,530
is a lot more of an exponential technology than anything that
868
00:55:10,530 --> 00:55:16,020
we've ever seen, including the internet. The just look back at
869
00:55:16,050 --> 00:55:19,080
that chart of how many tokens Open Rotary is doing of a
870
00:55:19,740 --> 00:55:24,560
25,000% increase-that's just astronomical-and there are going
871
00:55:24,560 --> 00:55:27,560
to be so many opportunities that are created out of this. Whether
872
00:55:27,620 --> 00:55:31,820
it's upleveling your own skills, creating new businesses out of
873
00:55:31,850 --> 00:55:36,290
this, I think it's just-I would highly recommend reading this
874
00:55:36,350 --> 00:55:41,260
report to see where, at least I think, value will accrue in this
875
00:55:41,890 --> 00:55:46,480
AI native world, and where it may give you ideas for starting
876
00:55:46,570 --> 00:55:49,300
businesses, et cetera, and really encourage founders who
877
00:55:49,390 --> 00:55:53,800
are on the cutting edge to reach out to us here. And in addition,
878
00:55:53,890 --> 00:55:56,920
too, I would just think about how to really become the most AI
879
00:55:57,220 --> 00:56:00,720
version of yourself because it will only bring you a lot of
880
00:56:00,720 --> 00:56:02,040
benefits in the future.
881
00:56:02,160 --> 00:56:04,920
I know research is a really core part of what
882
00:56:05,010 --> 00:56:08,250
Early Writers does or provides to the ecosystem. You guys have
883
00:56:08,340 --> 00:56:11,430
some great newsletters that I'd encourage everyone to sign up
884
00:56:11,520 --> 00:56:15,210
for. But Singularity Stack is not just, as I understand it, a
885
00:56:15,300 --> 00:56:18,150
one-off report, but you you you have a living part of your
886
00:56:18,180 --> 00:56:21,320
website that will continually update. Can you talk a little
887
00:56:21,350 --> 00:56:24,470
bit about how people can access what you've already done, and
888
00:56:24,470 --> 00:56:27,410
then how to stay on top of the developments that you see since
889
00:56:27,530 --> 00:56:29,180
you've published it?
890
00:56:29,329 --> 00:56:32,689
Yeah, for sure. We, yeah, please check out
891
00:56:32,869 --> 00:56:36,019
earlywriters.com. We publish a bunch of research there. You can
892
00:56:36,109 --> 00:56:39,079
subscribe to our newsletter there too, and then we also have
893
00:56:39,679 --> 00:56:43,569
a weekly newsletter that just goes into that tries to track
894
00:56:43,629 --> 00:56:46,419
the singularity over time, and the singularity, if if we
895
00:56:46,479 --> 00:56:48,999
haven't already mentioned it, it's just defined as the
896
00:56:49,809 --> 00:56:53,829
acceleration that is too great for the human eye to really
897
00:56:53,919 --> 00:56:57,519
measure. And so, while while we can't necessarily measure each
898
00:56:57,579 --> 00:57:00,519
single improvement over time, we're trying to really keep
899
00:57:00,609 --> 00:57:03,779
everybody abreast to the latest developments as it relates to
900
00:57:03,869 --> 00:57:06,899
open source advances, what it means for the economy and
901
00:57:07,169 --> 00:57:11,609
Bitcoin and companies in the space. And so, if you subscribe
902
00:57:11,699 --> 00:57:16,319
to our website, we'll add you to all of our research lists. And
903
00:57:17,429 --> 00:57:19,859
please reach out if you're building companies on the edge
904
00:57:19,949 --> 00:57:23,959
of this too. liam@earlygriders.com, and
905
00:57:24,109 --> 00:57:25,819
really appreciate you having me on today, Lynne.
906
00:57:26,929 --> 00:57:28,969
We thank you for your research. We thank you for
907
00:57:28,999 --> 00:57:31,699
all the thought that you're giving this, and the ideas for
908
00:57:31,759 --> 00:57:36,139
opportunities, both for founders as well as for investors. And it
909
00:57:36,199 --> 00:57:39,139
just gives a lot to think about in a positive way. That I think
910
00:57:39,619 --> 00:57:43,089
you know, with a narrative that is becoming more fearful toward
911
00:57:43,539 --> 00:57:46,329
AI, it's so important to look at what benefits it's achieving
912
00:57:46,509 --> 00:57:50,889
already and what the potential is for the just the advancement
913
00:57:51,249 --> 00:57:54,219
of of where we go from here. So we we really appreciate it.
914
00:57:54,279 --> 00:57:57,219
Thank you for listening to this episode of Build with Bitcoin.
915
00:57:57,369 --> 00:58:00,479
If you found it of value, please take a second to like,
916
00:58:00,569 --> 00:58:03,659
subscribe, or share, this helps the visibility and reach of our
917
00:58:03,749 --> 00:58:06,269
podcast, so that more people are aware of the innovation
918
00:58:06,359 --> 00:58:10,259
happening in and around Bitcoin. It truly means a lot. I'd also
919
00:58:10,319 --> 00:58:12,989
like to take this opportunity to invite you to join us at an
920
00:58:13,049 --> 00:58:16,949
event that Israel and I are hosting, Meeta Tech Talks 2026,
921
00:58:17,519 --> 00:58:21,469
taking place from October 25th to the 27th at the exclusive
922
00:58:21,649 --> 00:58:25,099
Puntamita Resort in Mexico. This will be an intimate gathering of
923
00:58:25,609 --> 00:58:29,689
125 invite-only participants, exploring how the future of
924
00:58:29,749 --> 00:58:33,439
capital is being shaped by the intersection of Bitcoin,
925
00:58:33,799 --> 00:58:37,459
artificial intelligence, and energy. To request an invite or
926
00:58:37,519 --> 00:58:41,829
for more information, visit mitatechtalks.com. We hope
927
00:58:41,889 --> 00:58:45,099
you'll join us, and thanks again for being a part of the Build
928
00:58:45,129 --> 00:58:46,179
with Bitcoin community.