Hacking Just Got 1000x Cheaper

Summary

• The discussion estimates cyberattack costs have fallen “probably 100, maybe 1,000 fold,” while defenders may spend 10–1,000 times what attackers spend. Compromised trust infrastructure can turn an old device into a starting point for further attacks. The exchange argues that better defensive models do not remove the attacker’s advantage: finding one entry point can suffice. One panelist calls the resulting demand “very bullish compute.”
• Anthropic’s S-1 eases near-term financing fears without resolving the long-term economics of frontier models. Ceteris highlights cash reserves and no customer exceeding 10% of revenue; open-source commoditization remains unresolved. The compute trade still depends on demand absorbing efficiency gains: a panelist warns that Jevons’ paradox taking five years would still be a problem.
• Jake expects SpaceX and Elon to “clean up this market in the next 12 months,” particularly through stranded compute and weaker operators. Execution speed, diversified revenues, and financing advantages underpin the call. Another panelist points to the cloud business’s short-notice contracts and spot-price exposure as substantial downside risks.
• Hypercall’s options thesis rests on making underwriting and hedging cheaper through Hyperliquid’s existing liquidity. Jake reports roughly $250 million of associated perp volume over 30 days and estimates equivalent Binance hedging would cost 10–15 times more under generous fee assumptions. Around-the-clock trading, fractional contracts, and composable collateral are the customer proposition; retail distribution remains the execution problem.
• Bitcoin’s quantum tail risk produces the episode’s sharpest disagreement, particularly over whether a fork could preserve value after compromised coins move. José argues holders would choose the fork that protects value. Other panelists warn that a fork or compromised coins could be damaging; one estimates roughly 90% downside. Jake expects quantum cracking eventually and stresses that compromised Satoshi keys would imply a much wider migration problem.
• Large IPOs and AI financing could drain public-market liquidity while recycling exit proceeds into venture. A panelist sees exceptional conditions for agile early-stage funds, supported by mega-fund demand and scarce investable companies. Yan warns that weak Anthropic trading could reset private marks quickly, while Kevin wants retention evidence behind startups reaching $100 million ARR before they have meaningful cohort histories.
• The crypto stance is patient rather than aggressively leveraged, with Bitcoin around $80,000 and a possible retest of the $70,000s still inside the bullish framework. Jason holds previously raised cash because “if I don’t have a read, I don’t play the game.” Ceteris remains constructive on 2027, while Yan sees on-chain financial activity and debasement—not a single treasury buyer—as the cycle’s underlying drivers.

Deep dive

1. Cash remains a position while crypto lacks follow-through

• Jason expects little surprise from the Fed meeting and admits his claim that the 10:30 seller had disappeared immediately failed. He holds cash raised during the earlier on-chain trading surge because his market read keeps reversing: “If I don’t have a read, I don’t play the game.”

• Ceteris does not think the cycle is over and expects 2027 to be good, but sees “no follow-through.” Bitcoin is around $80,000, Zcash between $1,100 and $1,300, and HYPE around $80. Midterms the following Tuesday might change conditions; meanwhile, patience beats pushing leverage on the first breakout. He acknowledges possible complacency.

• Yan expects chop followed by a stronger year-end, though Bitcoin could retest the $70,000s from roughly $80,500. His cycle thesis is TradFi activity moving on-chain through RWAs, stocks, and perp DEXs such as Hyperliquid and Lighter, potentially with institutions providing front ends. Debasement adds longer-term support; midterms could constrain Trump-related policy risk.

2. Anthropic’s balance sheet buys time for the AI trade

• Ceteris reads Anthropic’s S-1 as materially better than feared: substantial cash extends the runway, and no customer accounts for more than 10% of revenue. José notes that this counters speculation about extreme dependence on Meta or Jane Street. Ceteris thinks removing those near-term concerns helps money return to adjacent AI names.

• The distinction is between working capital and eventual business value. Cash “kicks the bucket down the road massively,” but cannot answer whether open-source competition commoditizes models. That remains unresolved; an S-1 could not have settled it anyway.

• José describes SK Hynix’s quarter as “absurd,” estimating “60 billion, something like that” in earnings. He sees some mechanical downside support when memory stocks trade around three times forward earnings, while acknowledging cyclicality. Neoclouds have also rebounded, but remain “very hard to understand businesses.”

3. Efficiency gains can redistribute value across the AI stack

• Jake questions why the whole stack trades together. He cites a Google paper he thinks was called TurboQuant cutting memory requirements by 80–90%, and SanDisk’s prospective high-bandwidth flash alternative to high-bandwidth memory. Advances across memory and fiber could “really nuke the value out of these neoclouds”; he expects greater dispersion.

• Ceteris sees upside in the roughly $500 billion of compute commitments being discussed because much of that capacity was secured below spot prices, which he says rose about 44% this year. José identifies the correlated downside: take-or-pay obligations leave unused capacity precisely when resale prices collapse. He cites telecoms in 2001 and SemiAnalysis’s 50-gigawatt forecast for the following year.

• Jake asks whether a sudden tenfold reduction in training and inference costs could expose finite demand, as cheaper blockspace did in crypto. José thinks AI demand is more real, but warns that Jevons’ paradox taking five years would still be a problem. Over half the Fortune 500 reportedly spend above $100,000 annually with Anthropic, yet that represents only 0.1% of the Fortune 500 companies’ overall expenses.

• José describes the frontier-lab CFO’s trap: underbuy compute and competitors undercut or out-train you; overbuy and a downturn can blow you up. Yan expects cheaper compute to deliver multiple expansion for beneficiaries outside AI. He cannot determine how much of Microsoft’s cited 60K layoffs reflects AI and rejects a neat layoffs-equal-spending explanation.

4. Cheap attacks compound through compromised trust

• Jake calls the PHP issue “kind of a nothing burger,” since hosting providers or Cloudflare can intercept many attacks. The scarier issue concerns an intermediary certificate provider whose name he tentatively gives as eMudra. Older devices retaining compromised certificate trust could accept impersonated services—for example, a gas meter receiving a malicious software update and becoming permanently difficult to patch.

• The discussion estimates attack costs are down “probably 100, maybe 1,000 fold,” and describes how one compromise can seed another: vulnerable devices become botnets and create further entry points. A panelist speculates that Railway’s problems might connect to the certificate compromise, explicitly noting the absence of a postmortem. Its GitHub build connections illustrate how stolen credentials could propagate damage downstream.

• A panelist asks whether defenders’ access to better models and more compute should give them the advantage. Another responds categorically: “No, not at all.” Attackers need one opening, potentially beneath an otherwise secure application. The discussion puts defensive spending at 10–1,000 times the attacker’s expenditure and raises the hope of an eventual equilibrium favoring defense.

• A panelist argues AWS and Google Cloud are safer than smaller hosts on scale grounds and points to national-defense data as potentially richer targets. José counters that cloud-hosted stablecoin minting keys could enable enormous issuance. Another panelist asks how much value an attacker could actually realize, even with access to massive holdings.

5. Bitcoin’s quantum risk divides the panel over recovery

• José says he considers Satoshi’s vulnerable coins less threatening than a few months earlier. He expects any breakthrough to involve a lab tied to the U.S. or China and doubts a government would destroy the holding’s value. Another panelist calls governments historically paper-handed; Yan says a Democratic administration “would just dump this to zero.” José replies: “No, they wouldn’t.”

• José thinks holders would favor a fork removing stolen coins over one allowing an attacker to dump them. Other panelists warn that a fork itself would be damaging; one estimates Bitcoin could fall roughly 90% in the adverse scenario. Jake adds that even burning compromised coins would be bearish because the underlying message is that someone accessed supposedly secure keys.

• Jake expects quantum computers eventually to crack Bitcoin and stresses that cracking Satoshi’s keys would mean other keys also need migration. Ceteris estimates moving everyone’s UTXOs could consume half a year to a year of block space, while new signature schemes introduce risks. José questions the appeal of governance that is extremely slow to change; Yan distinguishes immutability from security.

• Jake offers an explicitly unlikely, non-quantum scenario discussed with core developers around 2019–2020: an early Windows XP random-number configuration might expose one experimental key, allowing 50 BTC to move without compromising everything else. “I don’t think this happens,” he emphasizes. The point is that a nonfatal technical event could initially look fatal; the others expect severe market panic.

6. Synapse’s bridge economics explain its options pivot

• Jake says “bridging was dead” economically by 2022–2023 as Across started getting traction: margins fell roughly 100-fold in a year. On some chains, a million messages generated about $1,000. Demand was not expanding enough to offset that compression. Interoperability work—such as using bridged GLP as lending collateral—suggested a different application: improving options collateral efficiency.

• A perp market maker can operate with leverage and an ADL backstop; an option writer might post 10–100 times the premium received. Jake calls full collateralization “crazy.” Earlier Dopex work with GMX showed the concept, but fees and liquidity were inadequate. Hyperliquid supplied the missing liquid on-chain hedging venue and made building the product compelling.

• Keeping SYN reflects Jake’s preference for founders pivoting with their existing investors, rather than repeatedly launching tokens. He compares startups raising $5–10 million and still pivoting with one or two months of cash left. Hyperliquid also inspired him by refusing familiar crypto games and winning anyway. Ceteris subsequently discloses that he recently bought SYN.

7. On-chain options compete on access and capital efficiency

• A host asks what remains distinctive if Robinhood offers comparable assets. Jake identifies 24/7 trading, particularly when Middle East events drive volatility outside market hours, and fractional contracts. A one- or two-week at-the-money S&P 500 call can require a ticket of several thousand dollars; fractional options could keep even year-out contracts accessible for a few dollars.

• Jake says Hyperliquid delta-one positions can offset option risk, while purchased volatility can also support margin. Vaults add options to existing spot and perp strategies, enabling covered calls, basis trades, and more aggressive structures. His composability example: deposit Nvidia, write a covered call, collect premium, then borrow against the position and loop it. Yan also sees vaults supplying structurally needed volatility sellers.

• Hypercall generated roughly $250 million in Hyperliquid perp volume over the preceding 30 days, mostly taker flow from delta hedging, Jake says. Under generous Binance fee assumptions, his team calculated equivalent hedging would cost 10–15 times more there. Competing options venues therefore need competitive perp execution too—a harder undertaking than simply shipping options infrastructure.

8. Retail distribution and perp partnerships shape adoption

• Jake expects short-dated retail options first, with flows eventually supporting longer maturities; RWAs look like the clearest opportunity. “No one trades options” strikes him as “one of the dumbest takes,” given Robinhood usage and India’s daily notional. He anticipates a HIP-3-like trajectory: months of flat activity followed by a liquidity flywheel.

• A host presses on distribution: working infrastructure does not automatically attract traders. Jake’s practical answer is user-generated TikTok content, citing FOMO, plus builder codes and wallet integrations. His bridge background informs the warning: bridging looks seamless in quiet markets, then liquidity disappears when a price move makes everyone want to trade simultaneously.

• Jake sees Robinhood as a closer competitor than another crypto options protocol, with Polymarket or Kalshi potentially entering through a DCM. Structural difficulties remain, but they have substantial dry powder. He expects an oligopoly of two or three major players, each partnered closely with a major perp DEX, because managing both products together is difficult.

• Outside options, Jake favors new lending models. He mentions 3Jane’s loans against Binance accounts, while noting he has not reviewed it for several months. Non-overcollateralized credit remains underexplored; crypto-card providers can build repayment and spending profiles. “They’re credit card companies, right? They should extend credit.”

9. IPO supply could tighten public-market liquidity

• A panelist expects potentially the largest public-market liquidity drain in history over the next 12 months. Yan adds AI capex and tightening debt access, citing Paul Tudor Jones’s caution about 2027 despite a constructive year-end view. Kevin thinks markets can initially absorb Anthropic, but worries about later issuance turning equity supply positive after roughly 10–15 years of contraction.

• Yan would avoid Anthropic initially because other investments could benefit if it succeeds. Depending on pricing around a possible $2 trillion valuation, Ceteris expects a better entry after listing, potentially resembling SpaceX’s initial drawdown. Staggered vesting and unlocks matter. The balance-sheet reassurance does not make the opening public-market price automatically attractive.

10. Venture’s financing advantage still depends on exits and retention

• José thinks early-stage private markets have “never been in a better spot.” Mega-fund demand resembles index flows into scarce qualifying companies, amplified by “low float, high FDV”: companies sell perhaps 10% each round, while existing holders cannot freely sell. Few businesses can absorb nine-figure checks. He favors a small, agile early-stage fund over roughly five years.

• Ceteris finds private markets “spooky.” Yan warns that weak Anthropic trading would establish a lower clearing price, remove markups, and tighten funding throughout the stack, as in the previous venture cycle. José acknowledges that Instinct at $30 billion looks extreme. Yan’s qualification is blunt: “It works until it doesn’t,” and the reversal can travel downward quickly.

• Kevin questions what extraordinary revenue ramps actually prove. Startups now reach $100 million ARR so quickly that some lack usable retention cohorts. What does Instinct’s retention look like 12 months later? Fast growth followed by churn in intensely competitive AI products could force more disciplined underwriting and spread caution through private markets.

11. SpaceX’s operating advantage comes with spot-price exposure

• Jake expects consolidation among stranded compute and weaker neoclouds in 2027. The premium shifts from promised data centers to installed, working capacity earning revenue now. He cites companies paying substantially higher prices for compute secured into 2027 and says SpaceX and Elon are positioned to “clean up this market in the next 12 months.”

• Yan emphasizes financing: diversified revenues, scale, and lenders’ desire for future business can secure better terms. Kevin expects acquisitions of smaller operators whose delays and financing costs have undermined their economics. Their assets may retain value if a capable operator brings them online. Yan also expects TerraFab’s scale to support strong margins.

• José pushes back on valuation and downside. In his understanding, cloud contracts can be canceled by either side with three months’ notice, leaving substantial spot-price exposure. He estimates pricing near $50 per watt versus CoreWeave’s $10–15. SpaceX is therefore a leveraged AI bet that could suffer badly if compute prices fall; Yan thinks diversification improves its recovery prospects.

• A panelist raises Kessler Syndrome as a speculative tail risk: satellite collisions create debris that causes further collisions, potentially rendering low orbit unusable. He connects it to reported military satellite activity and a tweet about buying SpaceX puts. The concern extends beyond one stock to communications infrastructure; he sees little portfolio protection against such a scenario.

12. Familiarity remains the panel’s strongest allocation constraint

• José wants better exposure to power, grid construction, and infrastructure, already holding the GRID ETF. Commodities might become interesting over the next few years, but he would prefer specialist expertise. Yan thinks commodities call for active trading around events. José ultimately admits little appetite for distributing money across numerous managers.

• A panelist and Ceteris prefer staying close to crypto, where sustained attention occasionally reveals an unusually obvious opportunity. “I just wish I had more of everything,” Ceteris says. Jake counters that paying less attention sometimes worked better this cycle; group chats still surface opportunities for him. The disagreement is over attention, as much as allocation.