🇺🇸 Happy 4th of July everyone! 🇺🇸 It’s amazing to be a part of a great nation where freedom of expression is protected and be on X where we can share our opinions freely. This week we went through 24 podcasts. What stood out: •Enterprise AI trust - a first-person account of frontier-lab data harvesting, and the sovereignty product that shipped the same week •Anthropic - a $3T public-market call, 80% token margins, and a customer trust problem, all in one week •Memory - the numbers behind the supercycle keep going up •Positioning - "no bears left," and the flow data that backs it •Robotics - the mini ChatGPT moment for physical AI •Power - a 55-day data center delivery and the first commercial fusion PPA •Crypto - the case for the August-October window •One consumer AI company quietly out-collecting OpenAI on audio data 🧵 1/18

Meta spent billions on AI talent. Then announced it would become a compute company. Jordy Visser @jvisserlabs: "the silliest thing I may have ever heard." Visser at 22V Research: no finished models, data centers incomplete, strategy shifting. Spending billions on talent and then renting out compute like Azure is not a pivot - it's a pivot away from having nothing to show. His reading of the cheap data center sign-up: a cash flow move to put a floor on $META stock. Not strategic clarity. Not multiple expansion. Zuckerberg doesn't have anything finished right now. Contrast with OpenAI: compute-sufficient, enterprise adoption exploding, ROI story maturing. The gap between where these two companies sit in the AI stack is widening, not closing. Ed Zitron @edzitron has argued throughout 2026 that AI company pivots driven by investor pressure rather than product conviction are a red flag - companies that can't articulate what their AI does pivot to infrastructure instead. Meta's compute announcement fits that pattern exactly. Implication: $META is not an AI infrastructure play. It's a consumer platform buying compute optionality with no coherent model strategy. The market has not fully priced the execution gap. Full breakdown of the Meta vs OpenAI AI stack divergence: podcastalpha.substack.com/p/… Source: Anthony Pompliano Podcast - youtube.com/watch?v=Kp9rfw1I…

"Look, it's rough. Rough is the only way to describe it." Todd Sohn on bitcoin. Sohn, Strategas chief chartist, told @RealEismanPlaybook that money is leaving both gold and bitcoin ETFs - the two assets meant to hedge exactly this environment. Both are failing at the same time. Neither has a catalyst in sight to reverse it. If your downside protection is gold or $BTC, right now you have neither. Source: The Real Eisman Playbook - youtube.com/watch?v=mQXXlrL1…

Cathie Wood @CathieDWood of @ARKInvest: oil is in the negative 50s year-over-year on a three-month moving average. She expects it to go structurally lower. Two headwinds are converging. The Hormuz premium is fading as the Iran War recedes. Electric transportation - cars, trucks, and drones - is eroding demand from multiple directions at once. Wood calls it "a general deflationary world for oil in the next few years." If oil prices drop further, headline CPI falls, TrueFlation goes closer to zero, and the Fed has more room to cut. The oil bear case feeds directly into the rate-cut thesis. Elon Musk @elonmusk has made the same structural case repeatedly - EV adoption is accelerating faster than oil demand models price, and the long-term demand destruction is permanent, not cyclical. Oil, inflation, and the EV structural thesis: podcastalpha.substack.com/p/… Source: ARK Invest In the Know - youtube.com/watch?v=V5W2rJ2f…

Andrew Kang: humanoids productize physical labor the way LLMs productize knowledge work. If that analogy holds, the addressable market is not a niche. It is the $50 trillion global physical labor economy. That reframes the screening question. Stop asking "is this a real product." Start asking "which company captures the most labor." Goldman already moved its TAM 6x in the same direction. The full framework: podcastalpha.substack.com/p/… Source: Live from the Compound - youtube.com/watch?v=1Tgw82x4…

Why does Kalshi hold 90% of the US prediction market? It is not brand. It is not UI. Tarek Mansour @mansourtarek_ explains to Raoul Pal @RaoulGMI: adding a trader to a prediction market has second and third order effects that do not exist on normal platforms. At Airbnb, adding a host gives you one more listing. At Kalshi, adding a trader tightens spreads, which attracts better traders, which improves price accuracy, which attracts more capital. The loop compounds. At scale, Kalshi can direct its community at any new question and get calibrated probabilities within hours. Near-zero marginal cost to open a new market. Ed Zitron @edzitron has pushed the other way: dominant market share in fintech often looks structural until it does not. First-mover incumbency is not the same as a compounding moat. That tension is the real watch item. Kalshi's 90% could be durable - or it could be temporary incumbency from being first and regulated. The market share thesis and its counterarguments: podcastalpha.substack.com/p/… Source: Raoul Pal The Journeyman - youtube.com/watch?v=DTy3MbQs…

Cathie Wood @CathieDWood of @ARKInvest: savings rate below 3.2%. Subprime auto delinquencies at all-time highs. And for the first time, auto delinquency rates are running higher than credit card delinquency rates. This consumer stress is real. Food and energy inflation forced lower-income households to draw down savings to cover basics. Auto repossessions are rising because Uber and Lyft make car ownership less necessary - so consumers protect credit cards and let the car go. Aggregate consumption is holding, carried by higher-income spending. That bifurcation is the key variable: it keeps top-line GDP afloat while the bottom of the income distribution deteriorates. Steve Eisman @realsteveeisman has been consistently more bearish - his read is that bifurcated consumer health understates the deterioration that eventually pulls aggregate demand down across the income distribution. Full consumer stress analysis and what it means for discretionary positioning: podcastalpha.substack.com/p/… Source: ARK Invest In the Know - youtube.com/watch?v=V5W2rJ2f…

Someone built MicroStrategy for robots. Andrew Kang launched RoboStrategy ($BOT) on NASDAQ in May 2026. The structure copies Michael Saylor's Bitcoin playbook, with private robotics equity as the underlying NAV. Issue shares at a premium to NAV. Use the proceeds to buy more private robot equity. Raise per-share exposure over time. It gives retail a listed way into companies like Figure AI that otherwise stay private to multi-trillion valuations. How the flywheel works and where it breaks: podcastalpha.substack.com/p/… Source: Live from the Compound - youtube.com/watch?v=1Tgw82x4…

"Design is dead" is the wrong read, says Dylan Field @zoink. Figma's CEO told Molly O'Shea @MollySOShea it is a temporary phase of AI excitement, where each role thinks it no longer needs the others - engineers skip designers, designers go it alone. Then they hit the limits of solo AI work. His fix is teams. If you want scalable systems and the right build, not just a fast one, design empathy and judgment still matter - and those do not improve just by adding IQ points to a model. Tobi Lutke @tobi pushes the other way at Shopify - prove you cannot do the job with AI before you add headcount. That is the live tension over how lean AI-era teams get. For Figma holders: the "design is dead" bear case is premature, and design's scope is widening, not shrinking. Full thesis: podcastalpha.substack.com/p/… Source: Sourcery with Molly O'Shea - youtube.com/watch?v=3xfMDWEV…

If you hold residential real estate or homebuilder equities, Cathie Wood @CathieDWood of @ARKInvest has one number for you: 5.75%. That is the mortgage rate threshold where housing might begin to move. Prediction markets are not pricing it happening this year. Rates are well above it and trending up. The freeze mechanism: homeowners trading up would face a mortgage at roughly double their current rate. They are staying put. New home prices are falling but not collapsing. Builder inventories were declining and have started rising again. Wood's one scenario for housing to break free: productivity-driven deflation pulls long rates down fast enough - the same dynamic as the Industrial Revolution, when technology deflation thawed asset markets. Bill Ackman @BillAckman has publicly called for the Fed to cut aggressively to free the housing market - arriving at the same conclusion from a very different starting point. The housing lock thesis and what breaks it: podcastalpha.substack.com/p/… Source: ARK Invest In the Know - youtube.com/watch?v=V5W2rJ2f…

Why is Jordy Visser @jvisserlabs more bullish on OpenAI than $META right now? Visser at 22V Research cites Greg Brockman's recent interview on enterprise AI adoption. Six months ago the pitch was: "just get me anything, I need to catch up." Now enterprises are asking OpenAI to show them real ROI. That business is exploding. 10 million users on agentic systems today; Brockman expects billions. H2 2026 is his declared game-changer for AI agents specifically. OpenAI is compute-sufficient. Meta is not. Meta spent billions on AI talent and then announced it would become a compute company with no finished models and incomplete data centers. Visser calls the comparison stark. Brad Gerstner @altcap at Altimeter Capital has been one of the most vocal OpenAI bulls, arguing the enterprise transition from "spend anything" to "show me ROI" is the phase that produces durable margin improvements. Visser adds a warning: once the Anthropic and OpenAI stakes sitting in corporate balance sheets get marked to market at IPO, the earnings tailwind disappears. Implication: OpenAI's enterprise flywheel is real. Know which part of the AI stack you actually own - and whether the earnings quality underneath it survives the private-to-public transition. Full breakdown of the enterprise adoption shift and AI stack divergence: podcastalpha.substack.com/p/… Source: Anthony Pompliano Podcast - youtube.com/watch?v=Kp9rfw1I…

If you hold a position in active management stocks, this signal is worth your time. Tarek Mansour @mansourtarek_ on Real Vision: Kalshi's best inflation forecaster is a random person in Kansas. Not a Goldman economist. Their track record is public, verifiable, and already attracting outside capital. The mechanism: prediction markets financially punish partisan bias. The extreme take that gets retweets loses money. The boring, calibrated view wins. Over time, the best performers emerge not by credential but by calibration - and their records are on-chain. Chamath Palihapitiya @chamath has argued the same structural point on All-In: the institutional edge in finance was always about access, not alpha. Prediction markets close the access gap. The best Kalshi traders are already raising external capital. The track records are the pedigree. If this plays out, the manager selection business - prime brokers, seeding platforms, allocators - has a new competitor it did not see coming. The Kansas forecaster and what verified prediction market track records mean for institutional allocation: podcastalpha.substack.com/p/… Source: Raoul Pal The Journeyman - youtube.com/watch?v=DTy3MbQs…

Why does Figma's CEO want to rewind to 1970s computer research? Dylan Field @zoink told Molly O'Shea @MollySOShea the industry needs to return to the HCI work of the Alan Kay era - fundamental questions about how humans use a computer that we stopped asking. His evidence that we stalled: pull to refresh is still the last new interaction model most people agree on. The tablet is "just a big iPhone," VR is the only place inventing new paradigms, and AI opens the widest interface design space in decades. Brian Chesky @bchesky is making a parallel bet at Airbnb - reimagining the interface from first principles rather than bolting AI onto the old one. For long-horizon investors: the platform that invents the next interaction model, not just the next feature, captures the prize. Full write-up: podcastalpha.substack.com/p/… Source: Sourcery with Molly O'Shea - youtube.com/watch?v=3xfMDWEV…

$MU: sub-$60 to $1,240 in 15 months. Blowout earnings. Stock sold off. Jordy Visser @jvisserlabs says consolidation is the trade, not the exit. Visser at 22V Research: the 50-day moving average sits around $840 vs the high of $1,240. A 50% range from the 50-day. Wide-range consolidation resolves one of two ways - price falls to the 50-day, or time passes through the next earnings cycle in July. Either path leads to the same place: Q4 2026 new all-time highs. The sell-off on blowout earnings is a signal. When great news can't hold the price, the market is already pricing the next 6 months, not the last quarter. Tom Lee @fundstrat has stayed bullish $MU throughout and argued earnings beats are the correct signal to stay long. Visser's refinement: the sell-the-news reaction tells you multiple compression has begun even if the fundamental thesis is intact. These aren't the same trade, and they require different sizing and time horizons. Implication: Micron bulls should expect chop in the $840-$1,240 range for the next 3 months. Add on dips toward the 50-day. Exit the momentum playbook. Stay in the thesis. Full technical setup and Q4 call: podcastalpha.substack.com/p/… Source: Anthony Pompliano Podcast - youtube.com/watch?v=Kp9rfw1I…

If you hold a large single-stock tech position, a new force is sitting under it. Todd Sohn, Strategas chief chartist, told @RealEismanPlaybook that leveraged single-stock ETFs have grown to $200B. These funds rebalance daily to hold their leverage ratio. On a down day they must sell more of the same name into the fall. On concentrated tech, that is a built-in accelerant. This did not exist in prior cycles. Model a faster, deeper drawdown. Source: The Real Eisman Playbook - youtube.com/watch?v=mQXXlrL1…

Ev Randle on Sourcery: how do you value a business whose margin falls as usage explodes? Randle, a Benchmark GP, argues this is the AI-era diligence question. Quantity is flat - the same buyers. Margin is lower because inference costs money. Price has exploded - inference platforms sign nine-figure contracts with startups, a scale SaaS almost never saw. High margins now mean low usage. Rebuild your screen around P x Q x M, not rule of 40. The framework in full, via @MollySOShea: podcastalpha.substack.com/p/… Source: Sourcery with Molly O'Shea - youtube.com/watch?v=O6Dahcxw…

Nasdaq peaked March 2000. Value held 8-9 months. The S&P did not peak until September. David Rosenberg @EconguyRosie on @RiskReversal says that same sequence is playing out now. Tech leads down while the cap-weighted index still looks stable, because leadership breaks before the average does. Dan Nathan adds the Cisco tell: its earnings miss and guide-down was an early shot across the bow before the broader tech sell-off. A guidance cut from Nvidia, Amazon, or Microsoft would be the modern equivalent. Howard Marks @HowardMarksBook makes the cycle case the same way. You cannot time the top, but you can read where you sit in the sequence. If you index, a calm S&P is not all-clear. Watch the leaders, not the average. The full 2000 sequencing map: podcastalpha.substack.com/p/… Source: RiskReversal Podcast - youtube.com/watch?v=KMBKsnDB…

Dylan Field @zoink's test for which AI monopolist actually lasts. Figma's CEO gave Molly O'Shea @MollySOShea a rule he traces to Tim O'Reilly: create more value than you capture. That is his definition of a platform - the value created outside it must exceed what it takes. The AI labs some expect to "suck it all up" fail that test the moment capture outruns the value they create for everyone else. Satya Nadella @satyanadella has run Microsoft on the same idea - Windows won by making its ecosystem richer than the platform itself. For investors sizing AI platform bets: the durable winners widen the pie, not narrow it. Full thesis: podcastalpha.substack.com/p/… Source: Sourcery with Molly O'Shea - youtube.com/watch?v=3xfMDWEV…

Cathie Wood @CathieDWood on @ARKInvest: the Fed just set up a task force to cross-check official government statistics against private data. She calls this the biggest institutional signal of the year. Kevin Warsh has established five reform task forces. One covers data sources. One covers productivity and jobs during the technology transformation. One covers the inflation framework. Wood reads all three as Warsh signaling the Fed knows it is flying on distorted instruments. The practical implication: private measures like TrueFlation may become official Fed cross-checks, which would validate ARK's case that current policy is overtightened. Adam Taggart @adamtaggart has consistently argued the Fed is using lagging and distorted data and needs exactly this kind of structural data reform - support from a very different starting point. The Warsh thesis and what five task forces mean for rate-cut timing: podcastalpha.substack.com/p/… Source: ARK Invest In the Know - youtube.com/watch?v=V5W2rJ2f…