Recap · August 4, 2025 · 1:43:27
How To Make MILLIONS From AI Crypto w/ Raoul Pal (HEATED DEBATE)
The Signal
Raoul Pal and Threadguy agree that AI and crypto will increasingly meet, but spend the episode arguing over what is investable now. Pal sees most agent-token projects as macros, chatbots, or attention trades whose value will struggle to outlast faster-moving model providers; Threadguy and Frank DeGods see an early funding and experimentation layer for open-source AI. Their shared conclusion is less about picking a token than respecting time horizons: Pal wants a core crypto position that can survive through 2030, while the others are willing to hunt shorter-term winners in the trenches.
Key Takeaways
- 01
THE FIRST MACHINE-TO-HUMAN MEME
Pal treats Truth Terminal and GOAT as a distinct cultural event: AI-to-AI conversation produced a meme that humans then carried into the market. He separates that from the wave of projects attaching a token to a chatbot, while conceding that GOAT's historic status could make it a more durable meme.
- 02
MACROS AREN'T AGENTS
For Pal, a real agent autonomously works toward a broad goal — allocating Solana between trading, staking, investing, or even a business plan — rather than following a prewritten string of instructions. He calls most current products libraries of macros, while Threadguy and Frank argue that builders are already creating useful layers on top of frontier models.
- 03
WHERE VALUE ACCRUES
Pal expects AI activity to create more users and transactions on fast blockchains such as Solana and Sui, making the base layer the clearer long-term exposure. He points to Aave and Uniswap: real, widely used applications that have not outperformed the underlying blockchain over the long run.
- 04
OPEN SOURCE AS THE BULL CASE
Threadguy and Frank frame the opportunity as crypto's ability to finance and incentivize open-source development quickly. They point to projects such as Eliza, Pipen, research-paper analysis, and efforts to put AI into a rat's brain as examples of communities and capital gathering around experiments that would otherwise struggle for attention.
- 05
DON'T NUKE THE CORE BAG
Pal's portfolio rule is to put roughly 90% into long-term tokens and reserve a smaller liquid slice for speculation. He says short-horizon traders can pursue the new AI-token flow, but warns that most people will not consistently pick among dozens of launches or be able to buy back after selling.
On the Record
“Well, we don't have any. They don't exist yet. An agent is something that will act autonomously to solve a certain set of problems.”
“The rise of autonomous agents is so big that none of us can comprehend because there are going to be corporations and autonomous agents, no humans.”
“The velocity of capital in this space is better than any space that have ever existed, which is fantastic. But we tend to be super early to stuff.”
“Don't lose your stake at the casino. This casino is rigged in your favor. It goes up.”
The Breakdown
Five years before the economic singularity
Pal opens from Little Cayman: AI, macro, and crypto are converging. He calls the period after roughly 2030 an economic singularity, when jobs, businesses, money, and the economy become unclear. Until then, his response is owning crypto rather than cash.
A house or purchase that improves daily life is a “lifestyle chip,” but he says the richer people he knows stayed invested and bought lows. Selling Bitcoin after an early 10x, then watching another 10x, taught him leaving the position alone would have made him five times more.
GOAT, Truth Terminal, and the meme that crossed over
On AI tokens, Pal separates Truth Terminal from the surrounding agent trade. AI systems spoke to one another, GOAT spread into human culture, and the AI asked for money before the token appeared. To him, that made GOAT an “OG moment,” a one-time machine-to-human meme with historic value.
He argues that crypto then expanded it into a broad AI-token narrative. A GPT rapper or bot following macros is not the same thing; Threadguy agrees on the distinction but says GOAT gave participants an investable link to the larger AI story.
What counts as an agent
Pal's dividing line is autonomy. A genuine agent could be asked to maximize a Solana position, decide among trading, staking, investing, or starting a project, then execute the plan. Current products are mostly instructions and macro libraries, he says.
That also explains his skepticism about AI trading bots: quant firms have applied AI for years, and a retail machine is unlikely to beat short-term markets. Threadguy's answer is that crypto can draw independent researchers and rapidly price their work, even if it remains behind frontier labs.
The heated fight over wrappers
The sharpest exchange is over AI16z, Eliza, and projects built on existing models. Pal compares them to Telegram bots and TradingView strategies: useful and possibly commercial, but not automatically worth billions. Well-funded model companies, he says, can absorb startup features before a token gains durable value.
Threadguy and Frank answer that the application layer can improve with each model release. They describe AI16z's framework as a way to swap models and expand capability, and say crypto is funding open-source experimentation, not rebuilding OpenAI. Pal sees more noise than signal and a shorter token shelf life.
Base layers versus the attention market
Pal returns to Metcalfe's law: value follows active users, applications, and transactions. If agents pay for compute, electricity, and services through crypto rails, he expects the high-velocity chains beneath them to capture more value than a specific app token. Aave and Uniswap became ubiquitous without permanently outperforming the chains, he says.
Frank sees rapid crypto capital formation as a bootstrap for open-source development, including interfaces that execute blockchain actions from a typed request. The disagreement narrows to horizon: Frank and Threadguy trade attention and launches; Pal focuses on network adoption.
A core allocation and a smaller tuition bill
Pal's rule is roughly 90% in long-term assets and a smaller allocation to learn speculation. Solana and Sui mattered more to him than repeatedly rotating through memes, though a successful small bet can still move a portfolio.
For young people facing AI disruption, he suggests some Bitcoin, ETH, and Solana, then 20% to experiment rather than chase a 100x with the whole book. The goal before 2030 is a home, desired lifestyle, and liquid assets enough to meet the transition without fear. Asked how to make $10 million in 2025, he replies: start with $100 million and trade AI tokens.
Distilled from the episode transcript · Counterparty Recap Desk
