
Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.
The Trump–Xi summit is being framed as a stability play, but the real signal is who got invited — Trump showed up in Beijing with Elon Musk, Tim Cook, Jensen Huang, Kelly Ortberg, David Solomon, Stephen Schwarzman, Larry Fink, Jane Fraser, and Meta’s Dina Powell McCormick, making business leverage and supply-chain politics as important as the diplomatic optics.
Jensen Huang’s presence mattered because Nvidia sits at the center of the China chip fight — the hosts dwell on reports that he may have been a late addition to Air Force One, noting Nvidia is far more exposed to export controls and AI competition with China than most of the other CEOs on the trip.
The summit talks are touching semis, rare earths, Taiwan, and Iran — but not the big AGI questions Silicon Valley obsesses over — the hosts point out that Sam Altman, Dario Amodei, Demis Hassabis, and Sundar Pichai weren’t there, underscoring a gap between frontier-AI discourse and actual US-China statecraft.
Anthropic’s restricted ‘Mythos’ rollout to Japan highlights how inconsistent US AI policy looks next to Nvidia selling into China — Japan’s three megabanks, including MUFG and Sumitomo Mitsui, are getting access via Treasury-level approval, while critics argue it clashes with letting China buy the compute needed to build its own Mythos-class systems.
One of the most interesting unanswered AI questions is what China’s own ‘Mythos moment’ will look like — the hosts compare Anthropic’s diffuse, heavily financed American cap table to DeepSeek/High-Flyer’s tightly controlled structure and wonder how the CCP would react if a Chinese lab hit a genuine frontier-model breakthrough.
Outside geopolitics, the show swings to orbital data centers and inflation — two very different versions of infrastructure stress — they joke about blimp and undersea GPU farms while covering reports of SpaceX and Google talks, then end on inflation re-accelerating to 3.8% and producer prices at 6%, with fears of stagflation if AI capex stays concentrated while the broader economy weakens.
The episode opens on the Wall Street Journal’s Trump–Xi summit coverage, but the hosts immediately zoom in on the business roster, not the political one. Elon Musk, Tim Cook, Jensen Huang, Kelly Ortberg, David Solomon, Stephen Schwarzman, Larry Fink, Jane Fraser, and Dina Powell McCormick are the real tells here: this is diplomacy through supply chains, chips, aircraft, and capital markets.
They get hung up on whether Jensen was always supposed to be on the trip, noting the reporting drama over whether Trump personally invited him and the detail that he boarded in Alaska. The reason it matters is simple: Nvidia is where US-China AI tensions get concrete, and Huang showing up beside Trump and Elon makes the chip story feel central even if no one says the quiet part out loud.
In classic TBPN fashion, the geopolitics detour turns into a riff on Air Force One aesthetics and Jensen’s leather-jacket pack list. They make the bigger point that founder uniforms compound into brand mythology — Steve Jobs’ black turtleneck, Palmer Luckey’s Hawaiian shirt, Jensen’s jacket-and-sneakers look — while joking that the downside is old clips can go viral forever if you still dress exactly the same 10 years later.
Reading through the Journal editorial, they summarize the stakes as tariff truce, rare earth supply, advanced chip exports, Taiwan posture, and China’s support for actors like Iran, Russia, and North Korea. The hosts stress that everyone wants “stability,” but the Wall Street Journal’s point is that personal rapport won’t erase Beijing’s strategic aims, especially on AI and Taiwan.
One sharp observation: the people missing may be as revealing as the people present. No Sam Altman, no Dario Amodei, no Demis Hassabis, no Sundar Pichai — which, to them, shows that the live geopolitical conversation is still about semis and trade mechanics, not the “AI 2027,” post-AGI, or fast-takeoff scenarios that dominate Silicon Valley debates.
The show then jumps to Anthropic reportedly granting Mythos access to Japan’s three megabanks — MUFG, Sumitomo Mitsui, and Mizuho — with Treasury Secretary Scott Bessent apparently involved in the approvals. That sets up the tension they can’t shake: Washington is treating frontier AI as national-security-sensitive enough to personally gatekeep access, while also allowing the broader compute and chip negotiation with China to stay muddy.
This becomes the most genuinely analytical section of the episode. They wonder what happens if DeepSeek or High-Flyer reaches a truly frontier capability in China: unlike Anthropic, with many investors and stakeholders, a closely held Chinese AI company could collide much more directly with CCP control, and whatever that confrontation looks like may be one of the most dramatic AI governance stories ahead.
In the last stretch, they cover reports that SpaceX and Google are discussing orbital data centers, with Project Suncatcher and a timeline that sounds very “a decade or so away,” then riff about blimp-based and undersea data centers as the more absurd alternatives. They close on inflation: CPI at 3.8%, producer prices up 6% year over year versus a 4.8% estimate, and a worry that if AI capex props up headline growth while the rest of the economy crawls, the real threat is stagflation.
Share
Keep Reading
The Weekly Echo. The inbox-shaped summary of what mattered.
New editorials announced here.

Playbook
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.

Playbook
Learn how tasteful prompting helps you move beyond generic AI output by shaping context, style, and judgment from the start.

Playbook
OpenAI shipped /goal for the Codex CLI. It turns a prompt into a persisted, self-continuing contract.