Summary

Santos is winding down Inversion’s crypto thesis. After screening 20+ sectors for businesses that stablecoins could transform, he sent investors a letter admitting “filtering for crypto has led us to evaluate businesses that are more fragile than businesses that I would want to own” — and invited disproof: “I’d rather be proven wrong now.” The firm continues, but as a buyer of good businesses where “technology is orthogonal” — it enhances, it’s not the reason to buy.
The remittance autopsy is the tell: despite Mexican families losing 5-6% per transfer being his original crypto aha, he found Remitly and Wise “not making money,” with no pricing power — the agents and fulfillment partners are the ones who “break the laws of physics of making money available simultaneously in two parts of the world.” His verdict after talking to remitters and crypto founders alike: “we can’t make this business more efficient.”
The new hunting ground is boring, mandated cash flow: critical-power servicing, where hospital administrators “can go to jail” if generators aren’t inspected, revenue is contracted and all-weather, geography fragments the market, and emergency calls bill 3-4x — “like software but better,” with AI margin uplift on scheduling and reporting layered on top. Plus businesses whose owners are retiring with no succession plan.
His incentive design is the differentiator: a holding company, not a fund, because committed capital makes you “trigger happy,” and no 2% fee because “no matter how you look at yourself in the mirror,” incentives change you — take a fund from $20M to $2B and “you become less of an investor and you become an asset manager.” Three people, $26.5M raised, mostly unspent; “we will automate everything except judgment.”
On crypto itself: “crypto doesn’t fail” — he can “simultaneously love the technology and be incredibly bearish about valuations.” Stablecoin dollar rails remain the killer use case “no matter what the price of Bitcoin,” memecoins are lottery tickets of an affordability crisis, and with the regulator excuse gone the industry must self-police (he cites the token transparency act).
The AI trade is a barbell: own the extremes — compute, regulation, and energy that “does not go away” on one end, and the irreplaceably human on the other (World Cup and Wimbledon tickets at 15,000, vintage cars over hypercars). A host’s sharpest specific: long pharma with proprietary datasets, because for the first time one “scientist” — the host refers to “Fable 5” — has cross-domain expertise spanning neuroscience, microbiome, sleep, and cognition, and unlike academia’s “mafia,” AI “doesn’t care” about defending old theories.
The method underneath it all: “do you want to make money or do you want to be right?” — “making money is uncovering the truth before it becomes consensus.” Clever people distort the world to fit their thesis; likely Stan Druckenmiller saying “go long Nvidia” at a conference — to audible disappointment — is the model of a man with no identity crisis. “I’m an investor. I so happen to invest in crypto.”

Deep dive

1. He engineered away his accent — and treats every disadvantage the same way

Born in Mexico but routinely taken for American, Santos deliberately erased his accent while growing up — recording himself, studying how others spoke. His logic: “if you are sitting across someone… you want to remove any sort of degradation in the message,” and “consciously or subconsciously someone’s going to judge you.” His blunter version: “anyone can perfect their accent — I hate to say it, but it’s kind of lazy if you have one.” The host’s label: high agency.
His parents gave education and travel, not money — a middle-schooler wanting Harvard summer school had to “go to investment committee.” His parenting conclusion: kids need “time and love and a support system. Not money” — the feeling of “a net that will catch them if they fall.” He credits “all my success to that singular moment” of believing anything was possible.
Asked for the biggest risk he ever took, the answer isn’t a trade: it’s asking a girl to dance at a quinceañera. Betting on Bitcoin “felt like a walk in the park relative to the sweat and the anxiety” — and “the fear of rejection is probably the biggest bottleneck to unlocking potential.”

2. Crypto’s origin story: friction, remittances, and dollars in digital form

His investing method starts with irritation: “in your day-to-day interaction… observe where there’s friction.” The formative case — remittances are the first or second source of income for Mexican families, and “they get ripped off every time… at least 5-6%” moving money from the US.
The Bitcoin aha was cross-border money “faster, better, cheaper” — now expressed through stablecoins. “People want dollars, have always wanted dollars — give them dollars, especially in digital format.” That’s a killer use case “no matter what the price of Bitcoin… everything else is TBD.”
The path in ran through trauma: “talk to any founder… everyone has trauma” — his middle sister got sick, and after two years of investment banking plus two years at a growth equity fund he asked what else life could hold. He entered crypto at 24, noticed “the smartest people” in San Francisco on obscure forums, and his mom “broke down in tears” when he told her.

3. The Inversion letter: filtering for crypto was the mistake

The news of the episode — Santos sent shareholders a letter yesterday admitting that after 20+ sectors of work, “we haven’t found a sector.” The confession: “my mistake was filtering for crypto has led us to evaluate businesses that are more fragile than businesses that I would want to own.”
The remittance deep-dive carries the argument: a large public remitter said stablecoins would transform its business, so he interviewed a representative sample of peers plus crypto founders building cross-border rails. Finding — Remitly and Wise are “not making money,” unit economics are poor, regulatory footprints are huge, and the remitter has no pricing power: the agents and fulfillment partners are the ones who “break the laws of physics of making money available simultaneously in two parts of the world.” Conclusion: “we can’t make this business more efficient.”
He’s careful about scope: the technology is real, and crypto-native firms — Binance, Tether, exchanges, likely Hyperliquid — are “phenomenal” businesses. But for traditional FX, credit, and tokenization plays, the transformation was never “commensurate with the risk” of a turnaround, which demands “a very healthy margin of safety” he never found. He ends the letter inviting refutation: “I’d rather be proven wrong now.”

4. The pivot: technology is orthogonal — buy the business first

The one-liner he wants crystallized: “the mistake that I made was thinking that technology is at the center of why you want to buy a business. It’s orthogonal — it enhances, but you still have to buy a good business.” The filter is meaningful cash flow “through any market environment” plus low disruption risk — “and if you can’t find it, don’t invest. Buy the benchmark.”
The specimen thesis is critical power: every hospital, school, and hyperscaler is federally mandated to keep serviced backup generators — administrators who skip it “can go to jail.” Revenue is contracted and recurring “no matter what latest model comes out,” the market is fragmented by geography, labor is short, and emergencies bill “three, four times the rate.” “Like software but better” — with AI uplift on routing, scheduling, and client reporting as the margin kicker.
The supply side of deals is demographic: a wave of businesses with no succession plan in Germany, Japan, and the US (he focuses on the US “because of rule of law”). The template is Warren and Charlie: “Warren and Charlie were smart because they just found really good operators and gave them money and just compounded.”

5. Incentive design: a holdco, no 2%, and the psychology of fund size

Inversion is a holding company, not a fund, “for a very intentional reason”: committed capital plus a ticking clock makes you “trigger happy,” and sunk-cost pressure after a lot of sector work makes passing “demoralizing.” His counter: “my first and only order of business is protecting shareholder capital.”
Pressed on whether not deploying would have been harder on a 2% fee, his answer is unusually honest: “of course… you can’t look at yourself in the mirror and say your incentives are not going to change. You’re deluding yourself.” Take him from a $20M fund to $2B and “my psychology will change… you become less of an investor and you become an asset manager.” He credits Benchmark for holding its fund at $300-400M for years.
The scale is deliberate: $26.5M raised, very little spent, three people — and the reminder that Constellation started with $25M and Berkshire with small sums. Internal rule: “we will automate everything except judgment.”

6. “Crypto doesn’t fail” — but the industry has to look in the mirror

His stance splits cleanly: “I can simultaneously love the technology and be incredibly bearish about valuations.” Asked whether crypto’s incentive problem is worse than other industries’, he says no — “it just hurts more because it’s an industry that has very high principled values and stakes them.”
Memecoins get a sociological read, not a moral one: “people buy more lottery tickets in a bear market” — an affordability crisis plus crypto opening “the floodgates to people that otherwise wouldn’t have access to investment products.” He’s laissez-faire about it, but notes the dissonance of firms publicly preaching real-world assets while prioritizing memecoins.
The excuse era is over: “for so long we blamed a regulator… now that excuse has kind of gone away.” The industry can self-police — he cites the token transparency act as a good example — and “it’s not surprising that we’re not taken as seriously” when conduct doesn’t match the principles.

7. Truth-seeking beats cleverness — and identity is the enemy

His biggest investing mistakes came from “trying to be clever in how I make money… you want to feel smart, you want to feel validated.” The test: “do you want to make money or do you want to be right?” — because “making money is uncovering the truth before it becomes consensus,” and clever people “distort the world view to conform with their thought process.”
The exemplar is likely Stan Druckenmiller: at a conference (he thinks it was the Collison brothers or whatever), he said “go long Nvidia” — “really? that’s the best you got?” — and it was pretty solid. What Santos admires: a man “incredibly at peace with changing his mind” and no identity crisis about what kind of investor he is. His own version: “I’m an investor. I so happen to invest in crypto” — crypto having forced him to be “simultaneously a high-frequency trader and a venture capitalist… a great education to investing, full stop.”
Even Inversion was framed as truth-seeking, not a pitch: “I want to know before anyone else how big this market is” — trillion, ten trillion, or a couple hundred million — “and then make that trade.” A former boss’s response, as told: “you [__].”
The LP corollary — worth keeping: “some of the best fundraisers are extremely clever… you have to watch out for the people that sound incredible but actually can’t invest.” He points to the “forgotten billionaires” book about one of the guys who was running Long-Term Capital Management and to trust-fund heirs who buy the Goldman SPV “because they think it’ll differentiate them.”

8. The AI barbell: own the extremes and the irreplaceably human

His synthesis of what “smart guys” are saying — software investors declaring “software is obsolete, it’s a commodity,” services matter, hardware is back — is a barbell: “you want to be investing in the extremes.” One end is AI infrastructure — compute, regulations, and energy, which “just does not go away.” The other is the human experience: World Cup and Wimbledon tickets at 15,000, “people are craving the offline,” and vintage cars over futuristic hypercars — “the nostalgia factor… is going to be even more valuable.”
A host’s sharpest sector call: “I’m very long pharma companies that have a lot of data.” For the first time ever, one entity can “make relationships across data sets” with deep domain expertise spanning neuroscience, the microbiome, sleep, and cognition — “you’ve never had a single scientist” like that, and the host refers to “Fable 5.” Academia can’t do this because “the mafia of academia is worse than any other mafia” — a 60-year-old whose life’s aging theory gets disproved never says “I was wrong.” AI “doesn’t care, and it breaks through that.”
The story that carries his optimism, as told: an Australian with a sick dog and some biology knowledge iterated his way to an mRNA vaccine, fought regulators who said he wasn’t an expert, convinced a lab to manufacture it — and “the dog is cancer-free. That guy’s going to build a billion-dollar pharmaceutical company… If you’re not looking at that and not inspired, I’m sorry.”
His one caution: “the hallucination risk is real,” and model weights and “information that is being served” deserve attention — hyperconnectivity makes people “think that we’re informed and that’s very far from the truth… which means markets are more inefficient.”

9. Spiky beats well-rounded — and this is the generalist’s moment

The episode’s signature riff, aimed at his own mother: “She was like, ‘You need to be more well-rounded.’ I’m like, that’s the biggest load of [__] I’ve ever heard in my life… That’s mediocre. That’s literally a bell curve.” Education “sands even the spikiest of people” — he was an SAT tutor and insists gaming the score “doesn’t make you a first-principles thinker.”
Spikiness, in his telling, is made, not born: “spiky people have become spiky… in the same way that you perfect an accent” — and reinvention is always available: “you’re under no obligation to be the person you were two minutes ago.”
Hence the moment: “experts are overrated,” the marginally delusional outsider drives innovation (Musk isn’t an aerospace engineer; penicillin came by accident), and AI collapses the “degradation in the assembly line” between the idea in your brain and a built product — he cites “claw code” (likely Claude Code). His own state: “I’ve learned more in the last six months… it’s like a renaissance. It’s the most exciting time of my life.”

Inference Is the New Training — Philip Kiely and Ali Taha, Basten Why smarter AI models could drive up compute prices 10x