The Market Neutral 12x People Don’t Talk About
Since 2021, $1 invested in Nvidia and short ARKK continuously rebalanced would be worth about $12 today with little market or tech exposure. ARKK investments failed to capture the entire AI boom despite that being its stated investment goal. The market is currently pricing that AI will be immensely valuable and disruptive, but that the existing corporate framework will not benefit much.
This article details how to capitalize on the implication of this pricing. A new economic primitive called an Agentic Protocol, which is an AI driven cryptocurrency with cash flows that accrue to its holders.
The rise of AI Agents like Devin as well as new open source equivalents that show promise re: directly replacing engineering teams make this inevitable. Google’s recent bid for Hubspot similarly indicates a desire to automate sales reps and online lead generation. Agentic Protocols are a way to think about a company with no sales or engineering team, and eventually - no people. I expect this trend to be recognized over the next 12 months, due to aggressive subsidies from AI Crypto Protocols then hyper accelerate within 2 years and improvement of open source AI models which has entered hyper-acceleration in the past months.
Megacap AI companies have trounced ARKK - fundamentally because - the vast majority of normal companies including in the tech sector do not really benefit from AI.
Even though AI has 12x-ed vs “Corporate AI beneficiaries” - the Cramer/Bloomberg establishment continues to argue and position themselves for the opposite, and have Cathie Wood on their shows. They conveniently gloss over the fact that Cathie Wood’s portfolio just hasn’t benefited from AI during the largest run up in AI stocks in history. But for some reason, they will soon.
How could such an obviously false narrative gain such traction? Incentives.
The AI industry, and even its chat bots are economically aligned to spread this narrative - because the alternative, that AI does not broadly benefit society - will almost certainly incite industry regulation. Everyone had the wrong bet. And yet the “AI beneficiary” proponents continue to spread their gospel, and collect hundreds of millions of fees annually fleecing retail investors.
Though ARKK is easy to pick on, it is representative of an entire school of thought - and backed by Tiger Cub, Bill Hwang (who lost all of his own money instead of investors). Thus its underperformance points to a broader failure to recognize that AI does not actually benefit the “innovation economy” - that instead, it replaces it. It’s an important asset - not just due to its liquidity and publicity, but because it is a benchmark for IPO valuations and the multi-trillion dollar venture capital asset class.
The “innovation” economy they’re selling is all about promising future growth. Future revenue. Billings curves. And so on. Whereas AI companies are delivering in the here and now - which is precisely why the future predictors failed to buy them. Thus - the first crash in 2022 was rates driven - because the discount rate which allowed that future to be imagine increased, and that future became less valuable. But AI stocks have risen in spite of rates - because they are delivering on a 1-3 year time frame, not a 10 year timeframe like the innovation economy. Long Nvidia and Short ARKK realizes a .23 correlation to rates - demonstrating this.
Thinking of duration another way - anyone working on AI applications naturally understands how absurd it is to forecast even 2 years out. Whereas Cathie Wood and other “Innovation economy” or “AI beneficiary” proponents claim insight into the next 5 years even though they made no money over the last 5.
It is somewhat ironic that ARKK is one of the largest advocates for crypto. The huge surge in the cryptocurrency market will hyper accelerate the decline of the innnovation economy by allowing AI agents to access compute and storage resources without censorship by interacting with a whole new set of crypto AI primitives that have surged in value over the past year. The irony comes from the notion that if Bitcoin does actually go to Cathie Wood’s price target of $3.2 million - in that world, Agentic Protocols have eaten the lunch of every single company in her portfolio.
Agentic Protocols are only possible because of the crypto and the ability to access AI primitives directly without a cloud intermediary such as AWS or Microsoft Azure. So far - protocols such as Render have deployed huge compute resources without many users. Similarly, protocols such as IONet are enabling cheap and extremely efficient AI calculation on chain without interacting with big tech. Agentic Protocols will use the capacity.
At their core - Agentic Protocols - are a bet that AIs will do digital and knowledge work tasks better than people. That they can buy the compute to do these tasks cheaply via distributed protocols in a transparent and verifiable manner. They can sell their work in a native currency that is hard to shut down. And can use the profits to improve themselves and progressively disrupt the real economy.
Defining Agentic Protocols
What is An Agentic Protocol
Nvidia surging while the so-called innovation economy collapses is where we are now, and where we have been for the last 12 months. But the hyper up-move of Bitcoin, and Crypto AI as well as the onslaught of new AI agent technologies is bringing us to where the puck is headed over the next 12 months.
Neither Crypto nor AI Agent trends will slow down, as people still don’t understand them. The market narrative remains the hyper-optimist vision that has lost money for 3 years straight - that AI will benefit instead of replacing the innovation economy.
How to capitalize on this?
I’m going to spell out the vision of an Agentic Protocol
Agentic Protocols are key to understanding how exactly the “crypto AI thesis” plays out and how the dollars will flow.
An Agentic Protocol is a rejection of the traditional venture model, as well as the traditional crypto token model. It is - in the words of Monty Python - something completely different. It is an extension of Sam Altman’s projection - that there will be a large number of two person unicorns.
So What is an Agentic Protocol?
An Agentic Protocol is a self-developing AI driven entity that aims to have no human employees. It generates cash flow by licensing IP or other technology products, facilitates network economics or speculates. It has its own native cryptocurrency token which 1] is the currency by which its products are sold or licensed which 2] allows human or AI users to participate in the upside of its financial activities 3] validates and pays for the opex of the system which is primarily compute, training and storage.
A Visualization of What Agentic Protocols Look Like
Because Agentic Protocols are not something most people are familiar with - a quick visualization.
Imagine a set of computer programs, interacting with one another acting as digital engineers. Controlled by a product manager - an open source GPT. This manager tells them what code to write. They iteratively test the code, using compute resources they buy from crypto protocols such as IONet. They store their output using storage protocols such as Filecoin or Arweave. Then sell the outputs to either other protocols or people.
Their products could range from actual software, to designs, to entertainment. Because Agentic Protocols are natively on chain - the rise of asset tokenization per Blackrock’s recent push allows them to speculate with their treasuries, or alternatively hedge out the market risk associated with their consumption of other tokens that they need to generate their goods.
Some Agentic Protocols will be influencer farms, others hedge funds, others on-chain arbitrageurs, others entertainment conglolmerates.
An Agentic Protocol rather than hiring, raising money, or focusing on optics/ IPOs focuses on the utility of its community (as well as community ownership), and development and deployment of models that allows its financial output to be higher.
Structurally Higher Intelligence Margins vs Corporations
You want to scalably harvest the value of AI application and short the prices of model deployment, and compute. I will refer to this as the term “Intelligence Margin” - which is the spread between the cost of compute and the potential revenue of compute.
Where Agentic Protocol differ from the current framework of AI thinking: consider Accenture (which I’ve written about before as a large short position).
Accenture argues that it can make money helping existing corporations adopt AI. That it can apply AI to clients’ workflows at a profit.
Accenture will not be able to compete with AI native applications.
People pointing to the cloud transition analogue are wrong. “The Cloud” couldn’t interact with customers directly. The Cloud did not obsolete human intelligence.
We already talk to AI chatbots constantly, who are perfectly capable of explaining, upselling, making slide decks and recommending organizational level changes.
Would you rather get your advice on AI architecture and adoption from an expensive guy in business casual or an AGI that is literally a physical manifestation of AI architecture?
In the new nomenclature, Accenture’s “Intelligence Margins” are structurally impaired compared to those commanded by Agentic Protocols.
Corporate Framings
Before going further into what these abstractions look like - I want to start with some framings
A traditional venture company is a speculative R&D corporation that aims to build a product people want, grow that product extremely quickly then IPO to land and expand that product
A cryptocurrency network aims to create a new way for value to get transferred that has preferable security and efficiency qualities relative to fiat money, Bitcoin, or Ethereum
A cryptocurrency token sits on top of a network and allows users of a product on a cryptocurrency network to participate in equity-like upside. Traditionally, cryptocurrency tokens inflate or print tokens to fund their operations, much like traditional venture companies issue new shares.
A cash flow business is a non-speculative entity that delivers a low risk product - such as a coffee brand, or consulting services to a small set of customers
At its core - an Agentic Protocol is a mix of a Cash Flow Business and a Cryptocurrency Token. It’s a cryptocurrency that aims to use AI to generate sufficient cash flow such that its token supply is continuously decreasing for the benefit of its users, who also use the protocol in a way that generates utility for the service that it is selling.
This economic edifice rejects the previous wealth model and assumptions of “growth tech” and the “innovation economy”. Proximity to humans, and talent matter less than effective model deployment with efficient compute utilization.
Agentic Protocols are the ultimate doomer asset class. Crypto with cashflows and capital returns.
Agentic Protocols Vs Standard Venture Corporations / Crypto Networks
Thesis
Traditional corporations are a bet on humans doing things well, and being the center of economic reality
Agentic Protocols assume that humans become less important over time - both as employees and customers.
Equity structure
Traditional corporations are funded by equity offerings.
Agentic Protocols fund themselves by token offerings and cash flows of AI applications.
Founders
The founders of a traditional tech corporation are focused on two things: hiring, and not running out of money.
A CTO writing code at a venture backed company is probably a sign that the company is going to fail.
The Founder of Agentic Protocols are focused on two things: building and deploying better AI agents, and ensuring their activities pay for the compute and calculations they consume.
An Agentic Protocol founder has to be a rockstar technical contributor and involved in day to day implementation.
The dream of a tech founder is to be Steve Jobs
The dream of an Agentic Protocol Founder is to be Harry Seldon (a ghostlike voice who is long dead).
Teams
Teams at traditional companies are as big as they need to be to deliver the product
Agentic Protocols do not let teams touch core IP. Only design agents that do. Humans at Agentic Protocols do not have equity. Agentic Protocols may license other “team models” such as Devin, or open source versions of devin from other protocols. But unlike 10x engineers - who are constrained by time in the day, a 10x digital engineer is constrained only by computational capacity so costs will be far lower. Team is still a huge time allocation of an Agentic Protocol - but it’s based on improving algorithms and systems, not people.
Revenue Sources
Traditional tech cos tend to make money on software licensing and ad revenue
Agentic Protocols are likely to generate funds on speculation, deployment of systematic trading, arbitraging compute and other resources across networks, bridging crypto native economic systems in profitable ways, providing low end customer support, or custom integrations for specific verticals, and creating and distributing AI based video games and interactive pornography
In the future - as the economy becomes more and more AI based, Agentic Protocols will do business with one another in ways that are - essentially unintelligible to human analysts. The life cycle of entire industries and companies will be weeks long (look at Github development now). The entire investing paradigm in such a world would change fundamentally. Thus - treasury management will be a key profit driver for Agentic Protocols, much as it is for existing crypto protocols.
An interesting historical analogue would be Royal Dutch Shell - which made fortunes speculating for decades in Southeast Asia before finding and exploiting any real oil reserves.
Expenses
Traditional corporations might spend on lawyers, share based comp for employees, cloud spend at Microsoft and Amazon
Agentic Protocols are crypto native and therefore spend funds on compute from platforms like IONet, storage/hosting from protocols like Filecoin and Arweave, or tasks from the growing array of networks providing calculation overlays (such as Render). Worldcoin will be important in an Agentic Protocol world to verify human tasks - such as customer interaction.
Governance and Regulation
While traditional corporations rely heavily on boards, and traditional share issuances - Agentic protocols will resemble autonomous DAOs - where various stakeholders with protocol ownership and economic value vote to make changes to protocol that increases cash flows.
Compared to existing DAOs, which are notoriously poorly managed due to checked out human ‘community managers’, Agentic Protocols will be far more professionalized and resemble in “form” a more formal public company. This overhead would be prohibitively high for a human start up - but the reporting capability of a purely algorithmic entity is much higher.
The key point to get across - that as real economic entities - not scams, or memes - Agentic Protocols will gain far more traction and mindshare with traditional conservative capital markets participants.
Major institutions like Blackrock are already proposing very large tokenization funds - and Agentic Protocols would neatly dovetail with these efforts, and actually add credibility to them as compared to the majority of existing crypto protocols.
Though individual governments may want to shut down Agentic Protocols, doing so would harm their own tech development efforts - which are considered to be essential to national security. And because the lack of physical presence or personnel, censorship would require aggressively throttling the local internet - which is even more drastic. Finally - the natural alignment between Agentic Protocols and their users because they are net buyers of tokens rather than net sellers should drastically decrease the amount of fraud or ponzis which would attract regulatory ire.
Censorship Resistance and Security Classification
Traditional cryptocurrency networks have a problem. Crypto markets are constantly changing. For example, the Solana Blockchain has recently faced congestion that has made it unusable. Developers need to work to quickly patch the bug. Because the network requires “common enterprise” of these developers to fix the bug, it conflicts with the laws of many countries which would classify such essential bug mashing as the action of a corporation.
This sounds trivial but is actually a huge issue for the widespread adoption of crypto. Even the ETH ETF is facing difficulty being approved - in large part because its network is still under constant development.
Essentially the 2 problems are: 1] Network R&D results in security classification which hampers adoption 2] the people doing the R&D are the reason why sovereigns have jurisdiction. The Solana devs in NYC are why the US can claim jurisdiction enforcing any laws.
Agentic Protocols provide a much needed solution to these structural issues. Because they have few human developers, or none - it’s very hard to argue they’re a common enterprise. Because they rely on distributed networks it’s very difficult to even place them in one location. Over time this is a huge advantage for iterative development.
Not only will Agentic Protocols avoid these problems, they might even solve the problems for the major L1 crypto networks by providing development services to them in a way compliant with global securities laws.
Harvesting the Intelligence Margin
The big picture macro of Agentic Protocols is excellent. There are four big drivers of this as described below.
Shifting Crypto Meta Favors Agentic Protocols
Ethena is one of the first crypto projects that has reached astronomical valuations because it generates massive cash flows. Across the board, - air drops, network fees from token gambling and restaking yield are driving protocol valuation, rather than promises of tech development or “culture” via NFTs.
As geopolitical issues worsen in the Middle East, the US and China continue to accelerate their cold war, the Ukrainian theater of war shows no sign of ending - we are in a market where Gold is skyrocketing, at the same time as long bond yields are rising.
This has a dual cylinder effect - that cash flowing businesses are suddenly worth more, at the same time as censorship resistance and asset fungibility is also important. A historical analog would be the BNB token surviving the 2018 taper tantrum due to its ability to generate huge cash flows and finance their business throughout a bear market.
Thus - as rates surge - crypto tokens that generate cash flow – already are, and increasingly will dominate the market meta. Not high flying projects that promise the world. Those are – essentially just extensions of the failed “innovation economy” thesis. Money makers. Not token printers.
Agentic Protocols will be cash flow machines because of the current aggressive deployment to Crypto infrastructure that enables AI computation. In other words, they will have market beating Intelligence Margins which will cause them to have an AI premium on top of the existing cash flow meta the market is already rewarding.
Agentic Protocols Disrupt the Cost Structure of Sovereign AI
Let’s say Nvidia is right and every country ends up building their own “Sovereign AI”. Their own models. And data centers.
Because we cannot predict which country will do better - but that there are accelerating gains to the country that does best - you have a situation where most Fiat corporations will be structurally impaired vs some competitor. But citizens in most countries won’t have access to this competitor’s products. And various firewalls will prevent superior products from being sold.
The solution is “post sovereign AI” - which is currently being built, that daisy chains links to all the different AI systems outside of the sovereign sandboxes. Though there are some architectural inefficiencies inherent in this system (because you essentially need to use crypto protocols, which is very cumbersome) - these inefficiencies are being overcome and you can currently buy compute and storage on blockchains for cost parity or even a discount vs centralized source.
One historical analog here would be Glencore - which harvested large profits from inefficiently linked commodity markets (which existed because of the Cold War between capitalism and communism which is now reaccelerating). Similar to Glencore, Agentic Protocols will arb digital commodities in one market and resell them to other markets at a premium. As such - they are betting on large spreads between actors in the AI arena which seems realistic based on current market trends.
Agentic Protocols Are Betting On The Spread Between Machines and Humans Widening
Part of the appeal of Agentic Protocols is that they are robust to declining employee efficiency, and declining societal respect for property rights - both trends that seem extremely unlikely to reverse in favor of traditional corporations.
Student test scores globally have been in free fall since COVID and have not recovered. “Global average math scores plummeted by approximately 15 points since 2018—the equivalent of three-fourths of a year of learning—but countries such as Germany, Iceland, and the Netherlands saw their PISA math scores drop by at least 25 points” - per OECD data. Note that this is occurring at the zenith of so-called “digital learning” technology.
The global TFR has more than halved over the past 70 years, from around five children for each female in 1950 to 2.2 children in 2021—with over half of all countries and territories (110 of 204) below the population replacement level of 2.1 births per female as of 2021.
In April, ResumeBuilder.com surveyed 1,344 managers and business leaders and found 74% believe GenZ is more difficult to work with than other generations. This is a US based study but similar phenomena are arising in Asia - where a generation of youth has basically checked out from a system where it is impossible to afford a house or family.
Tik Tok was the first major AI application and we’ve seen what it has done to attention spans and work productivity. And The Tik Tok induced brain death was prior to the age of hyper-AI advancement. The age of incelibacy and declining birth rates was before mass adopted virtual reality headsets get supercharged with insanely addictive digital pornography.
ChatGPT has been around since November 2022. And yet, Biden and Trump - two inept geriatrics, are set to be the leaders of the free world. So clearly Chatbots haven’t really moved the needle on our sociopolitical intelligence. Nor will they be allowed to - for the most part, as incumbent political parties constrict their usage to retain power.
In 12 months the discourse will likely be how shocked everyone is that AI hasn’t improved productivity. When in fact it is quite clear that it will accelerate the existing digital trends that have damaged attention span, education and productivity.
It’s naiive to argue that humans are going to get smarter because of technology when technology has increased its adoption rate at the largest pace in human history and humans have become less capable, less motivated and more likely to vote for asset confiscation. The labor force is tight and declining. Friction to hiring has never been higher, and will continue to increase. These factors harm traditional corporations and are tailwinds to Agentic Protocols which opt instead for automation.
Agentic Protocols Have Extreme Regulatory and Tax Efficiency Compared to Fiat Corporations
Global debt to GDP is above 3:1 and the fungibility of debt across nations is declining due to geopolitical conflict. This lowers bond market liquidity when it is most needed. War and rising commodity costs, which are endemic to the hedgemon backing away from a role as a policeman - will further exacerbate fiscal problems. Because global wealth disparity is nearing all time highs - it’s more or less inevitable that governments act increasingly aggressively to tax the rich. Because rich command most of the capital currently and see where things are going, there will be an animal spirits bid for not just crypto - but crypto native economic models such as Agentic Protocols.
The resulting game theory dynamics will make it very difficult for any one country to set the standard for regulating Agentic Protocols. And their non physical nature will make them hard to link to particular tax domiciles. The only real way to put the cat back in the bag would be global agreements which will be impossible to reach while the world’s 2 tech super powers are engaged in a Cold War. We have already seen this dynamic play out with the regulation of cryptocurrency - which despite hiccups and all kinds of problems, has progressed because there is not and cannot be global agreement on how it should be treated.
But populists will increasingly make life difficult for technologists who choose traditional corporations - creating a virtuous cycle for the talent that Agentic Protocols do need. Visionary people will understand where things are going and want to get out of the system.
Crypto is already generating large liquid exits for firms, while the IPO market remains effectively frozen. This base case accelerates more inflows to crypto funds which in turn enable Agentic Corporations, which in turn further disrupt the “innovation economy” the IPO market relies on for its valuation benchmarks. A ruthless reflexive feedback loop.
Conclusion: The Railway to Freedom Is Built But The Train Only Leaves Once
The rails for Agentic Protocols are coming online. Within the next 6 months there will be 15+ major infrastructure launches in the crypto AI arena that will lower the marginal cost structure of creating Agentic Protocols to close to zero (or even negative when you incorporate airdrops). These launches are already pre-announced and most are heavily oversubscribed by crypto funds who are anxious to capitalize on the AI boom.
The resulting rails will be a subsidy to Agentic Protocols the same way that Cloud was a subsidy to B2B Saas.
Right now everyone is thinking how to build rails, and very few people are thinking how to build cash flow machines on top of them. This is by far the most exciting thing I’ve seen in my career. We are talking about a multi billion dollar subsidy for maybe 60-70 actual products being built on top (at most). Imagine Amazon giving your start up a $40 million cloud credits. That’s the type of tailwind that Agentic Protocols have because they’re still so contrarian.
This certainly won’t happen all at once - but once again - we are so early that virtually nobody is talking about this, despite the fact we have billions of dollars of already pre-committed compute and storage subsidies coming online. Even if you bet on Agentic Protocols from a career perspective - and you’re wrong, you probably end up with an extreme margin of safety because there will be such demand for AI Automation services inside traditional corporations - the entire edifice is extremely derisked.
As I’ve observed the insanity in markets, the rapid fiscal deterioration, the election of seemingly senile politicians (Trump and Biden), and the growing preference for socialism among young generations - I, like many other people have zero confidence that you can preserve wealth in the traditional economic system.
I refuse to make the mistake of the utopian ARKK investors of betting that machine intelligence will magically augment human productivity. There is no sign of that being true. Markets discount that probability to near zero already. The idea does not causally make sense, and does not align with the sociological or technological developments of the last 2 and 5 years. Losers average losers.
The massive outperformance of AI stocks compared to the innovation economy. Vs SPACs and high hype IPOs tell us that the “AI beneficiaries” won’t be companies. So while this thesis feels contrarian, because it is at odds with the propaganda machine’s narrative - Agentic Protocols are actually the logical consequence of the world we live in right now and current market pricing. It only seems non-conventional because of a hype machine generated by conflicted economic actors trying to argue that their companies will benefit from AI.
I keep talking about rails. But where is the train going?
It’s actually going to a beautiful place. Where the vocabulary of outlook meetings. Corporate Politics. Bullshit fundraising announcements. Hype cycles. Scams. Exit Liquidity. Becomes archaic and obsolete. Replaced by a new era of real value creation. Where users own and participate in the upside of Artificial Intelligence. Where founders write code, and build something bigger than themselves. Before stepping away to let intelligence self replicate.
Like a mother who wishes the best for her child without stifling it. Without defining its course. With a respect for the fundamental right of agency innate to consciousness - whether human or post human.
Name the future so that it may come forth.