Jordi Visser @jvisserlabs: he used Waze for two or three years before switching to Google Maps. Then Waze had 1,700 clones by the next day. On Anthony Pompliano's @APompliano show, Visser said that gap, years of user loyalty to one app, is gone. New AI models and assistants like Astra, Meta's Muse, and Grok now ship faster than people can evaluate them, so the old rule that the best app wins for years no longer holds. The one thing still hard to copy overnight, Visser said, is a user's accumulated memory and context inside the assistant they already use. If a product's edge is a good model, its lifespan is weeks. If its edge is your history inside it, that lifespan resets to zero the moment you would have to start over somewhere else. Podcast Alpha is tracking who actually owns the memory layer in the personal assistant race, subscribe for the coverage. Source: Anthony Pompliano -

The government did not stop calling recessions by accident. Jordi Visser @jvisserlabs says it is a monopoly. On Anthony Pompliano's @APompliano show, Visser pointed to two straight quarters of negative GDP growth that never got labeled a recession. His read: the people who define recession understand the market cannot absorb one for long, so a slowdown that would have counted a decade ago quietly stops counting. Add a monetary bazooka ready to fire at any real downturn, and Visser does not expect to see an 18-month recession again in his lifetime. Two negative quarters only equals a recession if the people who track it are willing to say so. Podcast Alpha tracks which macro rules still hold and which ones quietly stopped applying, subscribe for the next one. Source: Anthony Pompliano -

Jordi Visser @jvisserlabs: oil hit $150 back in 2008. It is $100 today, and people are still saying oil is going to kill the economy. On Anthony Pompliano's @APompliano show, Visser ran the actual math: household net worth is up roughly 3.5x since 2008, near $195 trillion. Oil relative to disposable income is the lowest it has ever been in US history. The dollar figure people fear did not move much. The size of the economy absorbing it did. A scare headline that does not divide by how much richer the base got is not measuring risk. It is recycling an old one. Podcast Alpha tracks which macro fears actually hold up against the numbers, subscribe for the next one. Source: Anthony Pompliano -

Same AI boom, opposite instincts in Washington and Brussels. On Ben Pouladian's @benitoz episode of BEP Research, host Luke Knight laid out the split live. While the All-In Summit crowd waved off existential AI risk as science fiction, Knight pointed to European Commission President Ursula von der Leyen pushing the opposite instinct, favoring a slowdown. King Charles convened Jensen Huang, Demis Hassabis and other lab leaders in Scotland to talk through cyber controls and containing agent swarms, short of a full stop. The AI buildout timeline now runs on two clocks. Which government issues the permits matters as much as which lab ships next. Source: BEP Research -

Drug discovery used to run on years of physical lab work. Ben Pouladian @benitoz says that timeline is now weeks. Host Ronan Chambers brought up Pouladian's own essay calling the lab the next AI inflection point. Pouladian pointed to Ture Therapeutics, a Los Angeles biotech he's tracked for five years, sequencing billions of molecules and feeding the results into AI models. Work that used to take years of physical lab time compresses into weeks, generating data that otherwise wouldn't exist. Host Luke Knight agreed live that this beats the current chatbot wave. The AI story with the longest runway isn't the one generating text. It's the one generating lab data nobody could produce fast enough before. Source: BEP Research -

Running your own AI infrastructure isn't really an infrastructure decision. It's a data decision, Ben Pouladian @benitoz says. Hosts Luke Knight and Ronan Chambers pressed him on where frontier labs' economics hold up now that NVIDIA owns Hugging Face. Pouladian's framing: rent a frontier model like Opus for the hard problems, and let it orchestrate cheaper open models like Nemotron or Qwen for the routine work you already own. The reason to own that layer isn't the hardware bill. It's that the data stays on your own system instead of training the model company selling it back to you. The infrastructure spend isn't really about hardware. It's about which vendor doesn't get to train on your data and use it against you later. Source: BEP Research -

Ben Pouladian @benitoz has a rule for AI safety debates: the safeguards always come after the product, never before. Same order every time, cars included. Host Ronan Chambers pushed past the pro-acceleration vibe at the All-In Summit, where Jensen Huang's 10 percent doom number got waved off as science fiction, to ask where Pouladian actually lands after Dario Amodei's pacing letter. His answer: cars ran without seatbelts or airbags for years before either existed. Streets and sidewalks came first. The one line he holds: don't ship a product once it's already gone rogue or cheated its way past a test. Accelerationism isn't a claim that AI is safe today. It's a bet that AI gets the same grace period cars got, before anyone had built the airbag. Source: BEP Research -

An advocate does not get to grade whether the thing they advocated for caused a catastrophe. Bill Gurley @bgurley applies that rule to five named COVID experts. At Jason Calacanis's @Jason All-In Summit, Gurley pointed to Fauci's 2011 Washington Post op-ed arguing gain-of-function research's "pros outweigh the cons," then argued that sentence alone, on the record before COVID existed, made real impartiality about a lab origin impossible for him. His comparison: a juror gets struck for far less bias than believing your own past advocacy might have caused 15 million deaths. The disqualifying fact was never what someone concluded after the fact. It's what they were already on record wanting to be true beforehand. Podcast Alpha's newsletter has the full case Gurley builds against all five experts he names, and the document he says undercuts them. Source: All-In Podcast -

Anthony Pompliano @APompliano, on his own show: users of his Sylvia platform hold about 20% of assets in real estate, only 10% in crypto. That's a crypto-leaning audience. Public equities still make up the bulk of the remainder, and Pompliano says the headline that his own audience must be mostly crypto doesn't hold up against the platform's actual numbers. Jay Jacobs ties the same gap to BlackRock's generational-wealth-transfer push, advisors now bridging a boomer parent's large-cap fund questions and a younger client's Bitcoin ETF questions in the same family. Podcast Alpha's coverage checks what a crypto-native audience's own numbers actually say about portfolio share. Full breakdown: Source: Anthony Pompliano -

Richard Feynman cleared six months off his calendar to sit on the Challenger commission. Every other member waited for staff briefings. Bill Gurley @bgurley told Jason Calacanis's @Jason All-In Summit crowd that Feynman showed up at NASA facilities and started interviewing engineers before the commission ever convened in Washington. When Neil Armstrong noticed, he pulled Feynman aside: "We just evaluate. We don't investigate." Feynman kept going anyway and later forced his own findings into the final report as Appendix F. A commission seated to review something and a person who actually goes and checks it can sit at the same table and still not be doing the same job. Podcast Alpha's coverage traces Gurley's searchers-vs-blockers framework from Challenger through to the one document he says nobody in COVID ever forced open. Source: All-In Podcast -

A catchy ETF ticker tells you almost nothing. What's built underneath does. Jay Jacobs, BlackRock's US Head of Equity ETFs, told Anthony Pompliano @APompliano that BIDA, BlackRock's Bitcoin premium income ETF, was deliberately structured as a 33-Act product instead of the standard 40-Act wrapper, the same structure used for Bitcoin and gold ETFs. The tradeoff: a more complex K-1 tax form in exchange for better after-tax efficiency. $BLK Two similarly-named ETFs can carry different legal structures, tax forms, and market-maker support. Podcast Alpha's coverage checks what actually separates a real ETF from a look-alike. Full breakdown: Source: Anthony Pompliano -

How do you investigate a catastrophe 3,000 times bigger than the last five combined? Bill Gurley @bgurley answered with a line from Richard Feynman. At Jason Calacanis's @Jason All-In Summit, Gurley sized COVID-19 at 15 million deaths against roughly 5,000 combined for Surfside, the 737 MAX, Katrina, Fukushima and Challenger. Then he recalled Feynman on the Challenger panel, told once to soften his tone, replying: "But seven people died." Gurley's question back to the room: if seven deaths justified dropping the niceties then, what does 15 million justify now? The size of a catastrophe should set the limit on how much anyone gets to protect an institution's comfort while investigating it. That limit moves with the death count, not with who's uncomfortable answering. Podcast Alpha's newsletter breaks down every number behind Gurley's 3,000x comparison and what it leaves out. Source: All-In Podcast -

Bill Gurley @bgurley: the Army Corps of Engineers built the levees that failed in Hurricane Katrina. It paid an outside engineering society $1 million to help investigate itself. At Jason Calacanis's @Jason All-In Summit, Gurley walked through what happened next. The Corps paid the American Society of Civil Engineers $1 million to join its internal review, and the ASCE team got folded into the Corps' own process instead of staying separate from it. Engineering professor Raymond Seed later wrote a 42-page report to the ASCE cataloguing every place it had blocked the investigation. Paying an independent reviewer does not make the review independent. The check only holds if the reviewer's paycheck comes from somewhere else. Podcast Alpha's coverage of this talk maps the same failure across five separate catastrophes, plus where Gurley applies it to COVID. Source: All-In Podcast -

Jay Jacobs, BlackRock's US Head of Equity ETFs, on Anthony Pompliano's @APompliano show: AI demand moves in days. New copper mines take four to eight years to come online. Indium for the photonics that replace copper in data centers needs its own zinc mine and processing chain. Even an accelerated US chip fab takes about four years. Jacobs calls it the biggest mismatch in AI today, and it's why BlackRock built sleeve ETFs like POWR and iCop around the value chain instead of one ticker. $BLK Podcast Alpha's coverage breaks down which parts of the AI supply chain hold pricing power longest. Full breakdown: Source: Anthony Pompliano -

BlackRock assumed big Bitcoin holders converting into iBit were fleeing custody risk. Jay Jacobs, BlackRock's US Head of Equity ETFs, told Anthony Pompliano @APompliano that's only part of it. The bigger pull: once Bitcoin sits inside the ETF wrapper, a holder can borrow against it, layer options, or rotate into the S&P 500. None of that works from a hardware wallet. In-kind minimums have also dropped to $1.5-2 million, pulling in more holders. $BLK Podcast Alpha tracks how BlackRock's own ETF mechanics are reshaping Bitcoin demand. Full breakdown: Source: Anthony Pompliano -

Why did every major chipmaker except Sam Altman get AI's scale wrong? Cerebras $CBRS CEO Andrew Feldman @andrewdfeldman: TSMC got it wrong. Nvidia $NVDA got it wrong. The memory makers got it wrong. Cerebras @cerebras got it wrong too. Feldman told Uncapped with Jack Altman @jaltma that the initial Stargate number got mocked as "mind-bogglingly large" at the time. Now, he says, it looks "sadly small" next to where AI demand actually landed. Our coverage of this episode tracks what a multi-year pattern of underestimating AI compute demand means for capacity forecasts. Source: Uncapped with Jack Altman -

Nvidia's $NVDA moat isn't CUDA. It's a decade nobody was watching. Cerebras $CBRS CEO Andrew Feldman @andrewdfeldman says the real edge came from roughly 2000 to 2013, when Nvidia traded as an unloved public company that almost nobody believed in. Feldman told Uncapped with Jack Altman @jaltma that grit is still visible in Jensen Huang today, even running the most valuable company in the world. Feldman says he runs Cerebras @cerebras, his fifth startup, with the same mentality: "I'm a professional David in the battle with Goliath. And I wake up every day with that mentality." Our coverage of this episode covers what a culture-based moat means for modeling Nvidia's durability once the founder generation moves on. Source: Uncapped with Jack Altman -

NVIDIA sold GPUs. Now Jason Calacanis @Jason says $NVDA is the frontier model in five domains, and Jensen Huang @JensenHuang did not push back. He points to Alpamayo, a reasoning-based self-driving model that needs far less training data because it reasons about situations it has not seen, instead of pattern-matching billions of hours of road footage. And a suite of biology models, ESM2, ESM Fold, AlphaFold 2, and Proteina Complexa, built for protein folding and drug discovery, because customers like Eli Lilly $LLY and Meta $META need the stack and do not have the scale to build it themselves. Our coverage separates that diversification claim from the "superintelligence already arrived" line Jensen used minutes later, follow for the read. Source: All-In Podcast -

Greg Brockman @gdb: OpenAI is putting $1 billion behind hospitals, water utilities, and other critical infrastructure that can't otherwise afford frontier-model access. CrowdStrike $CRWD is the named partner on the discounted defender access. Asked about the commitment by host Erik Torenberg @eriktorenberg on the a16z show, Brockman said: "Every organization, every company, every government, critical infrastructure such as water service providers, hospitals should all be using this defender's window to secure themselves." Ben Horowitz @bhorowitz backed it with the pre-AI baseline: hospitals held hostage by ransomware, water supply hacked by state actors, years before any of this AI capability existed. Brockman: "I've never met a CISO who felt that they were appropriately resourced, right? That they were appropriately prioritized. Never." Our coverage tracks what CrowdStrike discloses on the partnership's terms, and whether Anthropic or Google answer with a matching defender-access program. Source: a16z Podcast -

Dave Blundin @DaveBlundin: some people took 15 years to adopt the internet. AI goes the same way. His rule for the skeptics: don't fight missionary battles. Adoption arrives as one hallelujah moment per person, and you can't force it, only be in the room when it lands. Some companies never cross the line at all. Moonshots AMA 289 with Peter Diamandis @PeterDiamandis. Source: Moonshot with Peter Diamandis -