Hivemind: Ceasefire In Iran, Is AI a Bubble & Drift Exploited For $280M

Summary

• Twelve hours into the Iran ceasefire, the Delphi desk is tactically long but nobody believes it holds. Yan says positioning favors longs because most people were rightly sidelined, yet his base case is negotiations break down “maybe before the 14-day window or at the end of it” since the demands are “diametrically opposed… completely mutually exclusive.” Jason is more cautious still: markets sit a few percent off all-time highs “even though things are as bad as they’ve ever been,” so play defense or “just not play the game.”
• The research lead calls Iran charging tanker tolls in Bitcoin or yuan “one of the biggest fundamental developments in Bitcoin in a while.” His frame: “it’s becoming clear now that the US isn’t the world’s dominant power,” countries are cutting independent deals, and that multipolar order is exactly what BTC-as-neutral-store-of-value needs. The room’s pushback — no real dollars will use Bitcoin, “it’s more of a headline thing” — is followed by Yan calling it a good headline that could help Bitcoin fulfill that role.
• The AI IPO wave — SpaceX, Anthropic, OpenAI — is a “ticking time bomb” against finite dry powder. SpaceX going out around $1.7T, or roughly $2T in Santiago’s framing, on roughly $15-20B revenue, versus Meta at $1.5T on $200B, drew “the gall to IPO at a two trillion dollar valuation”; Jason sees high probability it trades below IPO price, while Nasdaq inclusion-rule pressure and the Fundrise VCX fund ($19 NAV trading to ~$550-560) show the retail exit being built. Databricks vs Snowflake — one around $150B and the other around $50B on roughly $5B ARR each — captures the private/public mark gap.
• Santiago’s regime call: “there’s more alpha than ever in public markets” after two decades where “nothing ever happens.” SaaS names got indiscriminately dumped 50-60%, AI supply-chain theses (energy, memory, electrification) are multiplying, and dispersion inside the S&P is where long/short wins — though everyone concedes most people should still just own the index, because outperforming remains hard by definition.
• The stablecoin endgame is every app internalizing its own float — bearish USDC, while “Tether might be insulated.” Polymarket’s new native USD keeps the treasury yield (“when you don’t charge fees, you got to make some money”), Solana earns nothing on ~$15B of stables from an issuer that “doesn’t even freeze 200 million when there’s an exploit,” and Sky’s USDS is recently around $12B after absorbing what seems like much of Ethena’s supply through better bear-market yield sources.
• The Drift exploit was described as a six-month, in-person North Korean social-engineering operation, not simply a code bug. Santiago withdrew everything from admin-keyed Solana DeFi, and an internal Delphi poll found the first ten respondents — all industry veterans — hold under 25% of their crypto in DeFi. Santiago’s sharper cut: an admin-keyed “DeFi” protocol is “the worst of both worlds,” and Anthropic’s new vulnerability-finding model reopens the smart-contract attack surface too.
• The tactical book around the desk skews BTC, HYPE and heavy cash, with selective alt adds. Santiago holds only BTC and HYPE long term and is otherwise very cash-heavy; Jason mentions a small Zcash position, up 30% in the last day. The research lead remains mostly long equities and lists HYPE, Bitcoin, Zcash, ENA and PUMP; Santiago bought the Trump-speech puke, owns ENA (“insanely under-owned, super beaten down”) on USDe’s supply floor plus new prime-lending yield, likes AI survivors NEAR, Grass and Venice, and is bullish Lighter — met with “when has buying the number two in crypto that looks cheap ever worked out?” HYPE’s HIP-4 testnet (prediction markets, unified spot-margin) is Jason’s watch item.

Deep dive

1. Ceasefire day: long the tape, short the negotiations

• Yan’s setup, 12 hours into the ceasefire: “positioning is favorable to be long right now” precisely because most people were sidelined — but his base case is a breakdown “maybe before the 14-day window or at the end of it,” since the two sides’ requests are “diametrically opposed” and in places “completely mutually exclusive.” Delay itself is bullish: unwinding the backlog of fertilizer, oil and other shipments buys everyone time, and a less-squeezed global economy gives the US more leverage.
• He adds the geopolitical whispers as caveats, not claims: China possibly pushing Iran toward resolution, rumors of BTC or yuan paying tanker tolls — and the real question is whether there’s “enough follow-through to make these markets move.”
• Jason’s caution, worth keeping in full: relief that “power plant day and bridge day” didn’t happen, but nothing is resolved — “unless you’re actively trading and really good at it… it probably just pays to be cautious and either play defense or just not play the game.” Markets are back to the levels of Scott’s recession report from two-three weeks ago, “a few percent off all-time highs even though things are as bad as they’ve ever been” — he expects things to break down further before improving.

2. The book around the desk: cash, HYPE, and buying the Trump-speech puke

• Santiago, disclaiming “I’m always just long assets,” actually runs the tightest book: only Bitcoin and HYPE held long-term, otherwise “very very cashed up,” fading headlines tactically, and untouched equities — “I think equities probably come down.”
• Jason mentions a small Zcash position, up 30% in the last day. The research lead stayed mostly long equities with hedges from the start of “Operation Epic Fury” roughly breakeven; his crypto is HYPE, Bitcoin, Zcash, ENA, PUMP and possibly another position, with mostly Bitcoin buying over the past month and recent nibbling at alts.
• Santiago loaded up after Trump’s speech because the deadline itself was information: “you don’t throw that out with zero intention” of working around it. He owns BTC, HYPE and ENA, expects HYPE to benefit as RWA becomes a larger share of its L1, and sees ENA as “insanely under-owned, super beaten down” now that USDe supply has hit a floor and the team is adding prime-brokerage and non-Treasury yield beyond the carry trade — “a biz dev plus ability to find creative opportunities game” that Guy’s team has proven it can win.
• On AI coins, bear-market attrition is the filter: the teams still building — NEAR, Grass, Venice — become clearer bets. Venice looks most interesting on size relative to growth, and Anthropic’s move pushing OpenClaw users off subscriptions toward the API should drive people toward more frontier models, since “paying for APIs is pretty cost prohibitive for a lot of people.”

3. Iran tolls in Bitcoin: multipolar fuel or just a headline?

• The research lead’s macro thesis: “the world’s more multipolar than ever,” Iran illustrated it — countries flying out to make their own deals, China still shipping through the Strait of Hormuz, Europe dealing with China directly — and Iran charging tolls in Bitcoin or yuan is “one of the biggest fundamental developments in Bitcoin in a while.”
• The room’s immediate deflation — doubt that “any real dollars” will use Bitcoin, and the research lead’s concession that it’s “more of a headline thing” — is followed by Yan calling it a good headline and saying this is what Bitcoin needs to fulfill its role as a global, neutral store of value, with quantum fears and “Michael Saylor fears” as his own hedges.
• Yan sees a healthier tape than the DCA-down-only period: TAO, Monad, Zcash on the month and HYPE are working, while SOL is noticeably weak — the Drift hack didn’t help, and he floats that Kyle Samani’s departure from Multicoin may mean LP redemptions force sales of a position that large.

4. The AI IPO ticking time bomb: SpaceX at $1.7T meets finite dry powder

• Santiago’s two-path frame: SpaceX, Anthropic and OpenAI go public into “only so much dry powder.” Either revenue keeps ramping, enabling cheap borrowing and stock sales so “the flywheel keeps going” — or a downturn plus massive insider supply (“it’s all institutions that are just going to sell”) breaks the funding loop for growth.
• The private/public mark gap in one comp: Databricks and Snowflake, both around $5B ARR, with one trading near $150B and the other near $50B — alongside Chime, SoFi and others where private marks can be far above public-market levels. Santiago says “it’s kind of like what happens in crypto… the private marks are just way higher than what can be sustained on public markets and it just takes a while to adjust.”
• SpaceX math from the room: Santiago thought roughly $12B had been raised and estimated Elon still owned 42%; revenue was variously put at roughly $15-20B. Jason cited a last round around $1T and a planned $1.7T listing, while Santiago rounded the proposed valuation to $2T — versus Meta at $1.5T on $200B revenue, “literally trading at 100x revenue… kind of silly.” Jason sees a high probability SpaceX trades below its IPO valuation; Santiago’s blunt line was “the gall to IPO at a two trillion dollar valuation.”
• The exit plumbing being built: pressure on Nasdaq to loosen index-inclusion rules so SpaceX gets in sooner (“so the passive index buyers can buy private investors’ bags”), and the Fundrise Innovation Fund (VCX) — NAV around $19, holding Anthropic and SpaceX shares — trading up to $550-560 as proof of pent-up retail demand. Santiago’s framing of the divide: public markets ask for ROI today, while private rounds price future growth; public-market investors are “guilty until proven innocent.”

5. Santiago’s whiplash — and the case for active management’s comeback

• Santiago’s change of mind, stated as method: “I’ve never changed my mind so often as I have recently… if you change your mind, you have a chance to be right, but if you don’t change your mind, you’re almost definitely wrong based on how fast things are moving.” He was bullish verticalized AI software — context pipelines plus a good UI — but now asks: “if agents get good enough, I don’t even know if we have UIs… maybe you just say what you want and it just happens.”
• His one conviction: “there’s more alpha than ever in public markets.” Two decades of “nothing ever happens” — the Magnificent 7 carrying the S&P and passive winning — is changing; SaaS names were dumped 50-60% indiscriminately, and energy, memory-stock and AI-chip electrification theses are multiplying. “A more interesting game than it ever was” is also “really hard to play part-time.”
• A tangle over who coattails whom: Santiago argues more active capital can make passive even better, while Jason argues active managers coattail on the large passive flows. The synthesis: most people should just own the index, but good active managers may do very well over the next 12-24 months versus the past 5-10 — “now’s a better time for active managers” than when Buffett made his 2000 bet.
• Jason’s add: the mega-IPOs will be funded largely by rotation, not fresh capital — institutional managers rebalancing sector weights means some names get decimated and “really big volatility around those IPOs,” which is exactly the dispersion a long/short manager can harvest.

6. Pre-IPO stocks on-chain: $500k of Anthropic on Solana

• Jason highlighted that someone had bought roughly half a million dollars of Anthropic exposure on Solana through PreStocks; he had been skeptical of these instruments as synthetics or cash forwards. Santiago said it was by PreStocks, that people he respected viewed it as legitimate, and that he was surprised by the size. “There’s going to be some way to bring private markets on chain… I just don’t know what the winning design is.”
• Jason’s design note traces the interesting structure to FTX’s original Coinbase/Robinhood pre-listings via a Swiss law firm — a trust holding the asset, cash-settled on IPO or exit. Meanwhile the real-world plumbing is “Russian dolls”: SPVs on cap tables of SPVs, “which hopefully the underlying is somewhere in one of those,” with carry and broker fees making even secondaries brutal.
• Santiago’s cautionary math: a friend in Anthropic’s $18M round recently sold and, after a roughly 50% dilution mega-round and all the fees, netted “like an 8X.” Prediction for SpaceX: “a lot of people are going to realize that these things did not have access to the underlying” — and if IPOs go south, “everyone who’s in late-stage privates is like, how do we run for the door?”

7. Prediction markets resolve on technicalities; Polymarket internalizes the float

• Santiago, crediting a Matt Levine column, says the boots-on-the-ground market resolving yes off a pilot rescue shows how markets can settle on “some outlier technicality” divorced from their true intent, rewarding nuance-hunters while undermining the product’s stated purpose — “if their whole selling feature is that they make information more accurate… you want markets that are measured by their true intention.”
• Jason takes the other side: remove subjectivity entirely, treat literal resolution “as a feature, not a bug,” and iterate toward tighter trigger conditions — while conceding the UMA dependency “has been less than honest sometimes.”
• Jason generalizes from Polymarket’s native USD: stablecoin float is crypto’s real fundamental revenue, and Solana hosting roughly $15B of stables, much of it USDC, gets “nothing for that” from an issuer that “doesn’t even freeze 200 million when there’s an exploit.”

8. The every-app-a-stablecoin endgame — and Sky’s $12B supply

• Santiago resurfaces the Nic Carter argument: the endgame isn’t a few giant stablecoins but many — float is too valuable to give away, white-label issuance is easy now, front ends abstract the ticker into “just USD,” and cheap interoperability reduces the network effect. Jason’s clean bucketing: yield-sharing stables versus “lubricant for the app” — lubricant gets internalized, so USDC’s integration-dependent model gets hurt while “Tether might be insulated” as the transactional rail.
• Santiago flags Sky’s USDS supply at roughly $12B, with a 3.75% base yield and higher yields when staked. He says its existing Apollo credit-fund and other integrations appear to have provided better bear-market yield sources and absorbed much of Ethena’s supply. Jason separately notes roughly $4.4B of DAI, putting the combined figure around $16B, though they are unsure whether Maker can still be called decentralized.
• The structural moat Santiago confesses to embodying: “forgotten supply that’s just willing to accept lower yield — and I’m one of them, honestly” — legacy float that never leaves can make an incumbent stablecoin’s staker yield effectively leveraged. Jason says that if Maker’s changes raise yield by 100 basis points, they could draw supply back from Sky.

9. Drift’s lesson: North Korea beat the multi-sig, not simply the code

• Santiago’s reconstruction: the Drift exploit was “a six-month in-person operation” — North Korea hired external quant traders who befriended the team, ran a depositor vault on Drift, then got the team to download a TestFlight app. Social engineering in Drift, Bybit and Radiant has become at least as salient as smart-contract bugs because battle-tested DeFi code has improved — though Anthropic’s vulnerability-finding model may reopen that front, especially for long-standing honeypots such as the possibly affected Balancer V2 contracts.
• The operational failure and the standard Santiago demands: Drift thought hardware wallets meant cold keys, but signing on a daily machine means “your machine is hot” — on a $200M-plus TVL multisig, no dedicated signing machine “is just lazy.” He withdrew everything from admin-keyed Solana DeFi; his internal Delphi poll — “do you have 25% or more of your crypto portfolio in DeFi?” — got ten straight nos from industry veterans, though Jason wonders how much is risk aversion versus simply “lack of compelling things to do.”
• Santiago’s category correction: “DeFi has kind of been hijacked” — anything with an admin key is explicitly not DeFi, and an admin-keyed protocol is “the worst of both worlds”: smart-contract risk, centralized key-management risk and multisig slowness. Neither Resolv nor Drift were DeFi hacks, but the brand damage spills over anyway. Santiago also questions treating HYPE as DeFi; Jason agrees that key management has always been the biggest risk, says he has to trust the team, and mentions an unverified possibility of a bridge migration.
• HIP-4 is in testnet with prediction markets and unified spot-margin — Hyperliquid’s first move beyond perps and a possible share-grab from Polymarket. Santiago says he was shilled into buying Lighter while in China; Jason is bullish, seeing a licensing path, RWA space and a potentially useful ZK eject button, though he is not sure the eject button currently works. The room’s own caveat lands hardest: “when has buying the number two in crypto that looks cheap ever worked out?”