Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.
Back to Podcast Digest
Musk Loses OpenAI Case, Leopold’s 13F, Data Center Backlash | Diet TBPN
TL;DR
Musk lost his OpenAI case on a timing technicality, fast — a U.S. jury deliberated for about 90 minutes before unanimously dismissing Elon Musk’s claims against Sam Altman/OpenAI because the lawsuit was filed too late.
Leopold Aschenbrenner’s much-hyped 13F was more ambiguous than the timeline wanted it to be — the filing showed huge notional semiconductor puts, including roughly $2 billion on SMH, but the hosts stressed that 13Fs are stale March 31 snapshots and omit crucial details like strike, expiry, premium, hedging structure, and shorts.
The AI buildout is running into a bipartisan backlash, but for different reasons — the left worries about job loss, art theft, and creativity, while the right increasingly frames data centers as surveillance infrastructure, creating a broad anti-build coalition.
Kevin O’Leary’s Utah data center became a perfect PR target even if the project itself looks reasonable — the proposed $100 billion, 40,000-acre, 9-gigawatt Stratos project in Box Elder County is pitched as self-powered, using existing water rights in a remote desert valley, but ‘Mr. Wonderful’ is exactly the kind of flashy face critics can turn into a villain.
Eric Schmidt getting booed at the University of Arizona captured the mood problem around AI — the hosts argued that telling graduates to help shape AI may land at Stanford, but at a public university where students hear ‘your jobs are going away,’ the message predictably detonates.
One surprisingly concrete fix for data center opposition: just pay people directly — citing Ben Thompson, the show highlighted a proposal where a 1.6-gigawatt data center generating about $3 billion annually could send each of 11,500 local residents around $10,000 per year for just 3.8% of revenue.
The Breakdown
Musk’s OpenAI lawsuit ends with a whimper
The show opens on the day’s cleanest headline: Elon Musk lost his case against OpenAI because he waited too long to sue. The jury apparently deliberated for only about 90 minutes, and the hosts marvel at how quickly it was over — a big win for OpenAI, even if it came on what they call a weird technicality.
The internet waits for Leopold Aschenbrenner’s 13F like it’s a product launch
They describe genuine timeline mania over Leopold Aschenbrenner’s hedge fund filing, with people refreshing all Friday and spinning theories that he’d either petitioned not to disclose or gone entirely to cash. TBPN lets the bit breathe with some studio chaos — bright new lights, haircut jokes, and an on-air gift for Tyler — before pivoting back to why people cared so much: Leopold’s whole brand is built on the thesis that frontier AI keeps improving and compute demand keeps exploding.
Why the 13F caused so many bad takes
Once the filing lands, the big takeaway is that he appears to have massive semiconductor puts, including a huge SMH position, while also showing interest in names like Nvidia. But the hosts repeatedly warn that 13Fs are easy to overread: this is only a March 31, 2026 snapshot, options are reported notionally, and nobody can see strike prices, expirations, hedge ratios, or broader structures, so the ‘billions in puts’ discourse can be badly misleading.
AI backlash is now coming from both political directions
The show says data center opposition has broadened into a true left-right coalition, something Saagar Enjeti had predicted earlier. On the left, the anger is about job displacement and stolen creativity; on the right, the newer framing is that data centers are ‘surveillance centers,’ with the added irony that activists are often using AI tools and AI-generated graphics to organize against AI.
The Utah megaproject and the problem with making Mr. Wonderful the mascot
A big chunk of the episode digs into Kevin O’Leary’s proposed Stratos data center in Utah: a $100 billion, 40,000-acre project that could use 9 gigawatts at full buildout. The hosts think the actual plan sounds fairly by-the-book — remote desert valley, self-generated power, purchased water rights, limited local disruption — but they also point out that O’Leary, with his dual luxury watches and TV persona, is almost comically easy to cast as the face of extractive AI infrastructure.
‘23 atom bombs a day’ and the data center rhetoric war
They react to a viral critique claiming the Utah project could create thermal load equivalent to ‘23 atom bombs’ per day, and mostly treat the comparison as classic fear-mongering by scale analogy. The more interesting discussion is around water rights: they walk through the ‘I drink your milkshake’ aquifer logic, noting that owning water rights doesn’t necessarily mean the water was actively being used for crops right now.
Eric Schmidt gets booed, and TBPN says the industry should not be shocked
A clip of Schmidt being booed during a University of Arizona commencement becomes the emotional centerpiece. The hosts say this is what happens when AI’s public story is all vibe coding, productivity talk, and job anxiety, while the magical consumer moments people remember from earlier tech waves — Uber, Yelp, Groupon — feel mostly absent.
Ben Thompson’s blunt fix: give locals a check
To close, they run through Ben Thompson’s argument that tech has a messaging problem, a misinformation problem, and a deep inability to relate to normal people. His most memorable solution is the simplest one: if a 1.6-gigawatt data center can do $3 billion in annual revenue, then cut residents in the host town a yearly check — say $10,000 each — and stop pretending vague tax benefits are enough to win consent.
Share
Keep Reading
Two Ways to Follow
The Weekly Echo. The inbox-shaped summary of what mattered.
New editorials announced here.
