Mark Cuban on the AI Bubble: Who Actually Gets Wiped Out?
⏱ 41 min listen · ~3 min read
🎙 Mark Cuban
Forty-five minutes of Cuban with a thesis on exactly who eats the losses if AI unwinds, plus the mechanics of the Yahoo collar told first-hand.
Key takeaways
- 01Mark Cuban puts the AI downside on the VCs, private equity and private credit funds that bought at peak prices — not on retail investors, as in the dot-com bust.
- 02Funding hundreds of billions of capex with 50-year bonds on top of a private-credit problem is what Cuban calls 'planning for perfection'.
- 03Cuban's dark-fiber analogy has surplus data centers ending up as pickleball courts if a price-performance breakthrough strands the capacity.
- 04Lovable, one of Cuban's investments, ships 770,000 applications a week with only 20% of its users engineers.
Chapters
0:00 · Not a dot-com rerun: who actually gets wiped out2:38 · 50-year bonds and data centers as pickleball courts5:13 · Go public: stock as acquisition currency, the M&A freeze6:30 · The Yahoo collar story7:46 · Enterprise AI is harder than anyone expected10:29 · Lovable's 770,000 apps a week+7 more
Read the full summary →
Senators John Fetterman and Dave McCormick: Bipartisanship, Money in DC, Datacenters, Graham Platner
⏱ 43 min listen · ~4 min read
🎙 John Fetterman and Dave McCormick
A sitting Democrat and Republican from the same swing state stress-test each other's talking points in real time, with actual numbers — on data centers, fentanyl, the filibuster and what their races cost.
Key takeaways
- 01Fetterman and McCormick flipped Pennsylvania Senate seats in back-to-back cycles off the same coalition of electricians, pipe fitters and steam fitters.
- 02Fetterman calls the 2020 Democratic push to kill the filibuster a collective mistake and now says defending it is a hill he would die on.
- 03Fetterman breaks with his own party on Iran and on the data-center moratorium, which he calls a China-first policy, and refuses to call Trump a fascist.
- 04McCormick's evidence of progress: Pennsylvania fentanyl deaths down from 4,000 to 1,600 in a year, and $92 billion committed at his state energy summit.
Chapters
0:00 · Country over party: the odd-couple opening statements3:51 · Pennsylvania as microcosm — the coalition both men won6:28 · J.Cal's charge sheet: debt, inflation, record-low approval7:46 · Iran, $5 gas and the lone Democrat defending the strikes10:30 · The filibuster mea culpa13:09 · The K-shaped economy and the bottom half+9 more
Read the full summary →
Former Intel CEO on What Went Wrong, What's Next + Lovable CEO on the Real Promise of Vibe Coding
⏱ 49 min listen · ~3 min read
🎙 Pat Gelsinger and Anton Osika
A rare insider's account of all three of Intel's epochal misses — Apple silicon, CUDA, TSMC — from the CEO who lived them, plus concrete numbers on how fast vibe-coding platforms are actually growing.
Key takeaways
- 01Gelsinger blames Intel's decline on business people replacing technologists at a company whose executive staff once held 15 PhDs out of 20.
- 02$100 billion went out in Intel buybacks and dividends in the five or six years before Gelsinger returned as CEO, while no new factory was built for a decade.
- 03Steve Jobs had Apple's operating system ported to x86 four releases before revealing it, so Intel never saw silicon independence coming.
- 04TSMC turned the foundry model Intel dismissed into five times Intel's wafer volume, now closer to seven.
Chapters
0:00 · Inside Intel's decline1:18 · Technologists vs. bean counters5:10 · Steve Jobs' secret x86 port and Apple silicon7:51 · Jensen, CUDA, and the Larrabee project killed10:33 · TSMC and the foundry model Intel ignored14:32 · Taiwan: three weeks of energy reserves+9 more
Read the full summary →
The Trillion-Dollar Industries AI Is Disrupting: Voice, Law & the End of the Billable Hour
⏱ 51 min listen · ~3 min read
🎙 Mati (ElevenLabs) and Max (Legora)
Two founders competing directly against frontier labs explain, in concrete operational terms, what they think the labs cannot take from them — and neither answer is scale.
Key takeaways
- 01ElevenLabs shipped the first human-sounding text-to-speech model in early 2023 and reached roughly $600 million in ARR with 600 employees.
- 02ElevenLabs credits its voice lead to architecture rather than scale, and calls labelling the real bottleneck — over 1,000 in-house contractors label audio.
- 03ElevenLabs has never had product managers, and Mati embeds engineers in talent, legal and go-to-market teams to review what those teams now build themselves.
- 04Legal is 4% software and 96% service — about $40 billion of software against a trillion dollars of annual spend — in a market short of lawyers.
Chapters
0:00 · The revenue ramp: 20 months to $100M ARR5:07 · Engineers embedded in talent, legal and go-to-market7:44 · Running the company without product managers15:20 · Voice cloning, impersonation and identity19:13 · Tracing, moderation and detection for other labs' models20:29 · A voice marketplace that has paid out $22 million+9 more
Read the full summary →
Open Source Wins, AGI Is Here, and Scorsese's AI Toolkit with CEOs of Cerebras & Black Forest Labs
⏱ 63 min listen · ~3 min read
🎙 Andrew Feldman and Robin Rombach
Two people who actually ship silicon and models both argue, independently and in the same hour, that architecture rather than scale is now the thing that moves — which is not what the surrounding discourse says.
Key takeaways
- 01Feldman says Cerebras has left Moore's law behind and expects better than 2x in 18 months, crediting the architecture's youth over any new process node.
- 02A GPU design two decades old buys its gains from smaller geometry, while a new architecture still has unexploited knowledge of the work it is handed, per Feldman.
- 03Reasoning is inference, and reasoning models burn enormous token counts internally, so Feldman's 15x-faster machine turns a day of runtime into weeks of thinking.
- 04Cerebras reports a $25 billion backlog, with buyers ordering chips before the chips were finished.
Chapters
0:00 · Andrew Feldman on the scale of the AI build-out2:38 · A $25 billion backlog and demand booked ahead of supply5:14 · Token-maxing, and the AWS credit-card era repeating9:10 · Watching a reasoning model choose its own sources11:44 · Why reasoning is an inference problem13:15 · Breaking off Moore's law with a young architecture+9 more
Read the full summary →
Autobiography of a Yogi49
The Daily7
Crime Junkie6
Huberman Lab6
MrBallen Podcast: Strange, Dark & Mysterious Stories6
My First Million6
Planet Money6
Rotten Mango6
The Mel Robbins Podcast6
The Rest Is History6
The Why Files: Operation Podcast6
SmartLess5
This Past Weekend w/ Theo Von3
The Shawn Ryan Show1