Brian Armstrong | No Priors Summary

1. Key Themes

The "Everything Exchange" Is Already Mostly Non-Bitcoin

Coinbase has quietly transformed from a crypto trading venue into a diversified financial exchange, with the vast majority of revenue now coming from beyond its original asset class.

"Actually 88% of our revenue is from non Bitcoin trading at this point. That's an interesting stat. I think it's funny because if you look at our stock price, it does trade like fairly correlated with the price of Bitcoin." 00:18:36

Banking the Agents: A New Financial Customer Class Is Emerging

Armstrong frames AI agents as an entirely new category of financial customer that needs its own accounts, wallets, and credit infrastructure, distinct from serving humans.

"We don't want the AIs to be unbanked. You know, if we want to bank the AIs, they deserve financial services as well. The AI agents themselves are going to have to have their own accounts." 00:00:00 "We're also going to provide self-custodial wallets to any AI agent out there that wants to hold a balance in stablecoin... It could actually raise money. It could issue a token." 00:04:51

Micropayments Are the Actual Agentic Economy — and Legacy Rails Can't Serve Them

The economics of existing card networks structurally exclude the transaction sizes AI agents actually need, which is why crypto rails become "essential" rather than optional for agentic commerce.

"About like 76% of the agentic commerce transactions we're seeing are under 30 cents." 00:00:00 "It doesn't make sense to pay 30 cents to send a one cent payment... a lot of it is gathering information... it's venture firms doing research compiling from a paywall data, or it's recruiters who need to scrape data from LinkedIn." 00:00:00

Tokenization Is the Real Long-Term Trend, Not Just Stablecoins

Stablecoins were the proof of concept; Armstrong sees essentially the entire financial system — stocks, credit, treasuries, deposits — being rebuilt on tokenized rails.

"This is like broadly a trend called tokenization... stablecoins were the first major use case of this with a token that represents one-to-one a dollar... now we're seeing it with stocks... you just go on down the list, like private credit and treasuries and bank deposits. Like everything's going to get tokenized." 00:19:05 "There's like something like 4 billion people in the world who don't have access to any brokerage or US investment account." 00:19:34

Recursive Self-Improvement as an Internal AI Architecture, Not Just a Buzzword

Coinbase has operationalized "recursive self-improvement" as a literal internal system — a living, editable knowledge base ("brain") per team/service that agents consult and update, compounding accuracy over time.

"You have a brain for the company... a brain for each team... a brain for each individual, which by the way, I believe they should be able to take with them when they leave Coinbase." 00:12:13 "When you go to make that change and edit the agent's thinking, that context has to go back into the brain so that you not only fix it in this case, but in all future cases going forward... the accept rate for one-shotted PRs... just starts to tick up over time." 00:13:35

Productivity Gains Won't Shrink Companies — They'll Accelerate Output Per Person

Armstrong pushes back on the "10-person unicorn" narrative, distinguishing task automation from headcount reduction, while acknowledging small teams can now do outsized things.

"Productivity is definitely increasing... that's slightly different though than saying the average company size will be smaller... tasks are being eliminated. I think tasks are being eliminated. People are not being eliminated." 00:16:57 "You're going to see some companies that are two people or five people or whatever, and they're able to do a lot more than what we would have historically thought was possible. And that's also true." 00:17:26

Prediction Markets as Societal Infrastructure, Not Just Sports Betting

Armstrong sees prediction markets extending far beyond entertainment into policy-making tools and even open-ended "opinion markets" on unresolvable questions.

"Within months of launching it, it had to hit like a hundred million dollar revenue run rate. It's growing really nice, like over a hundred percent quarter over quarter." 00:21:54 "You could actually say, all right, if we were to implement this policy, what would the unemployment rate be within one year, two years, three years? And then a prediction market could form on that. It could help inform what policies out there." 00:22:23

Founders Running Multiple "Act Ones" Simultaneously

Armstrong frames his New Limit involvement as part of a broader pattern of successful founders becoming multi-company builders/capital allocators, naming specific peers doing the same.

"There's a bit... I think there's a blurry line between like, I'm the CEO of one company and like, I'm a pure investor... in some sense, you're both capital allocators at a certain size of company." 00:30:28 "I probably tried like 10 different ideas before Coinbase started to really work, right? And then you want to make sure you don't try to go do multiple things too early... the distraction can be really dangerous." 00:29:59

Longevity Science Via Epigenetic Reprogramming Is Now an AI-First Drug Discovery Problem

New Limit's approach treats aging as a single meta-problem (restoring youthful cell function) rather than disease-by-disease, using AI to navigate a massive combinatorial search space of transcription factors.

"It starts with AI, as with all things these days. You can test millions of hypotheses about what sets of transcription factors... could reprogram different cell types. It's a massive search space... We've built the leading frontier model for epigenetic reprogramming." 00:28:05 "We've demonstrated successful reprogramming of at least one human cell type in these humanized mouse models... the first phase one clinical trial will be launching next year." 00:29:03

Special Economic Zones / "Freedom Cities" as an Underrated Policy Lever

Armstrong sees deregulated geographic sandboxes — not broad national policy change — as the more tractable path to unlocking innovation in nuclear, drones, drug discovery, and crypto.

"You could take a piece of this federal land, 10 or 100 square miles and say... in this area, you don't need to go through all the normal red tape on EPA... you can build data centers there... nuclear power plants... unlimited drone delivery." 00:43:06 "The problem right now is, in many cases, there is no sandbox to go try this innovation. And that is just incredibly stifling." 00:44:03

2. Contrarian Perspectives

Money Doesn't Disappear Even in a Post-Scarcity AI World

Armstrong directly disagrees with Elon Musk's view that money becomes obsolete, arguing scarcity persists structurally even as AI/robotics collapse the cost of goods.

"I've heard people like Elon say that they think there won't be money in the future... even if the cost of goods and services really falls... there's always still going to be scarcity of certain things like land... energy, like chips. Dyson spheres are going to be expensive, I think, pretty far into the future. So I think we're still going to need a medium of exchange." 00:09:30

Gene/Embryo Editing Will Increase, Not Decrease, Human Diversity

Against the intuitive fear that genetic "packages" will homogenize humanity (a concern Lon raises), Armstrong argues the opposite — diversity of parental preference plus "raising the floor" rather than optimizing a single trait will expand variation.

"I actually think it's the opposite. I actually think it could increase diversity. The reason is that there's just unlimited diversity of human preferences... in the limit case, you could actually get diversity that's really out there. Like you could have kids born with gills." 00:40:08

Embryo Editing Will Flip from Controversial to Morally Obligatory

Rather than treating embryo screening/editing as a fraught ethical frontier, Armstrong predicts near-total normalization within a generation, comparing avoidance of it to reckless behavior.

"80% of Americans, embryo editing for disease prevention is supported by the population... I think it'll be more normal over time. And at some point it'll be abnormal to have children without doing this kind of screening and editing... it's like driving without a seatbelt." 00:38:43

Most Longevity Science to Date Has Been "Snake Oil"

Armstrong is unusually blunt in dismissing the existing longevity field's rigor before founding New Limit, rather than positioning it as merely underfunded.

"I didn't see like that many great teams working on longevity. I thought a lot of it was like kind of snake oil... all the science is like a little bit bogus... people have tried these things for a long time, like caloric restriction or whatever. We found a bunch of ways to make mice live longer and it doesn't work in humans." 00:26:10

Regulatory Appetite for Alternative Jurisdictions Has Cooled Precisely Because US Deregulation Is Working

A non-obvious claim: interest in seasteading/charter cities-style solutions declined because domestic US policy improved, undercutting the usual narrative that such projects gain momentum over time.

"There was actually... I'd say there was a little bit more interest in this before the most recent Trump admin, just because the deregulatory environment has created a lot of opportunities here in the US. And that's a nice, that's a great substrate and market to build on." 00:42:38

3. Companies Identified

Coinbase — Publicly traded crypto exchange evolving into an "everything exchange" for stocks, commodities, derivatives, prediction markets, and agentic finance. Mentioned throughout as the core subject; 88% of revenue now non-Bitcoin, tokenized stock product just launched outside the US, prediction markets hit "$100 million revenue run rate" within months.

"It's now possible to trade everything in one place with good liquidity and good cross margin." 00:02:12

New Limit — Longevity/anti-aging biotech co-founded by Armstrong, focused on epigenetic reprogramming to restore youthful cell function; lab of 50-60 people in South San Francisco; first phase one trial (alcoholic liver disease) launching next year.

"We've built the leading frontier model for epigenetic reprogramming... first phase one clinical trial will be launching next year." 00:28:33

Tether — Referenced as the pioneering stablecoin success story that proved global demand for dollar-denominated digital accounts, used as the analogy for why tokenized stocks will also succeed.

"Much like Tether was very successful ex-US because a lot of people wanted dollar denominated accounts. There's like 4 billion people who want access to high quality investments." 00:19:34

Prospera (Honduras) — A special economic zone/charter city project that Coinbase invested in to learn about the model; described as having had a rocky relationship with the local government but nearing resolution.

"They've had a mixed up and down, I'd say, with the local government there. But they're getting closer to a resolution, is my understanding... I think they're going to do a good job." 00:42:09

Preventive — Embryo editing/screening company Armstrong has personally invested in, focused on disease prevention.

"There's a company I invested in called Preventive that's doing embryo editing, which is a controversial topic." 00:38:16

Harbor — Real estate tokenization startup Lon invested in roughly seven-to-eight years prior to the recording, cited as a case of being early but too early to the tokenization trend.

"I remember actually seven, eight years ago, I invested in a company called Harbor that was trying to tokenize real estate. I think that was a little bit too far ahead." 00:21:12

Erebor — Palmer Luckey venture referenced as an example of a founder building a "handful of companies" in parallel, similar to Armstrong's own Coinbase/New Limit dynamic.

"Palmer Luckey has now a handful of companies like with Erebor." 00:29:59

4. People Identified

Brian Armstrong — Co-founder/CEO of Coinbase and co-founder of New Limit; described by the host as one of the most forward-thinking people across science, technology and finance, and self-describes as an entrepreneur first, investor second.

"I like being a builder or like an entrepreneur. I'm probably better as an entrepreneur than I am as an investor." 00:24:41

Shinya Yamanaka — Scientist whose lab's early results on cell reprogramming (Yamanaka factors) directly inspired New Limit's founding thesis, showing cells could be reprogrammed to restore youthful function.

"One of them they came back with was this epigenetic reprogramming idea, which some labs had seen early results with like Shinya Yamanaka's lab." 00:27:06

Jacob Kimmel — CEO of New Limit, identified as part of the founding team assembled through Armstrong's dinner series on longevity science.

"It turned out to be, you know, Jacob Kimmel, who's now the CEO of New Limit." 00:27:36

Blake Byers — Co-founder of New Limit.

"Blake Byers, one of the other co-founders." 00:27:36

Greg Johnson — Co-founder of New Limit.

"Greg Johnson, who co-founded the company." 00:27:36

Elon Musk — Referenced twice: once for his (disputed) view that money will disappear in the future, and again as the most famous example of a founder running multiple simultaneous companies.

"I've heard people like Elon say that they think there won't be money in the future, right? And I mean, I actually agree with Elon on most things, but that one I was like, okay, I think I know what he means." 00:09:30

Joe Lonsdale — Cited as an example of a founder who has successfully incubated multiple companies in parallel.

"You know, Joe Lonsdale has done this." 00:29:59

Sam Altman — Cited alongside Lonsdale as a multi-company incubator/builder.

"Sam Altman has done this." 00:30:28

Palmer Luckey — Cited as running multiple companies simultaneously, including Erebor.

"Palmer Luckey has now a handful of companies like with Erebor." 00:29:59

Daniel Lech — Referenced by Lon as another example of a founder who successfully transitioned into incubating multiple companies over time, with Armstrong agreeing.

"He's made that successful transition to incubating more companies over time and figuring out where he can plug in to help some capital, but it's really his time is probably the most valuable part." 00:31:30

5. Operating Insights

Force Context Corrections Back Into the System, Not Just the Output

Armstrong's most tactically valuable point: the discipline isn't using AI agents — it's building a mandatory feedback loop where every human correction to an agent's work updates a persistent shared knowledge base, so mistakes are fixed permanently rather than just once.

"The key step that we're trying to really enforce now in our software factory or software development lifecycle is when you go to make that change and edit the agent's thinking, that context has to go back into the brain so that you not only fix it in this case, but in all future cases going forward." 00:13:35

Structure Agent Brains at Multiple Granularities — and Make Them Portable to People

Rather than one company-wide knowledge base, Coinbase deliberately segments "brains" by team, by service/repository, and by individual — and Armstrong has set policy that an individual's personal AI brain should be portable when they leave the company, which is a notable talent-retention/goodwill lever few companies consider.

"You can have a brain for each team. You can have a brain for each service... you can have a brain for each individual, which by the way, I believe they should be able to take with them when they leave Coinbase. So we're creating some good policies on that." 00:12:13

CEO-Level Parallel Task Delegation to Cheaper, Specialized Models

Armstrong personally demonstrates a workflow of decomposing a feature into phases, then routing individual sub-tasks to the cheapest sufficient model (mixing open-source models and Grok) rather than defaulting to frontier models — a cost/speed optimization playbook operators should copy.

"Go, go execute phase one, go spin up 10 separate agents to go execute each of those pieces in parallel. And by the way, recommend the right agent to use because they might be cheaper models. And so it spun up a bunch of open source. It spun up a bunch of Grok." 00:15:02

Fine-Tuned Small Models Beat Frontier Models on Narrow, Proprietary Tasks

A direct, numbers-backed claim that smaller open-weight models trained on internal proprietary data outperform frontier models for specialized enterprise tasks — a build-vs-buy signal for any company evaluating whether to keep routing everything through GPT/Claude-class APIs.

"If you train it on 100,000 compliance cases within Coinbase proprietary data, a small open weight model can actually outperform a frontier model." 00:08:20

Cap Your Personal Portfolio of "Act Twos" to Avoid Simultaneous Failure Risk

Armstrong's explicit risk-management rule for founders running multiple ventures: keep the number of concurrent high-stakes commitments small enough that a simultaneous crisis in two of them doesn't compound into personal/organizational failure.

"I think having like a small handful of things you're trying to do and you don't want to get overwhelmed where if a couple of these things have a problem at the same time, you're just going to not be sleeping and probably ignoring your family and stuff like that, which is not balanced and healthy." 00:30:28

6. Overlooked Insights

X402 as a Quietly Forming Standards Coalition Around Agent Payments

Buried in the conversation is a significant infrastructure signal: Coinbase incubated the X402 payment protocol and then donated it to the Linux Foundation, and it's now being adopted by Google, Cloudflare, and AWS. This is effectively the early formation of a neutral, cross-industry rail for agent-to-agent micropayments — a potential "TCP/IP moment" for agentic commerce that got only a single passing mention rather than the deep dive it likely deserves.

"We're seeing a lot of agentic payment now flow over these new protocols like X402, which is one that we incubated at Coinbase that we actually put into the Linux Foundation. So Google and Cloudflare and AWS, a bunch of people are working with us on that now." 00:05:45

An Emerging Marketplace of Narrow "Specialist Agents" as a New Company Category

Armstrong floats, almost as an aside, the idea that the agentic economy will spawn a layer of hyper-specialized agent vendors (an agent that's just excellent at ordering pizza, or one trained purely on proprietary Figma design data) that other agents transact with as tool calls. This implies an entirely new class of thin, vertical AI companies whose customers are other AI agents rather than humans — a B2B2Agent business model that wasn't explored further but has significant venture implications.

"You're going to see in this marketplace a bunch of specialist agents. It might be like, hey, this agent is just really good at ordering you pizza if that's what you want. Or this agent is like a really incredible designer... it's been trained on is like tons of proprietary Figma data or something like that." 00:08:20