Rob Wallace | Bitcoin News

Speakers argue AI deflation and infrastructure bottlenecks position Bitcoin as the optimal settlement layer for agentic commerce.

1h 1m Listen to the original Episode page
Artificial IntelligenceBitcoinDeflationNvidiaData CentersAgentic CommerceCryptocurrencyEnergy

Summary

Mark Suman and Liam Nelson discuss the 'Singularity Stack,' arguing that AI progress will cause significant deflationary pressure on wages and prices, potentially breaking the debt-based economy. They contend that Bitcoin serves as the primary asset for capturing value in this deflationary environment, acting as both a store of value and a native settlement rail for AI agent microtransactions. The speakers highlight severe bottlenecks in energy and compute, noting over 2,000 gigawatts in interconnection queues, while discussing the risks of Nvidia's customer concentration and circular financing. They also explore the shift toward local AI models for privacy and the prediction that 2027 will mark the rise of agentic commerce.

Episode sections

  1. 0:15 – 2:08

    AI and the future of money

    Host Rob Wallace introduces the topic of the AI trade, noting recent market volatility and political shifts. He frames the discussion around whether AI is breaking and who will ultimately win, suggesting that whoever holds non-printable money will benefit from the deflationary nature of AI.

  2. 2:08 – 5:53

    The singularity stack

    Liam Nelson presents the 'Singularity Stack' thesis from an August report, arguing that AI progress will cause significant deflation. He suggests that while value will accrue at various layers, Bitcoin is the primary asset for capturing value as the economy adjusts to cheaper intelligence.

  3. 5:53 – 10:09

    Bottlenecks in the AI stack

    Speakers discuss bottlenecks in energy and compute, noting over 2,000 gigawatts in interconnection queues. Mark Suman highlights the fragility of the infrastructure's financials, comparing the current state to the dot-com bubble, while Liam Nelson points to SpaceX's high pricing as evidence of demand.

  4. 10:09 – 16:07

    Local vs. cloud AI models

    Mark Suman argues for a hybrid future where local models handle routine tasks and cloud models handle complex ones. He believes local AI will grow as models become cheaper, but cloud data centers will remain superior for powerful computing due to infrastructure advantages.

  5. 16:07 – 21:55

    Data center controversy

    The speakers analyze public resistance to data centers, attributing it to low daily usage of AI tools and fears of a K-shaped economy. They discuss how political majorities might slow permitting but cannot stop the software distribution of AI.

  6. 21:55 – 26:44

    China and open-weight models

    Mark Suman suggests China may release open-weight models to crash AI inference costs and undermine US frontier labs, similar to OPEC's strategy with oil. He notes studies showing US models are more censorious regarding Chinese propaganda than Chinese models.

  7. 26:44 – 30:59

    The Jensen/Nvidia situation

    The hosts discuss Jensen Huang's personal scandal, with Mark Suman praising his innovation and questioning US intelligence's prior knowledge. Liam Nelson views the scandal as a single point of failure risk but believes Jensen will likely remain in charge.

  8. 30:59 – 32:52

    Nvidia's market concentration

    Liam Nelson argues that Nvidia benefits from open-weight models by standardizing infrastructure, predicting they will catch up to frontier capabilities. Mark Suman notes the extreme difficulty in acquiring GPUs, describing the market as cutthroat.

  9. 32:52 – 41:29

    Nvidia Concentration and Circular Financing Risks

    Rob Wallace asks about Nvidia's customer concentration, noting five customers make up 75% of sales. Speaker 2 compares this to Bitcoin miners in 2021-2022, citing circular financing as a risk, though notes chip customers are currently becoming more profitable. Speaker 1 agrees circular references are risky and could lead to legislative intervention after a catastrophe.

  10. 41:29 – 45:02

    Privacy and Data Security in Cloud AI

    Speaker 1 explains that using cloud AI models like Claude or ChatGPT risks leaking proprietary business secrets, as labs may train models on user data. He notes that OpenAI and Anthropic have disabled 'zero data retention' features for safety scans, making encryption and local processing necessary for security.

  11. 45:02 – 49:21

    Maple Work: Local AI and Multi-Agent Architecture

    Speaker 1 introduces Maple Work, a local AI app that runs in secure enclaves to prevent data leakage to frontier labs. It supports multi-agent workflows, allowing users to keep core data private while selectively sharing small snippets with external models like Claude for specific tasks.

  12. 49:21 – 56:38

    Bitcoin as AI's Native Money and Microtransaction Rail

    Speaker 2 argues AI will cause deflationary pressure on wages, making scarce assets like Bitcoin valuable stores of value. Speaker 1 adds that AI agents will need to perform microtransactions for compute, favoring Bitcoin over credit cards or fiat due to efficiency and the ability to hold non-debased currency.

  13. 56:38 – 1:01:17

    Agentic Commerce and 2027 Outlook

    Speaker 1 predicts 2027 will be the year of 'agentic commerce,' where AI agents interact directly with other agents rather than humans. Speaker 2 expresses strong bullishness for Bitcoin in 2027, citing market apathy and miner selling as contrarian signals.

Assets discussed

AssetKindStanceContext
BitcoinBTCcryptomixedAnd much of that is Bitcoin, similar to how during the first industrial revolution there was significant deflation and holding gold was one of the best investments.
Goldcommoditybullishand holding gold was one of the best investments.
NVIDIANVDAequitymixedAnd NVIDIA stock has taken a big hammering.

“Stance” is how the speakers talked about the asset, not a recommendation.

Claims worth verifying

The NASDAQ had its worst day since 'Liberation Day' six days prior to the broadcast.

Rob Wallace ·0:35

How to verify: Check NASDAQ composite index daily performance data for the week prior to 2026-11-12 and search for the term 'Liberation Day' in financial news archives to identify the referenced historical event. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

The incoming governor of Texas wants a freeze on nuance.

Rob Wallace ·0:55

How to verify: Search for news articles or official statements from the elected Governor of Texas regarding 'nuance' or data center regulations in the week following the 2026 election. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Over 2,000 gigawatts are in interconnection queues for data centers.

Unidentified speaker (speaker_2) ·7:01

How to verify: Review reports from the American Clean Power Association or the Electric Power Research Institute (EPRI) on grid interconnection queues. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

American models are more censorious when it comes to Chinese propaganda than Chinese models.

Unidentified speaker (speaker_1) ·22:53

How to verify: Search for academic or industry studies comparing content moderation policies of US-based LLMs (e.g., OpenAI, Anthropic) versus Chinese LLMs (e.g., Qwen, DeepSeek) regarding political topics. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

There are tens of millions of people who use an AI agentic harness on a daily basis.

Unidentified speaker (UNKNOWN) ·17:15

How to verify: Search for market research reports from firms like Gartner, IDC, or Statista on daily active users (DAU) of AI agents or agentic platforms. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Nvidia has five customers that make up 75% of their sales.

Unidentified speaker (UNKNOWN) ·32:52

How to verify: Check Nvidia's latest 10-K filing with the SEC EDGAR database under 'Risk Factors' or 'Financial Statements' for customer concentration disclosures. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

Anthropic and OpenAI have disabled zero data retention features for safety scans.

Unidentified speaker (speaker_1) ·43:27

How to verify: Review official announcements or policy updates from Anthropic and OpenAI regarding their data retention policies and safety scanning procedures. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Bitcoin miners sold a lot of their Bitcoin, with miners exiting leverage at the top.

Unidentified speaker (speaker_2) ·59:09

How to verify: Analyze on-chain data from blockchain explorers or reports from crypto analytics firms (e.g., Glassnode, CryptoQuant) showing miner reserve changes and exchange inflows during the relevant period. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

AI agents prefer Bitcoin over any other payment method, according to a BPI study.

Rob Wallace ·54:22

How to verify: Locate and review the specific study published by BPI (Blockchain Privacy Initiative or similar entity) cited in the transcript. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Maple Work is in beta and will have a public release soon.

Unidentified speaker (speaker_1) ·45:08

How to verify: Check the Maple AI website (maple.ai) or official press releases for product launch status and beta availability. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Maple Work uses confidential compute with secure enclaves.

Unidentified speaker (speaker_1) ·45:47

How to verify: Review technical documentation or whitepapers from Maple AI describing their security architecture and hardware enclaves. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Rhino offers loans through Arch.

Rob Wallace ·48:43

How to verify: Check Rhino's website or partnership announcements for details on their lending integration with Arch. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Speaker 2 is bullish on Bitcoin for 2027 due to market apathy.

Unidentified speaker (speaker_2) ·59:09

How to verify: This is a subjective opinion expressed by the speaker; verify the sentiment by reviewing the transcript or subsequent public statements by Liam Nelson. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Primary source

Maple accepts Bitcoin for payment and offers anonymous accounts.

Unidentified speaker (speaker_1) ·1:00:02

How to verify: Check the Maple AI website or payment terms for Bitcoin acceptance and anonymity features. For recent events, verify 2026-08-12 through 2026-11-12. Use another period only when the speaker explicitly names it.

Do your own research

  • Verify Nvidia's Q3 2024 earnings report for customer concentration data (specifically the top 5 customers' share of revenue).
  • Check the US Energy Information Administration (EIA) or grid operator reports for current interconnection queue capacities and wait times.
  • Review Anthropic and OpenAI's current privacy policies and terms of service regarding data retention and model training on user inputs.
  • Research the 'Singularity Stack' report by Liam Nelson to assess the original thesis against current market developments.
  • Analyze Bitcoin miner profitability and hash rate trends to evaluate the 'contrarian signal' claim regarding miner selling and market apathy.
  • Investigate the technical specifications of 'Maple Work' and 'confidential compute' enclaves to verify data isolation claims.
  • Compare historical deflationary periods (e.g., Great Depression, 1920s) with current AI productivity metrics to assess the validity of the deflationary wage argument.
  • Check for recent legislative proposals or regulatory actions targeting AI data privacy or cloud computing infrastructure in the US and EU.
Full transcriptexpand
0:15
Rob Wallace: Hello, hello, hello. Welcome back to Bitcoin News Live. I'm your host, Rob Wallace. It's been a couple of weeks since we've done a live stream. I've been traveling the world, it was at a conference down in Punta. Mita, Mexico, then over in Istanbul. But now I am back in the home base in the studio and happy to be having this conversation today. We are going
0:35
Rob Wallace: to dive into the AI trade. Is it breaking? What does it mean for the future? And how does Bitcoin come out the ultimate winner? All right, let's dive right into it. Six days ago, the NASDAQ had its worst day since Liberation Day. Voters handed Democrats both houses of Congress. and a lot of them did it to stop the data centers. The incoming governor of
0:55
Rob Wallace: Texas wants a freeze on nuance. And the Board of Nvidia is investigating its own CEO, the one man holding the whole entire AI trade together. And the biggest AI IPO in history, Anthropic, keeps getting marked down before it even hits the market. So here is the really, really, really important question. If
1:15
Rob Wallace: AI is the most important technology of our lifetimes, and the chips, the power, and the politics can all break in a single week, who actually wins? My guests today say it's not NVIDIA, and it's not whichever lab has the best model this month. AI is going to make intelligence nearly free and almost everything else cheaper with it. The winner is
1:35
Rob Wallace: whoever holds the money that cannot be printed while that happens.
1:39
Unidentified speaker (UNKNOWN): All
1:40
Rob Wallace: right. Today I am joined by Mark Suman, the CEO of Met Maple AI, and Liam Nelson of Early Writers. Gentlemen, let's get you up on stage. Guys, how are we doing? Mark, nice to meet you.
1:54
Unidentified speaker (speaker_1): Doing great. Nice to meet you as well.
1:56
Unidentified speaker (UNKNOWN): Thanks
1:56
Rob Wallace: for coming on. And Liam, how are you?
1:58
Unidentified speaker (speaker_2): I'm doing fantastic. I love the intro and I love the energy.
2:01
Rob Wallace: Let's go. Yes. That's what we do here at Bitcoin News. We bring the energy and the information. Keep the people going. All right. So, Liam, let's start with you. You in August released a report called the Singularity Stack. Give us the thesis in 60 seconds.
2:18
Unidentified speaker (speaker_2): Yeah, for sure. Um, so essentially the singularity is when the things are moving so quickly and progress is happening so fast that it can't be recognized. by the human eye. And we're always trying to really understand and show the market where value will accrue over a long period of time. So it goes through every part of the layer that is
2:39
Unidentified speaker (speaker_2): needed and efficient in the stack, everything from energy, the base, compute, models, harnesses, agents, identity, coordination, value transfer, and settlement.
2:51
Unidentified speaker (speaker_2): And a lot of this go uh essentially argues: is information or is value going to be captured at each of these layers, or will it just accrue to the consumer? Much, much of the progress that is going to happen is going to end up accruing to the consumer, creating significant deflationary impacts, which is
3:11
Unidentified speaker (speaker_2): going to really break the current economy based. on how much debt that there is if wages and prices continue to fall based on all the progress. And so while there is significant value to be captured at some areas, like harnesses and identity coordination and payments, that
3:31
Unidentified speaker (speaker_2): uh I think will be a little bit difficult to completely um understand exactly how that's going to work today. Much of this is going to be held and captured into where value actually settles at the end state. And much of that is Bitcoin, similar to how during the first industrial revolution there was significant deflation
3:51
Unidentified speaker (speaker_2): and holding gold was one of the best investments.
3:57
Unidentified speaker (speaker_2): Revolution holding Bitcoin will be. Um, But there will be significant great investments too. uh, like happened with the robber barange and and significant value created too. But for many people, just holding core money like Bitcoin or or gold will be the best way to play this.
4:14
Unidentified speaker (UNKNOWN): Thank
4:14
Rob Wallace: you. You just uh summed it up quite nicely there. Uh and Mark, you are a first-time appearance on Bitcoin News. Give a little bit of your background and whether you agree with this thesis.
4:25
Unidentified speaker (UNKNOWN): Sure,
4:25
Unidentified speaker (speaker_1): yeah. So my background. I come from the tech industry. I've been working at tech startups, done multiple different startups as an engineer, and then also worked at the biggest companies. I was at Apple for six years working on AI, machine learning, technologies with a big focus on privacy while I was there. So I'm coming at it from a technologist
4:45
Unidentified speaker (speaker_1): standpoint. Now I'm working on Maple AI, and Maple is a private AI, and we we do it using end-to-end encryption. We use confidential computing. We very much have a uh let's call it a secure cloud for um for processing AI data. So we use a lot of these GPUs, we use uh you know
5:05
Unidentified speaker (speaker_1): a lot of this infrastructure that's out there. And so we're paying attention to this uh this shift the that's going on. And um no, I I think that there's a lot of interesting stuff that that you just chalk. talked talked about in the thesis that you've put together there. Um we we build in the harness area. So there's there's so many different layers on the stack that you can be at, you know, um so
5:25
Unidentified speaker (speaker_1): we've we do a little bit on the inference side, but our our biggest focus right now has been on building at the application layer, building a harness, building a chat product, building integrations for people to bring all of these tools together so that um so they can really harness everything that's out there. And we're leveraging these open models, these um, you know, these cheapers models, the things that um
5:45
Unidentified speaker (speaker_1): every every few weeks, every few months, something better comes out and we're able to stick it in there and give our users the the latest and greatest.
5:53
Unidentified speaker (UNKNOWN): So
5:54
Rob Wallace: let's go through all the different layers of AI, right? I mean, to go bring it back to Adams, it really starts with with land and electricity, right? Like these are the two most fundamental components of the data center build-out and whatnot. Uh Liam, talk about all these different pieces that need to come together for AI to exist the way we have it today. And
6:13
Rob Wallace: is this just a bottleneck trade? And what do you think is the most important part of this stack?
6:20
Unidentified speaker (speaker_2): Yeah, I mean, um, to an extent it it is a bottleneck trade in terms of what's actually going to accrue value today. but that doesn't necessarily mean that it's always going to be that in the future. So right now, the some of the biggest bottlenecks are uh, and there are multiple across the entire stack, but are energy and compute, getting those stood up.
6:40
Unidentified speaker (speaker_2): And um that's why you're seeing companies like SpaceX being able to charge an arm and a leg like for some of their data centers and their spot pricing to companies like Anthropic, because everybody. wants more um inference and to be able to run it at a cheaper level. So there's a significant amount of um over like
7:01
Unidentified speaker (speaker_2): 2,000 gigawatts that are in interconnection queues trying to get stood up and they are, you know, everything that you're seeing about both the existing permitting and it's going to be significantly tougher uh, just like looking online and seeing everything out there about how every single politician, whether it's honestly
7:21
Unidentified speaker (speaker_2): right or left, is going to try to slow down the build out of data centers in their community. And so I anticipate that that will continue to be a bottleneck. And um, there are people that are a little bit above my pay grade that significantly think that all of this will end up going into space over time just because of the limited.
7:39
Unidentified speaker (speaker_2): land and compute and then just the uh ability to power um data centers through the sun. Um that's a little bit above my pay grade, but I think that right now, yeah, there's a lot of money to be made in uh buying the bottlenecks because there the demand for all of this is so significant that um, you know, everybody can win in
7:59
Unidentified speaker (speaker_2): the short
8:00
Unidentified speaker (UNKNOWN): term.
8:01
Rob Wallace: Yeah, Mark, I'd would love your take on. on on the on the stack.
8:05
Unidentified speaker (UNKNOWN): Yeah
8:05
Unidentified speaker (speaker_1): on the stack. I mean there there's uh I I can't even go into the whole you know CapEx table and how there's circular references everywhere. Um it it does feel like there are a few people holding up this entire industry right now when it comes to the financials behind it. And so I I uh that's that's well
8:25
Unidentified speaker (speaker_1): beyond my ability to understand right now. And so that's actually one reason why we haven't uh why we haven't really gone into that market so much is that we've tried to stay in a part of the stack that we have more control over. And I think really that's the that's the part of the stack that benefits most consumers and benefits most businesses. Businesses are really these individual entities
8:45
Unidentified speaker (speaker_1): that work on their own.
8:47
Unidentified speaker (speaker_1): A capital, a free market is really is really this decentralized market of entities all working for their own best interests and then finding reasons to work together. And so to me, that's the most exciting place to play is to to play in the area where you can help bring people together so that they can find additional value from
9:07
Unidentified speaker (speaker_1): that. So yeah, the
9:10
Unidentified speaker (UNKNOWN): yeah,
9:10
Unidentified speaker (speaker_1): I mean, I I think the that infrastructure thing is very is is kind of. fragile in the sense that I think that all of the the levered debt that is up against this trade right now, it could break quite easily.
9:25
Unidentified speaker (UNKNOWN): And
9:25
Unidentified speaker (speaker_1): I know a lot of people talk about the dot-com bubble bursting, and um, I do see a lot of similarities there. And what what shakes out of it is I think it's a temporary break. I think we see some things go down, and we're already seeing um what what's it called? Uh Intel stock is already down almost 50% from where it was just last month. And NVIDIA stock has taken
9:45
Unidentified speaker (speaker_1): a big hammering. So I think we might see some temporary things that hurt really bad for a while. But then those who are positioned well from the technology standpoint and from really good business relationships and are selling products and creating products that people really need, they're gonna they're gonna continue to thrive. and they're gonna continue on. You just have to
10:05
Unidentified speaker (speaker_1): outlast whatever kind of downturn there might
10:07
Unidentified speaker (UNKNOWN): be.
10:08
Rob Wallace: And Liam, in your report, you bring up the point that like light, electricity, it became too cheap to meter, right? Uh no one thinks about the cost of turning on a light bulb in their home. Uh is that what happens with intelligence? Is intelligence gonna going to be too cheap to meter in the future?
10:24
Unidentified speaker (UNKNOWN): I
10:24
Unidentified speaker (speaker_2): mean, yeah, I think it's so funny that people used to work in television. entire days uh for one hour of uh additional light. Um and now that's something that nobody thinks about.
10:34
Unidentified speaker (speaker_2): For high quality intelligence, I think that that's the direction that we go for the average Joe. But um at a larger scale, I think that um industrial like you know Claude and and Anthropic will have. to meter uh how much intelligence that they can produce in order to
10:54
Unidentified speaker (speaker_2): really uh derive the benefits from it. But uh like similarly, I think Mark, you were you're making a few great points. Like this is going to be a logarithmic decline in terms of the cost per task that you can accomplish. And there are going to be so many businesses that could be built on the outgrowth of it. And while there will definitely
11:14
Unidentified speaker (speaker_2): be some blood in the streets, um, like thinking about like Instagram being able to benefit from both the faster computing, more connections on the network overall, and significant declines in the cost to take a picture over like a certain amount of time. That created one of the biggest businesses ever with the um
11:34
Unidentified speaker (speaker_2): internet, and you know, came out of the dot-com dot com era a little bit on the later side. I think that right now we can't even necessarily fathom what will be fantastic businesses as some of these uh costs that decline all at the same period of time will continue to, you know, just create some of the most incredible opportunities that we've
11:54
Unidentified speaker (speaker_2): ever seen.
11:55
Rob Wallace: Yeah. And Mark, I would love Of your take on that. How do you plan for a business where your the thing that you're selling gets cheaper and cheaper?
12:03
Unidentified speaker (speaker_1): Yeah. Well, um, I mean, to your point, Liam, yeah, I do think that these models just make it so much cheaper over time for people to run. We have people talking about how they want infinite tokens. And so they look to local AI to be a way to run infinite tokens. And local AI is this term that I think a lot of people picture is it's just like
12:23
Unidentified speaker (speaker_1): my laptop or my phone, right? But local is really it's it's the infrastructure you control. So local for some people could be a mini server rack that they have either in their house or they have in you know some little shed in their backyard, or or maybe they own some land somewhere or they have a business where they run their own stuff. Local
12:43
Unidentified speaker (speaker_1): can actually take on a lot of different shapes and forms. Um my uh my honest take and long-term take is that the cloud is. not gonna go anywhere. And the cloud is always gonna be able to run more powerful computers. It has it has easier power requirements, it can run much larger um power conduits into it,
13:03
Unidentified speaker (speaker_1): it can have better cooling, and it has it can deal with loud noise, right? And so data centers will always be able to run better stuff. And so to me, that means we're just gonna have a hybrid in the future. when it comes to compute and when it comes to AI, where I think that people will be most benefited, and I just I just see it going this way, to running local models first. So
13:23
Unidentified speaker (speaker_1): they'll have smaller models. They're gonna be very capable. I mean, we're already seeing really capable models, both the Deep Seeks, the Quens. You see these models that are really good and powerful, but they're still not as good as the ones that can run in data centers. And so you just need a mix. You start small. you start local, you use what you can out of that. And then what and then it can be
13:43
Unidentified speaker (speaker_1): intelligent enough to know hey, I need something stronger for this one specific task. Then it flexes up to the cloud, gets that task done, and comes back. Um and to me, that's uh, you know, that's that's a much better arrangement for companies and for the consumer rather than just constantly going to the cloud for every single request and using this massive.
14:03
Unidentified speaker (speaker_1): model to answer, you know, what is the temperature today? Like uh we don't need to be doing
14:08
Unidentified speaker (UNKNOWN): that.
14:09
Rob Wallace: And out of curiosity, what kind of tasks right now are out of the reach of the open weight models that you need to go to the big boys for?
14:19
Unidentified speaker (speaker_1): I mean, so in theory, any task can be done, right? It's just how much uh how smart do you want it to be when doing it? So if I'm going to do corporate strategy and I want to build out this really lengthy multi year plan, maybe I can start local to figure out okay, what are some of the pieces I need in place? But when it
14:39
Unidentified speaker (speaker_1): comes time to actually like putting on the big brain thinking, I might want to flex up to cloud to do that. And then you look at these these swarms of agents and these swarms of AI that have solved these math problems that OpenAI has talked about. and some of the stuff that Anthropic is doing with their chemistry labs. You you might not want to just have and just kind of YOLO with a smaller model on your
14:59
Unidentified speaker (speaker_1): laptop and say, like, I'm gonna do this. So you but you can use your local models to coordinate those kinds of efforts and then say, okay, now let me farm out this thinking to something that that can think on a bigger level.
15:11
Unidentified speaker (speaker_2): Yeah, exactly. I I really see like to your point. thinking about like corporate strategy, like you can really just talk into Claude, OpenAI, or or any of the most capable models about exactly the inputs, outputs you want, like how your a deck should be formatted and it will come out with the most incredible um model that you've ever seen that is significantly
15:31
Unidentified speaker (speaker_2): more than I could ever do in my former life in and finance, as well as a deck like fairly instantly. Um but the the local models are
15:39
Unidentified speaker (speaker_2): are kind of like chat GPT like a year or two ago if if they're just like running on on your local computer with a very limited amount of RAM, but they don't necessarily uh have all the capable tool uses or at least um for most people's harnesses like they're they're not quite there. And so um they they most people will just
15:59
Unidentified speaker (speaker_2): go to the larger ones out of convenience for the larger multi-step processes I'm seeing today.
16:07
Rob Wallace: Now one subject I want to jump into is of course the election happened last week. And one of the main issues on the ballot was data centers, right? Uh people across America seem to hate them. They do not want to be around them. People would rather live close to a nuclear reactor uh than they would next to a data center. Mark,
16:27
Rob Wallace: what do you explain? What do you think explains this abject hatred for data centers in America?
16:34
Unidentified speaker (UNKNOWN): I
16:34
Unidentified speaker (speaker_1): think it's a combination of a few things. I think people in general,
16:40
Unidentified speaker (speaker_1): it the
16:42
Unidentified speaker (speaker_1): AI um products that are out there and the the AI usage, if you look at how much it's being used on a daily basis,
16:49
Unidentified speaker (UNKNOWN): it
16:49
Unidentified speaker (speaker_1): it kind of feels like you can. talk to anybody on the street and they've heard of ChatGPT and they say that they have used it, right? But the actual like daily pervasive usage of of AI and seeing real benefit from it, using these these tools every single day to gain real value out of it, at that uptake is still very small. Um
17:09
Unidentified speaker (speaker_1): I mean, we're talking tens of millions of people who actually use an AI agentic harness.
17:15
Unidentified speaker (UNKNOWN): uh
17:15
Unidentified speaker (speaker_1): on a daily basis. And so I in my view, the way that that this is playing out right now is that not enough people have actually seen the value of AI personally in their lives. And so if someone comes to you and says a bunch of scary things about a data center,
17:32
Unidentified speaker (UNKNOWN): um,
17:32
Unidentified speaker (speaker_1): if you don't actually feel like that's something that that you need in your life, then are you going to take the time to really look into it and research all the points that they're making? Um, or are you going to take the word of someone who seems more credible than you? Um, and I think it's just natural human behavior. So I can understand why people might be
17:52
Unidentified speaker (speaker_1): fearful or voting one way or another on this if they haven't actually taken the time or haven't felt the need to have that in their life.
18:01
Unidentified speaker (UNKNOWN): And
18:02
Rob Wallace: how do you see
18:02
Unidentified speaker (UNKNOWN): that?
18:02
Rob Wallace: it, Liam?
18:03
Unidentified speaker (speaker_2): Yeah, I would completely agree. And I would also say, in addition to, and other people have talked about this as well, but people are just seeing data centers as a conduit for the furthering of the K-shaped economy too, of all these, especially Dario saying like half of entry-level jobs are going to be gone within a certain amount of time and all the fear
18:23
Unidentified speaker (speaker_2): mongering that has stopped because you know they're doing their road shows. now. But uh that's a different topic. Like I think that hearing that, not necessarily seeing the benefits in your life, and then seeing a worsening of a K-shape economy that isn't just like K from like a linear aspect, it's K from like exponentials on both sides of
18:43
Unidentified speaker (speaker_2): like the people who are getting the best with these tools are becoming more and more valuable. They're having to hire less employees and thus they're able to scale really large businesses while um you know the founders can keep a very significant amount of profits. Whereas people who are on the the lower shape of the key that um you know have a just
19:03
Unidentified speaker (speaker_2): wage job and and some of their tasks can
19:06
Unidentified speaker (UNKNOWN): you know
19:07
Unidentified speaker (speaker_2): agreeably like potentially be deterministic in terms of like there is a correct answer and and those could be potentially automated away. While there will definitely be other jobs that can come and fill that void, I think that they're increasingly seeing just concern about where the economy is going and using uh data centers as one of the
19:27
Unidentified speaker (speaker_2): conduits is like uh pushing back on on what's happening globally.
19:32
Unidentified speaker (UNKNOWN): And
19:32
Rob Wallace: Mark, do you think that the new Democratic majorities in Congress are going to be able to slow down AI in America?
19:41
Unidentified speaker (speaker_1): I think that they, well,
19:43
Unidentified speaker (speaker_1): so yes, in one sense, I think if they want to slow down where the AI is computed,
19:49
Unidentified speaker (UNKNOWN): they
19:49
Unidentified speaker (speaker_1): can slow down AI in America in that's you know, in that sense. They can they can slow down where the data centers are built if they want to. They can slow down permitting.
19:58
Unidentified speaker (UNKNOWN): Um
19:58
Unidentified speaker (speaker_1): they talked about it on uh the all-in podcast over the weekend, right? That these these data centers. need so many different kinds of permits that even those that are already in flight and being constructed could be held up by the the last permits that they need. So I think that um if they want to, they can try and slow down that part of it. But AI usage is much more difficult to slow down
20:18
Unidentified speaker (speaker_1): because this this is just software. And we've seen time and time again that open source software will find where where it needs to go. People who need it will find it and will use it. And these open weight models are already out there and they are going to continue to be developed and pushed out there. And um, they can try to legislate all they want to around it. It's gonna be very difficult.
20:38
Unidentified speaker (speaker_1): They can certainly slow down aspects of it, but the usage uh will will find the people that want it. And so what they could end up doing is just driving the compute outside of the
20:50
Unidentified speaker (UNKNOWN): computer.
20:51
Unidentified speaker (speaker_1): country into other countries. And so now all these tokens will be generated and used on someone else's soil, which I don't think benefits America. If we're looking at it from an America first standpoint, we should be wanting to have that that technology happen here within our borders.
21:10
Unidentified speaker (speaker_2): Yeah, a hundred percent. Yeah. I think that um both the jobs aspect of it too and and um just by creating some of these data centers and completely understand if people don't necessarily want it in their backyards, but uh that's not necessarily where most of these are coming from. And I do think that um similar to your point, like they can slow down the second derivative of AI,
21:30
Unidentified speaker (speaker_2): meaning like regardless if there's not another data center constructed, there will be algorithmic improvements and improvements in training that will allow for increased usage for everybody today. And then people will want to necessarily use them on their own. So it's only going to slow down the growth of it rather than like the
21:50
Unidentified speaker (speaker_2): total um the rather than slowing down AI as a whole.
21:55
Rob Wallace: And speaking of other governments, I mean we've got China who plays a big role in this, especially with the open weight models. Mark, are there any concerns about using open weight models from China? And didn't they just recently in the last couple weeks put out some proclamations about wanting to keep that technology on their soil too?
22:16
Unidentified speaker (speaker_1): So there's a lot of benchmarks out there and studies being done. And we we should always be continuing and furthering these studies. But
22:24
Unidentified speaker (UNKNOWN): one
22:25
Unidentified speaker (speaker_1): of the big theories around why China is doing these open models is that they want to get their ideology out into the world, right? And the easiest way to do that is just to spread it for free. And um, so I think people have a lot of fear using these models because they think that they are going to be Chinese propaganda. And uh, on
22:44
Unidentified speaker (speaker_1): that front, some of the studies that have come out are actually showing that the American models, the the frontier lab models,
22:53
Unidentified speaker (speaker_1): are actually more censorious when it comes to Chinese propaganda, and that you can you can actually talk about more things when talking to the Chinese models than you can when talking to the frontier lab models. Um that's been quite a surprise.
23:07
Unidentified speaker (UNKNOWN): I
23:07
Unidentified speaker (speaker_1): think another one is that China has the ability to undermine this whole debt bubble thing that we're blowing up with the AI with the AI build out that we talked about at the beginning of the show. And it's a it's a similar playbook to what happened in the United States when it came to oil and
23:27
Unidentified speaker (speaker_1): fracking back 10 plus years ago, where America was finally a net exporter of oil because of fracking, but fracking was more expensive than just traditional oil extraction. And so the OPEC companies in the Middle East saw an opportunity where they could just crash the price of oil and put these fracking companies out of
23:47
Unidentified speaker (speaker_1): business to a point where they can no longer, they can no longer handle the low cost and they eventually folded, and then you can just raise the price of oil back up. So I think that might be an opportunity here that that China's looking at is like, hey, let's just flood the market with these cheap AI models and let's make it just. so cheap to run AI inference that these frontier labs
24:07
Unidentified speaker (speaker_1): buckle underneath the pressure.
24:09
Unidentified speaker (speaker_1): So I think that there's a couple of incentives out there. There might be more. And so, should we trust the Chinese models? I think, I don't think we should just trust anything. I think we need to be verifying it. That's probably one of the mantras of this show and the people who listen to this show, right? Is that we should be constantly running back. benchmarks against these models. We should be looking at what they're
24:29
Unidentified speaker (speaker_1): willing to talk about, what they're not willing to talk about, as well as looking for any kind of trip mine that might be inside these that could potentially be sending data back to a place we don't want it to go. Um and uh we should always be just kind of scanning these things and making sure that that they're not doing
24:46
Unidentified speaker (UNKNOWN): that.
24:48
Rob Wallace: Nice. Yeah, you have anything to add, Liam?
24:51
Unidentified speaker (speaker_2): No, I was just gonna say I I 100% agree with all that, and there's a lot of potential for malware that's included in the um models too that can infect your devices. Um so always like think about that and make sure that you go to the a reputable firm if you're going to download these and run them yourselves, um, rather than just going
25:11
Unidentified speaker (speaker_2): to like a Quinn or or something like that. Although there will probably be somebody who can uh verify that um but i also think that in terms of incentives too i think that there's a few others as well like both on the open source side and like who really wins if a lot of these models are available to the public and and can be commoditized. Like when looking at
25:30
Unidentified speaker (speaker_2): China, I think that's a lot of the robotics firms. Right. Like if you have those smaller models that can be input into robots and those are the most effective, like American firms will likely do that too. And maybe they'll post train on them a little bit, but they'll be significantly similar to those models. And China, I think,
25:50
Unidentified speaker (speaker_2): is an industrial and many manufacturing-based economy. And so I think that they do have some incentive to do that. Just and uh I think that the government is probably t a little bit top-down top down in that aspect. And then I do also think like who who wins in the US from open weight models and and open source is mostly NVIDIA. And they've been uh making a bigger push in
26:10
Unidentified speaker (speaker_2): terms of publishing their own open weight models and um, you know, acquiring a few teams that have have been creating them too. Just because the more open weight models the US has that are standardized on their infrastructure, et cetera, I think um they they'll obviously sell more chips. And so I would imagine that while most firms are kind of counting NVIDIA
26:30
Unidentified speaker (speaker_2): out at this point, and there's a lot of stuff going on with Jensen, but uh I do think that they will probably continue to make a push and uh and catch up towards the frontier if I had to make one prediction for this time next year.
26:44
Unidentified speaker (UNKNOWN): Well,
26:45
Rob Wallace: you brought it up, so now we got to talk about it. Uh Jensen, the man who's at the center of this whole entire story.
26:51
Rob Wallace: is uh finds himself in some hot water. Uh, he admitted to an affair with a woman, which the New York Times claims is a spy for China and helping to get the NVIDIA chips sold to our largest competitor in the AI race. Mark, tell us, uh, how important is Jensen himself, a singular man, to
27:11
Rob Wallace: the whole entire AI build build-out, out? And what do you think this means for? AI moving forward?
27:18
Unidentified speaker (speaker_1): Yeah, it's it's really unfortunate that this is going on because Jensen, he's not only super important to the AI build-out, but he has been one of the unique innovators in America for the last few decades. He has he has completely reinvented entire industries on the compute side. And um, he's
27:37
Unidentified speaker (speaker_1): done it rather quietly. Most people didn't know Jensen's name a year ago, right? Two years ago. They had no idea who he was. A lot of people don't know how to pronounce the company NVIDIA, even, right? This is uh this is something that started in the gamer world and then moved into cryptocurrency and then moved into
27:57
Unidentified speaker (speaker_1): AI. And it wasn't really until it moved into AI that people started noticing you know, who this company was. and who this who this man was. So I think it's just really important to recognize that he's on the level of the Steve Jobs and the Elon Musk and these names that come to
28:16
Unidentified speaker (speaker_1): exemplify American ingenuity and innovation. So
28:21
Unidentified speaker (UNKNOWN): I think
28:21
Unidentified speaker (speaker_1): it's I think it's unfortunate that this is going on and and that this might be. the thing that takes him down. I I I hope to see him continue to innovate and find ways to work. I also find it odd and interesting that supposedly the White House and supposedly American intelligence has known about this
28:41
Unidentified speaker (speaker_1): for many months, possibly even dating back to last year. We don't know which facts we can truly believe from reports, but if if the United States States intelligence community knew about this and he was continuing to go meet with the president and continue to make deals, I'm still not sure what to make of all that,
29:01
Unidentified speaker (speaker_1): right? Like it seems like there it wasn't as bad or that America was actually okay with what was going on if they knew about it and allowed it to continue and not just allowed it to continue, but continued continue to meet with him and invite him in. on very sensitive conversations.
29:18
Unidentified speaker (UNKNOWN): It's
29:19
Rob Wallace: crazy. Yeah. And uh Liam, what's your take on the the Jensen situation?
29:23
Unidentified speaker (speaker_2): Yeah. I mean, taking a step back, the entire AI trade has way too many single points of failures for many folks to get comfortable with, whether it's kind of towards the top of the supply chain and in the machines for that ASM ASML makes. for EV lithography, TSMC, NVIDIA, and
29:43
Unidentified speaker (speaker_2): its position within the chip space. Um, there are significant companies that are driving a lot of this. Jensen, completely agree with you. Mark has been uh incredible in the space. Um and I think that
29:57
Unidentified speaker (speaker_2): I've heard a lot of things out there as many as much as like two years ago about how uh the NVIDIA space is there's there's a lot of scandalous things going on there. Uh so I don't think that this uh this event with um I forget the woman's name is is that out of the ordinary. And so I think that that um that potentially is
30:17
Unidentified speaker (speaker_2): a reason why this was perhaps known by the United States government, not really thought of as super con uh super serious until after the midterms. It's definitely a little bit odd that with the shifting political ties with respect to the Democrats coming into power, like this is now front and center. It's been
30:37
Unidentified speaker (speaker_2): reported on Twitter uh as much as like two years ago that this relationship was happening. And um, I do believe that it the the White House probably did know. Um, so I think it's a little bit odd on the timing, but I mean just Jensen started this company and has been there for a very long time. I would imagine that he likely doesn't get pushed out
30:57
Unidentified speaker (speaker_2): over this.
30:58
Rob Wallace: And Mark, in terms of grand macro strategy for America, do you think it makes sense to restrict the chips that we sell to China in order to try to keep our lead in the AI race? Or is that pointless if we if we uh do sell to them that makes them not want to catch up? But if we don't sell to them, then they will try. What do you what's
31:18
Rob Wallace: the strip strategy there?
31:20
Unidentified speaker (UNKNOWN): Well,
31:20
Unidentified speaker (speaker_1): chips are very constrained right now, uh, just from what we're seeing when when we try to get more compute. It's it's incredibly difficult. The the way to get compute right now is is is almost you gotta go pound the pavement and go meet someone in a back alley in order to buy GPUs. Like it's it's uh it's a kind of a cutthroat world out there, right? now. So if
31:40
Unidentified speaker (speaker_1): we have this resource that we think is is really fundamental to the future of the most transformative technology that we've seen in the last couple of decades, maybe our lifetimes, then we should be trying to find ways to get the most out of it for ourselves. Um, and if America wants to win the
32:00
Unidentified speaker (speaker_1): AI race, they need to lock down the supply for themselves. So I do think that it's it's not out of the question for America to say, listen, these chips need to stay inside the borders so that we can use them more than other people can, especially if it's a company, uh country that that is a formidable size and could compete against you head
32:20
Unidentified speaker (speaker_1): to head. Um, then yeah, I think I think it's totally within within the power of the United States to say we want to keep them here. And I think it would benefit our our country greatly to lead in that AI race. Now I'm all for free markets and letting people buy what they want. Um, so it's it's a bit at odds with that in a sense, but
32:40
Unidentified speaker (speaker_1): uh America really needs to figure out like where do they want to play in this race? And these chips are are not prevalent, like they they're they're very constrained on supply.
32:52
Unidentified speaker (UNKNOWN): And
32:52
Rob Wallace: Liam, keeping it on NVIDIA, uh in your finance background, NVIDIA has like five customers that make up 75% of their sales. That concentration risk. Uh how bad is it? And should people be scared of that fact?
33:09
Unidentified speaker (speaker_2): I mean, the
33:11
Unidentified speaker (UNKNOWN): the
33:11
Unidentified speaker (speaker_2): amount of circular financing that has happened is um definitely above my pay grade and and honestly a little bit too complex for me to get into really deeply just given the large customers and um and just like the amount of just the different deals that are going on out there. I think that there,
33:31
Unidentified speaker (speaker_2): if I'm if I'm taking the bear case, which uh I'll do for this example, there's a lot of similarities that I'm seeing right now. between what Bitcoin miners did back in uh 2021 and uh led to the unraveling in 2022.
33:49
Unidentified speaker (UNKNOWN): Um
33:54
Unidentified speaker (speaker_2): the the difference is Bitcoin I mean mining definitely goes through a period where there's there's volatile upswings and then down uh bear markets as well. as just an increasingly competitive market that erodes margins over time. Whereas on the chip side, there is um all their customers are getting significantly more profitable over time. Like Anthropic,
34:14
Unidentified speaker (speaker_2): OpenAI are able to drive down their own costs of completing certain tasks and and um their cost per token significantly on a year over year basis. Um so honestly, I can't necessarily uh give a perfect answer, but I think it's definitely a risk that um
34:32
Unidentified speaker (UNKNOWN): that
34:32
Unidentified speaker (speaker_2): I can't quantify and and you know could topple the the market based on where we are today.
34:38
Unidentified speaker (UNKNOWN): And
34:39
Rob Wallace: Mark, what's your feel for the circular financing of it all?
34:42
Unidentified speaker (speaker_1): Yeah, no, it's
34:45
Unidentified speaker (speaker_1): it's uh it's just that. It's a circular reference and it seems like those things never go well.
34:51
Unidentified speaker (speaker_1): Uh and I and I do wonder
34:54
Unidentified speaker (speaker_1): you look at things that have happened in the past with other legislation that has been pushed through, things like the Patriot Act.
35:01
Unidentified speaker (UNKNOWN): Um,
35:01
Unidentified speaker (speaker_1): and you you wonder, is there some kind of catastrophe that is going to happen, whether it's manufactured or not, with this circular trade that um will be used as a way to push through some kind. of
35:17
Unidentified speaker (speaker_1): legislation
35:19
Unidentified speaker (speaker_1): that's
35:38
Unidentified speaker (UNKNOWN): not a
41:29
Unidentified speaker (UNKNOWN): Um if you
41:31
Unidentified speaker (speaker_3): lost
41:32
Unidentified speaker (UNKNOWN): the
41:41
Unidentified speaker (UNKNOWN): And people want to
41:45
Unidentified speaker (speaker_1): vertical business. Um
41:46
Unidentified speaker (speaker_1): the anthropic one would want to come after. Um one of the problems that you have there is when you're building a company, um, let's say let's let's take AI out of the equation for a second. If you are running a business, you have strategic conversations with your co-founders, with your executive team. with your, you know, with your engineering team, and you come up with how how
42:06
Unidentified speaker (speaker_1): can we beat the competition? How can we come out with something better? And you do it and you execute. But that secret sauce still stays within the walls of your organization. Now you introduce AI back into it. And if you're having all those conversations and then running them through Cloud Claude or running them through ChatGPT, sometimes even the actual conversations themselves, right? People are recording
42:27
Unidentified speaker (speaker_1): their conversations and sending them through. But let's just say, like even just typing up reports and things, that is now being put into those labs. And then they train the next model on all of that alpha that you just gave it. So now your competitor comes along and it starts using Claude. And Claude's not going to say, oh, your competitor did this and this and
42:47
Unidentified speaker (speaker_1): this. You should do
42:47
Unidentified speaker (UNKNOWN): it
42:47
Unidentified speaker (speaker_1): too. It's it's not going gonna to be that brazen. But what it is is you have now trained AI to be better at your specific niche. And so now the next company comes along and says, I want to compete in this niche. And it's like, well, actually, I know all about this. I'm super smart at it. And it will help them compete directly against you. And that's what's
43:07
Unidentified speaker (speaker_1): happening with these frontier labs as they take up all this data. And they even have the toggles you can turn on, like zero data retention. You have to pay money for some of that. But that is really, it's it's not privacy and it's not security. It's a business promise. And both OpenAI and Anthropic have turned off zero data retention at different times because of different things. For
43:27
Unidentified speaker (speaker_1): example, when Fable 5 came out, Anthropic said this model is too dangerous for the general public. We need to do safety scans. And so they started doing 30-day 30 day retention on data, even if you had ZDR. And uh I believe OpenAI has announced similar things too. So so it's it's uh it's really something where you need to you need to secure
43:47
Unidentified speaker (speaker_1): your data with real things like encryption. You know, use cryptography to secure your data. Don't just go off of business promises that can be turned turned off at a whim.
43:59
Unidentified speaker (UNKNOWN): Yeah,
43:59
Rob Wallace: and Liam, are you concerned about any business secrets leaking out? through your use of AI?
44:04
Unidentified speaker (speaker_2): Yeah, for sure. I mean, like, um, I think that David Friedberg gave a good example of talking to Claude, I think it was, about esoteric poop uh potato seeds and like engineering and c and chemistry related to that. Obviously, there's not a ton of that that uh is on the internet it probably. And and uh he was talking
44:24
Unidentified speaker (speaker_2): to one model and he was talking to them about. like whatever chemistry it was and it didn't necessarily know the answer but it went back and forth with it and was able to get it the next model came out and then it was kind of he was asking something similar and it kind of prompted him the whole answer all the way there. And so I think that whether you're a potato farmer uh like
44:44
Unidentified speaker (speaker_2): is something as uh crazy as that example like that can realistically have your competition be able to go out and get almost your trade secrets by even asking it those questions. So I do think that it is it's definitely a concern for um anybody who uses these frontier models.
45:02
Rob Wallace: And Mark, explain to us Maplework and the benefits of using it.
45:08
Unidentified speaker (speaker_1): Sure, yeah. Maplework, Work. that's the new app we have coming out this month. It's in beta right now, people can get it, but uh it'll it'll be a public release very soon. It is very similar to something like Cloud Code or Codex or Grok build that you would run on your laptop. You install the app and you give it access to your files. And so you can say, I want you to work in this folder, or here's multiple
45:28
Unidentified speaker (speaker_1): folders, and it can now start generating files for you. It can make presentations, spreadsheets, it can read files. But then it gets even more powerful when you connect it into your email, to your calendar. You start pulling in Notion, you pull in, you connect it to Vercel so it can publish websites for you. You can you can really put it at the center of all these. But
45:47
Unidentified speaker (speaker_1): then the important piece is that the AI that it's using is encrypted end to end. And so it's running in this confidential compute with these secure enclaves Waves that we run. And so now you are no longer feeding all your data into a frontier lab or some. to a Chinese lab. Instead, these models are running in a box. It's got
46:07
Unidentified speaker (speaker_1): four walls around it. And when you're running it, it's basically like taking you and taking AI, putting you in a room, closing the door, and saying, okay, talk about all the things you want to talk about. And when you're done, you walk out with the data and it stays on your laptop. It doesn't, it doesn't go to some other company. It's not being passed back to be trained into a model.
46:26
Unidentified speaker (speaker_1): Now I will say though, one of the I think one of the more powerful things that we do with Maplework is we understand that this is a multi-agentic world. This is a world where people want to use different AIs and they might have a subscription to Claude. Um and maybe you really like the way that Opus 5.5 writes, uh, you know, writes a press release or writes an article. Um in when operating
46:46
Unidentified speaker (speaker_1): a normal business, if if you have a strategy meeting, if in a room, but then you say, we need to go engage a PR firm to help us write some press release and market this thing, you leave that room and you give away, you selectively give away the information you need to that PR firm. And so we've actually built a similar process into
47:06
Unidentified speaker (speaker_1): Maple Work where you can stay private inside this Enclave, you can do all the work you want to, but then you say, you know what, I want, I want Claude to write this this thing for me. And so Maple Work can actually spin up a subagent inside of Claude and it can keep things private within Maplework and then say, This is the small bit of information Claude needs to write a really banger of a
47:26
Unidentified speaker (speaker_1): press release for us and get it out there. And then it can bring that information back into Maple. And so we're trying to we're trying to be realistic and make make tools that people can use to get their work done, but then have the security model.
47:39
Unidentified speaker (speaker_1): that they feel is appropriate for them at that time.
47:43
Rob Wallace: That's awesome. Because I've heard you say this before. But yeah, like when you're working with AI, you self censor, right? Like you're, you don't want to give everything away. You hold some things back. And so to have that private environment would be such a benefit. Cool. Well, in the last portion of the show, I want to bring it back to Bitcoin. This is Bitcoin news and why Bitcoin is going to win at the end of
48:02
Unidentified speaker (UNKNOWN): the day.
48:03
Rob Wallace: at this AI trade. Uh, and if you're interested in Bitcoin, you got to check out our friends over at Rhino. Because before, getting access to Bitcoin financial services was slow, clunky, and scattered across all these different platforms, each with their own limitations. But Rhino is changing that. Rhino is a next generation Bitcoin banking app designed to handle all of your Bitcoin needs in one place. Buy
48:23
Rob Wallace: Bitcoin with some of the lowest fees anywhere. That means more sats for every dollar. you spend. Make payments over the Lightning Network around the world instantaneously. Send cash over Bitcoin Rails to friends and family anywhere on planet Earth and set up automatic saving plans, DCA, so you stack SATS on your schedule and build wealth over time. They also just signed up with Arch so you can get
48:43
Rob Wallace: loans through Rhino and Arch right on your Rhino app. Rhino protects you with the highest level of security, biometric verification. 24-7 fraud protection, and the confidence of knowing your Bitcoin is never rehypothecated. Go today to rhinobitcoin.com, download the app for iOS or Android and use code Bitcoin News.
49:03
Rob Wallace: One word, Bitcoin News, and Rhino will drop you $21 of free Bitcoin into your account when you deposit your first 100. Rhino is built to keep you in control of your Bitcoin every minute of every day. Go check them out. That's RhinoBitcoin.com. Promo code Bitcoin News.
49:21
Rob Wallace: All right. So everybody's talking about AI, AI, AI. But Bitcoin's here chugging along, putting out a new block every 10 minutes here to save us from the endless fiat that will obviously be a part of this whole entire AI trade, whether it blows up or people need UBI. just to survive, no matter what situation we go
49:41
Rob Wallace: down, uh Bitcoin I think is going to play a pivotal part. So, Liam, explain to us how does digital scarcity work in a world of just digital abundance that AI can provide for
49:55
Unidentified speaker (UNKNOWN): us?
49:56
Unidentified speaker (speaker_2): Yeah, that's uh so essentially uh how I see it is tails Bitcoin wins. uh like heads fiat loses. And meaning by that, like if everything goes right and uh AI is um as amazing as we think it is and and it even accelerates even further, I think that there
50:16
Unidentified speaker (speaker_2): are going to be negative implications in terms of the economy from falling prices, right? Like we're in a heavily indebted society debt to GDP is very, very significant. And the average consumer, they have significant um debt on their homes, et cetera. And so if they
50:36
Unidentified speaker (speaker_2): if the ability to accomplish certain tasks declines by uh you know 300% again and again and again, which it's been doing over the past two plus years, I think that we're going to see just um falling prices in the form of wages. And there's going to be a lot of dislocations in the economy,
50:56
Unidentified speaker (speaker_2): especially talked about it a little bit before, but there will be
51:01
Unidentified speaker (speaker_2): deterministic tasks where there's a right answer. Those may get uh replaced. Now there will obviously be other ones that can come along and And replace it and be good for c uh those net consumers as a whole, but there's going to be a messy middle path. And so w in a world where there's falling prices, there needs to be
51:21
Unidentified speaker (speaker_2): some sort of money that can actually retain your value over time. So as the economy gets more and more efficient, Bitcoin is is completely scarce. And so it should allow you to uh buy more of anything over time, just given the deflation that will occur. Well uh in US dollars, I think that that
51:41
Unidentified speaker (speaker_2): will primarily be inflated away, both uh just given the inflation rates today, and then additional stimulus just given the um the amount of bailouts that will need to happen in order to make consumers whole on their homes, etc.
51:54
Rob Wallace: Very well said. Yeah Mark, how do you see AI playing out in this Bitcoin playing out in this AI world?
52:00
Unidentified speaker (UNKNOWN): Yeah,
52:01
Unidentified speaker (speaker_1): it's it's all uh uh so I think Liam, you just touched on a lot of a lot of aspects of kind of debasement and how Bitcoin really fits into that. Um, you know, I we we do what we can just to save in a currency that uh you know that doesn't get debased by a small room of people at will.
52:20
Unidentified speaker (UNKNOWN): Um
52:20
Unidentified speaker (speaker_1): so there's definitely that aspect of it. And then I think the other the other things to look at are the technology itself. Um it's not just the finite supply, but it's the ability to be a native internet money. And so now we have these native internet citizens because we have AI agents. My view is that AI is just a tool, right? But we wrap
52:40
Unidentified speaker (speaker_1): these personalities around these AI agents to make them easier to interact with. It's just an interface for humans to be able to interact with this tool. And so you have these AI personalities that are now getting. their own accounts online and they're starting to transact online. And yes, they could use credit cards, but uh the credit card is useful if you want to have your AI buy something for
53:00
Unidentified speaker (speaker_1): you on Amazon. But when your AI needs to actually pay for compute and it's paying for fractions of a penny of compute or a few cents here, a few cents there, it needs the ability to do microtransactions. And what is better built for microtransactions than a cryptocurrency that knows. how to how to speak internet protocol. And so I think that um I think that's a great place where
53:20
Unidentified speaker (speaker_1): Bitcoin can benefit because if the AI is able to study economics and can see that currency is being debased, and it will see which currency is better for it to hold on to, then it will likely make some of the same trades that that humans have started to make as far as okay, here's this money you gave me. I'm going to s I'm going to use it to to
53:40
Unidentified speaker (speaker_1): pay for compute. And when I get extra money that I've made, I'm going to save it in this other currency. And um, and then if it if it needs if it's if it's told, no, I want you to use US dollars, but save in Bitcoin, then it's gonna see the slippage that comes when it tries to convert from convert from Bitcoin to fiat to pay for things on back. Um, why not just stay in the
54:00
Unidentified speaker (speaker_1): native currency? There's no need to actually convert and and pay on these old rails. Just stay within the currency that you have. So I think there's an opportunity here for Bitcoin really to
54:10
Unidentified speaker (speaker_1): move into some of these other realms of like medium of exchange and maybe eventually unit of account kind of stuff because of the AI technology and how it's playing
54:21
Unidentified speaker (UNKNOWN): out.
54:22
Rob Wallace: Yeah, it does seem as if Bitcoin has this a huge opportunity. when it comes to agentic payments. And I mean, BPI put out a study saying that AI agents prefer Bitcoin over any other payment method. Uh, but right now a lot of it is being set up for stable coins. So, Liam, do you think that Bitcoin is going to
54:41
Rob Wallace: outcompete stable coins when it comes to agentic payments?
54:46
Unidentified speaker (speaker_2): I mean, I think that um just the entire economy as a whole is probably. going to be increasingly moving towards a dual currency role for at least a period of time. I think that we've seen that the US dollar and uh Bitcoin are both superior to international currencies, just given the amount of
55:05
Unidentified speaker (speaker_2): liquidity and uh marginally less inflation over time than than most other currencies out there. And so I think that as the dollar gets digitized and the access of opens up to more and more folks, it's just going to create a lot of the the second and third world currencies to decline in terms of dollar terms. But I do think
55:25
Unidentified speaker (speaker_2): so I think that that's a factor too. And then I think that as the um as AI continues to grow and get implanted not just in uh agents on your computer, but into pretty much every part of life that we interact with. Um, you know, maybe, maybe not quite on your refrigerators, but in terms of all the robotics that are
55:45
Unidentified speaker (speaker_2): going to come with respect to the uh humanoid robotics as well as incorporated into uh manufacturing and things like that that are non-humanoid shaped. A lot of them will need to have AI right into their um
56:00
Unidentified speaker (speaker_2): like built into their system, and they're going to need a uh ways to interact and barter with each other as they increasingly take humans out of the loop. I doubt that they're going to increasingly use stable coins. Um I think that that will probably be something that happens for humans, just given the amount of tax complexity
56:20
Unidentified speaker (speaker_2): that unfortunately still exists in the world out there, or very small payments if you're only using it for like a day or two. Um corporates don't necessarily want to deal with a lot of the tax headache right now, unfortunately. But for people that don't necessarily have to pay taxes, um, yeah, I think that it's just a superior form of money too.
56:39
Unidentified speaker (UNKNOWN): All
56:40
Rob Wallace: right. So now we're getting towards the end of the show. Let's look forward. Uh Mark, 2026 was definitely the year of the agent, right? When it comes to AI, open OpenClaw clause and now Grokbot and Maple and Hermes, everything is agentic. Uh, what is going to be 2027? 2027 will
56:59
Rob Wallace: be the year of blank when it comes to AI.
57:03
Unidentified speaker (speaker_1): Yeah.
57:05
Unidentified speaker (UNKNOWN): Uh
57:05
Unidentified speaker (speaker_1): it's it's funny because yes, 2026 was the year of the agent, but how many people do you know in like regular life, if you walk around, are actually using an agent? Uh I think it's very small. So in the news cycle and in the technology forefront of AI, is it's been the year of the agent. So uh yeah, I'm I'm happy to kind of to talk there. I still think there's a lot
57:25
Unidentified speaker (speaker_1): to do within getting this technology in the hands of people to continue to further AI chat and AI agents. Um, I think that the personal agent is is is a is a step beyond the normal AI agent. Um, we're just seeing it now. with GrockBot and Muse. We had our own version that we didn't ever launch and we might like bring up bring back around now. I think
57:45
Unidentified speaker (speaker_1): 2027 will be this year where we start to see what are the points in our everyday flow and in our um our our commerce that we have. What are the points that can be solved with AI from an interaction standpoint? So I put out a tweet yesterday that was just kind of like a silly tweet of, you know, the meme of the guy
58:05
Unidentified speaker (speaker_1): ripping his headphones off. And it's like you're on your third hour of your call center shift and you're sick of talking to grok bots that are calling in. Right. And uh it's that I think I think 2027 is going to be the year that we start to find out what are the places that can have interactions agent to agent. And so it's not just going to be on like one side
58:25
Unidentified speaker (speaker_1): of me using my agent to go out and talk to the world. It's going to be what are the points on the other side of that line are also agentic. And so I think maybe maybe agentic commerce is maybe the the the phrase for 2027. Um but it's going to be like completing the connection now of rather just one maverick
58:45
Unidentified speaker (speaker_1): on their computer with an agent trying to figure out how to go out in the wild, wild west. It's going to gonna be that, okay, now there's actually actual infrastructure there that we're putting into place so that these agents are not just talking to humans who don't want to talk to them. They're they're moving at a faster speed because they're talking to an agent on the other side.
59:01
Rob Wallace: Nice. There you have it. And uh Liam, in conclusion, 2027 for Bitcoin. Bullish, bearish. What's going on?
59:09
Unidentified speaker (speaker_2): Obviously bullish. Uh everybody's gone. Um, miners, you know, Mero. who levered up at the top, uh, sold a lot of their Bitcoin. The the Bitcoin miners have a lot of like mostly exited, don't get really texts uh from friends and family anymore about uh you know what's going on with micro strategy or Bitcoin. It's
59:29
Unidentified speaker (speaker_2): uh it so you know, a lot of people like it just feels like peak apathy out there. Um and you know, every time it's it's not even fear, it's apathy. I always think that's the the best time. So uh I I don't know if I've ever been more bullish right here. Um and and we could use it after these past five years. Yeah.
59:48
Rob Wallace: Yeah, I think we could. The community could definitely use a nice strong bull run going into twenty seven, twenty eight, and beyond. Well, gentlemen, that was a fascinating conversation. Thank you so much for your time. Uh just tell people where to go. Mark, where should people see uh Maple?
1:00:02
Unidentified speaker (speaker_1): Yeah, go go check out Maple. It's maple.ai. You can't forget the domain name. And we accept Bitcoin for payment. Um, and so we actually have not only regular AI accounts, but we have a fully anonymous account that can only be paid for with Bitcoin. So you get a unique ID that's auto-generated. You set a password, you pay in Bitcoin, and you can just you can
1:00:22
Unidentified speaker (speaker_1): just cook on your own.
1:00:24
Rob Wallace: Nice. And Liam, where should people learn more about early ride? riders and what you're doing?
1:00:28
Unidentified speaker (speaker_2): Check out earlyriders.com. You can send me an email at Liam at Earlyriders.com. Uh if you we put out a lot of research around Bitcoin, venture, and uh yeah, if you're building it kind of the intersection of Bitcoin and AI, uh, please reach out to me at Liam at Earlyriders.com.
1:00:44
Unidentified speaker (UNKNOWN): All
1:00:44
Rob Wallace: right, guys. Thank you so much. Everyone at home watching along, thank you for tuning in. This has been Rob Wallace, host of Big Bitcoin News Weekly. And we'll be back next Wednesday, eleven AM Eastern with some more great guests. Until then.
This page was produced automatically: the audio was transcribed with nvidia/parakeet-tdt-0.6b-v3+sortformer-v2.1 and interpreted by Qwen3.6-35B-A3B-UD-IQ3_S.gguf. Transcription and interpretation can contain errors; always confirm against the original episode. Nothing here is investment advice.

More from Rob Wallace | Bitcoin News

Rob Wallace | Bitcoin News

The NVIDIA Scandal Goes Deeper Than You Think | Prof St Onge

Episode discussion: Prof St Onge discusses AI monopolies, Nvidia chip smuggling, and Bitcoin as a monetary hedge

Speakers’ claims: Professor St Onge argues that open-source AI is essential to prevent a closed oligarchy controlled by companies like Google and Anthropic, citing Jensen Huang's advocacy for open models. He details allegations that Nvidia channeled $7.5 billion in chips to China through Singapore, warning that sanctions may inadvertently accelerate Chinese innovation. St Onge contends that AI is a leveling technology that commoditizes expertise, benefiting hardware manufacturers with strong moats like ASML and Micron due to IP restrictions. He advises young adults to focus on skilled trades and human-centric services, while positioning Bitcoin as a hedge against monetary debasement and systemic collapse compared to gold.

Artificial IntelligenceSemiconductorsBitcoin
1h 12m3 assets17 claims to verify
Rob Wallace | Bitcoin News

AI Will Force Trump and The Dems To Print Trillions | Marty Bent and Matt Dines

Episode discussion: Speakers argue AI infrastructure debt will force US Treasury to print money, boosting Bitcoin

Speakers’ claims: Speakers Marty Bent and Matt Dines argue that the AI infrastructure build-out has created systemic risk through negative cash flows among major tech companies, necessitating future bailouts or monetary expansion. They claim the Trump administration is shifting power from the Federal Reserve to the US Treasury, moving away from CBDCs toward a Treasury-centric model that aggregates government-held Bitcoin. The speakers contend that physical constraints like power and financial constraints like credit spreads are peaking, forcing hyperscalers to take on massive debt. This debt creation is described as the primary mechanism for future monetary inflation, which they believe will drive demand for Bitcoin as a hedge against political and economic instability.

BitcoinAI InfrastructureUS Treasury
1h 6m3 assets19 claims to verify
All insights