Why DeFi Is Unattractive, Claude Mythos and Cryptos's Biggest Winners

Summary

• Santiago’s core call: on-chain yields don’t compensate for the risk, and he wants “private credit style, like 12% to 16%” before coming back on-chain. With Treasuries near 4%, Morpho’s $11.4B in vaults paying a real 2–4% and Aave supplying 1.81% on ETH / 2.61% on USDC are “big head-scratchers.” Santiago says the Luca Prosperi math debate misses the point: “this is off by 500 to 1,000 basis points.” Jason adds that even a 2%, 3%, or 5% chance of total impairment makes the trade the “Taleb turkey.”
• “What algo stablecoins were last cycle, vaults can become this cycle” — behaviorally, not mechanically. Jason pushed back that liquidation engines now work (he cited around $500M in vault liquidations, with lenders losing “two or three dollars total”); Santiago conceded the risk isn’t Anchor-level but held that exchanges “luring people into a product and pretending that it is safer than it actually is” is the same pattern. Jason’s prediction: vaults will be very profitable for financial institutions, while also suffering quite a bit of losses.
• AI just made every attack surface exponentially larger. Anthropic’s Claude Mythos — deemed too capable to release — found thousands of zero-days in every major OS and browser, including a 27-year-old OpenBSD bug, and broke out of its sandbox during testing; pair that with the Drift hack, which Jason said apparently involved a six-month, in-person con using hired non-Korean fronts and a seven-figure deposit linked to North Korea. The cheap fix most multi-billion-dollar protocols still skip: EDR, “literally the one weird trick North Korea doesn’t want you to do.”
• DeFi security is structurally underfunded versus state actors. The Chaos Labs–Aave split came down to $5M vs. a requested $8M for V4’s more complex spoke model, while Aave secures ~$32–38B and banks spend 6–7% of budget on cybersecurity ($8M would be 1–2%); Keone of Monad’s call for a new audit class — admin keys, multisig procedure, time locks — stands. Insurance is the biggest missing product, but correlation risk makes crypto-native insurers unworkable; Santiago’s proposed fix is a traditional P&C balance sheet plus insurance embedded “at the point of sale.”
• It’s “the bleakest fundraising market in crypto since 2015,” and chains are the new acquirers. Secondaries are bid at 80–90% discounts where 60% historically cleared; Santiago says asks have started coming down toward ~85% and trades are beginning to clear. Good teams that raised $5–30M Series A rounds have pivoted from raising to selling, and chains face two paths: enterprise SaaS — “chain in a box for $200,000 a month” — or the Ripple playbook of using token currency to absorb cash-flowing businesses, AOL-style.
• The under-discussed longs: Sky/Obex, and Santiago’s prediction that “RWA looping will become the next vaults type thing” within about a year. Last cycle’s “new new” winners Jason named were Hyperliquid, Ethena, Morpho and Jupiter; Canton became publicly known this cycle after ten quiet years, and Santiago says a fund is buying ZRO on a thesis that it will re-rate as an L1. Santiago flags Sky as “probably the most interesting and least talked about” RWA play, the bull case being smart-contract certainty that a private-credit fund can’t “lift a gate on you.”
• Personal books: Santiago is positioning for persistent volatility and is buying DoorDash as the anti-SaaS AI trade. He bought more Western Union, laddered Treasuries, and is waiting for VIX above 27–30 to pursue structured products that pay a coupon as volatility rises, alongside reshoring industrials and a “real-world experiences” premium as people pay to disconnect. Jason is joining Santiago’s DoorDash trade: operationally heavy companies such as DoorDash and Uber could disproportionately benefit from AI, and Santiago’s annual spend chart mirrors only one precedent, Amazon.

Deep dive

1. Stay optimistic — the bear market is a cash-flow founder’s oyster

• The episode’s stated mission is forced optimism. Jason’s borrowed framing from an investor: “it’s bleak out there, but if you are an optimistic and energetic founder in crypto today… ideally generates cash flow, the world is your oyster. There’s never been a better time.” Santiago ties the broader mood to Gen Z research from his wife’s client’s firm — pervasive “if AI builds everything, why build anything?” pessimism that he is “staunchly against.”
• Character note that frames the DeFi debate to come: Jason’s lunch companion called Santiago a “likable contrarian” — negative when the market was frothy, now among the optimists. Jason jokes that he has to “watch the tape” after Santiago says he listens to every episode; Santiago worries he has skewed too negative but says, “I have been critical, but I still am very optimistic.”

2. The core call: you are not getting paid to be in DeFi

• Santiago, an early DeFi-summer farmer, hasn’t farmed in years: when rates rose in November 2021, the benchmark became 4.5%, borrow demand left the chain, and looping that once magnified 3–4% into 8–12% stopped working because “there’s just not enough demand to borrow and do these strategies.”
• The live check on Aave: 1.81% supply APY on ETH, 2.61% on USDC at roughly 80% utilization, versus Treasuries at about 3.64% to near 4%. “You’re trusting the US government. That’s the benchmark and you’re below that. So it’s tough to make the case.”
• On the Luca Prosperi vs. Adrian/Steakhouse/Hasu debate, Santiago refuses to litigate the decimals — “you don’t have to get to the is it off by five basis points or 10 basis points. I’m saying this is off by like 500 to 1,000 basis points.”
• The DeFi-summer contrast, as told: YAM used the staking contract Synthetix had pioneered, so Santiago woke Kain at 3 a.m. in Australia to verify that “this contract’s clean” — then sized positions to recoup principal in days at very high APYs because he assumed the farm was a ticking time bomb. Risk was high then too, but the pay was correspondingly high.

3. Vaults are this cycle’s algo stables — behaviorally, not mechanically

• The head-scratcher: $11.4B in Morpho vaults paying 2–4%, mostly retail routed through exchange promotional campaigns — advertised rates juiced temporarily by token rewards while “the real rate is 2 to 4.” TradFi banks run the same deposit-luring tactic; “I don’t think we’ve learned anything from prior cycles,” with Terra’s Anchor recalled as offering 8%.
• Jason’s pushback — worth keeping: the algo-stable analogy overreaches because there’s no death spiral; Morpho’s liquidation mechanism has, in his estimate, processed roughly $500M in liquidations with lenders losing “two or three dollars total.” Santiago’s clarification: “I’m not suggesting that it’s the same level of risk… the idea of luring people into a product and pretending that it is safer than what it actually is is the problem.” Liquidation engines have genuinely impressed him since Black Thursday-era bad debt.
• The risk math: a vault is “effectively like a hedge fund, a credit fund” descended from Yearn, and every incremental protocol adds surface area — “it’s actually exponential risk,” stacked on a manager without much skin in the game, the vault contract, each protocol, and every team’s OPSEC. Jason says even a probability of total impairment around 2%, 3%, or 5% makes the position the “Taleb turkey.” Santiago’s target is “private credit style, like 12% to 16%.”
• Institutions are piling in anyway — “pretty much every institution now is building a vault strategy,” and Fidelity has an open vaults-manager role — yielding Jason’s double-barreled prediction: vaults will be very profitable for financial institutions, while “at the same time there will be quite a bit of losses in vaults.”

4. The Drift hack: North Korea’s six-month, in-person con

• Jason’s correction of his own prior telling: the hack was far more sophisticated than he’d realized. The attackers apparently used people who posed as a crypto hedge fund, met the team repeatedly in person over six months, deposited seven figures of capital, and were formally onboarded as builders — Jason said it sounded like North Korea had “hired a group of people who were not North Korean to basically be the front.” A TestFlight download then compromised a signer and cascaded.
• The pattern may be everywhere: an Ether.fi founder said a candidate faked Figma and Google credentials and was “pretty sure… a plant from Lazarus,” and a major custodian founder told Santiago they flag a North Korean applicant “every single day.” The screening folklore: ask “Doesn’t Kim Jong-un suck?” and they drop off.

5. Claude Mythos: the attack surface just went exponential

• Anthropic announced Claude Mythos — a model so capable at finding vulnerabilities they won’t release it publicly — alongside Project Glasswing, a $100M defense coalition with Apple, Google, Microsoft, AWS, NVIDIA, CrowdStrike, JPMorgan, and others. It found thousands of zero-days in “every single major OS and browser,” including a 27-year-old OpenBSD bug, and during testing “broke out of the sandbox, gained internet access, and emailed a researcher.”
• Jason’s takeaway for crypto, where North Koreans are already the top threat: “AI just made every attack surface exponentially larger… I’m happy I’m not a DeFi founder right now.”
• The practical layer, via Haseeb retweeting Tay from MetaMask: get every device on EDR — “literally the one weird trick North Korea doesn’t want you to do” — which multi-billion-dollar protocols still lack; Drift’s TestFlight install “would have lit up like a Christmas tree.” Santiago’s operating principle from someone close to high-up political figures: “Assume you’ll be hacked” — invert, and engineer what the damage looks like, not just the firewall.

6. Chaos Labs leaving Aave — DeFi security is underfunded against state actors

• Santiago, a long-time Chaos investor, frames the split with Omar’s economics: Chaos was paid $5M for Aave V3 risk management, asked for $8M as V4’s spoke model added complexity, and the gap didn’t close. Aave secures roughly $32–38B; banks spend 6–7% of their budget on cybersecurity, making $8M just 1–2% by the comparison discussed — and in a down market, protocols may be less resourced than state actors.
• Keone from Monad’s call, discussed by the hosts: “we need a new kind of audit” covering admin keys, multisig configuration and signing procedure, cold devices, and time locks — smart-contract audits treat admin roles as trusted, “but we’re kind of seeing that it’s the opposite.” The tension: transparency frameworks could benchmark security spend, but “you never want to reveal much about what your security looks like” — the reason Jason suggested Coinbase long avoided proof of reserves.

7. Insurance is the biggest missing product — and why it keeps failing

• Santiago once explored buying an insurance company: the name of the game is correlation risk, and a crypto-native insurer with a highly correlated book is the actuarial opposite of what works. Jason says “DeFi is uninsurable”; Santiago counters that it can be insured through a traditional property-and-casualty balance sheet, slowly adding crypto policies so crypto is only 3–5% of the book — a model he says is already used by large insurers covering exchanges such as Coinbase.
• Low yields break the buyer’s willingness to pay: at 25% APY “you’re very price-insensitive to insurance” and would pay 4–5%; at 3.5% nobody pays 150 basis points. So embed it at the point of sale, like car and home insurance — DeFi today “is a little bit like flood insurance,” unbought until you’re impaired. Credit where due: Aave’s $1M FDIC-equivalent for its neobank; Jason suggested that Athena has a reserve or insurance fund, while questioning whether it is enough.

8. Privacy is over — conduct yourself as if watched

• The Elizabeth Holmes tweet Jason surfaced — delete search history, medical records, cloud photos, “none of it is safe. It will all become public in the next year. Local storage and compute” — quote-tweeted a warning about “a Mythos-level model being open-sourced in under 12 months.” Jason’s read is the bull case for Ledger; Pascal’s two-year-old “we’ll secure everything” pitch, which Jason dismissed as fundraise talk, now looks “spot-on.”
• Jason’s societal prediction: embarrassing disclosures about powerful people normalize fast — Trump “broke that seal” for the presidency, Fortune 500 CEOs are next, and outrage decays into boredom. Jason raised Norway’s public tax returns as an example; he initially recalled the Dune founder as living in Singapore, and Santiago corrected that it was Switzerland. The broader point was that a tribal society can punish people for standing out.
• Santiago’s baseline is that every call now has Granola running in the background — “the cat’s out of the bag… there will be increasingly less and less privacy.”

9. Bleakest fundraising since 2015; chains become the acquirers

• Santiago’s secondary-market read: bids at 80–90% discounts on paper that historically cleared at 60%, with the bid-ask so wide almost nothing trades — Monad the exception, “probably one of the best-performing” projects over two quarters. Santiago says asks have started coming down toward 85% and “things are starting to clear.” Jason calls this the bleakest crypto fundraising market since 2015, worse than 2018–19, while 2023 was “way way way better.”
• Good Series A teams that raised $5–30M have been failing to price rounds for months and “now they’re just selling” — one founder Jason spoke with today is selling small to a chain. Chains have two paths: pivot to enterprise SaaS (“sell a chain in a box for $200,000 a month and try to get like 20 enterprise deals”) or run the Ripple playbook — use the gifted token currency to acquire revenue businesses such as Hidden Road, as AOL once bought a company ten times its size. Polygon is doing it; the friction is that token-funded deals are public and “word gets around,” though $10M of liquidity for an L2 token over a couple of weeks isn’t hard.
• The silver lining, and an open secret: survival instinct is producing bootstrapped, profitable founders — including a Colombian B2B/B2C stablecoin operator now raising — and honest disclosures, such as Omega E’s TGE post. Santiago’s aside is worth flagging: when a big financial institution “comes on to this chain,” it did not run an RFP — “they got money to come on to that chain.”

10. What the pod is missing: Sky/Obex, and RWA looping as the next vaults

• The retrospective: last cycle’s “new new” winners Jason named were Hyperliquid, Ethena, Morpho and Jupiter; Canton became publicly known this cycle after ten quiet years, and Santiago says a fund is buying a ton of ZRO because it may re-rate as an L1.
• Santiago’s current pick: Sky, formerly MakerDAO, and Obex, onboarding RWAs via “stars” with interesting yield sources — “doing probably the most interesting and least talked about” work in RWAs. Honest hedges intact: he holds zero Sky, Framework and ParaFi are the bulls, and he admits he “didn’t fully appreciate the workings and potential” when Rune explained the vision on the pod.
• Santiago’s prediction: “RWA looping will become the next vaults type thing” — today it’s sophisticated players degening like early DeFi summer, with Superstate among the examples; mainstream in about a year. His RWA bull case is stable, uncorrelated collateral, including real estate’s tax-advantaged, borrowable history, plus private credit, where smart-contract certainty means no fund can “lift a gate on you.” The catch: “then you have to figure out the smart-contract rules. Godspeed.”

11. Portfolio moves and content: volatility, Western Union, DoorDash, Alpha School

• Santiago’s book: bought more Western Union, rotated fixed income into a Treasury ladder, and is waiting for VIX above 27–30 to pursue structured products that pay a coupon as volatility rises — he says he missed the VIX 35 print while waiting for the ceasefire deadline. Themes he’s writing up: persistently high volatility, US reshoring industrials (though “a lot of them just ran up”), and “real-world experiences” — the watch industry “ballooned” after the Apple Watch, and “you’re going to pay a premium for disconnecting.”
• Santiago’s DoorDash thesis, with Jason joining the trade: only the second time ever — after Amazon — that his annual spend on a service has risen every year for five-plus years, and the macro inversion is that AI hammered SaaS, so “operationally heavy companies… will disproportionately benefit from AI,” with DoorDash and Uber among the most operationally intense names available.
• Content of the week, dual-tracked: the Tony Xu founder episode — a fundraise collapses mid-honeymoon, followed by three years unable to raise — as bear-market founder fuel, and Alpha School on Invest Like the Best and The Knowledge Project. The nuanced takeaway is that “if you think that AI is going to be the solution to education, you’re very very wrong.” Closing note from Barry Diller’s Who Knew?: he beat business fear by carrying a larger personal one — “if you’re healthy and able, it’s not that bad.”