whatimreading

Wednesday, June 17, 2026 · 140 newsletters

Trump's Katrina Moment

iran · anthropic · ai-capex · openai · deepseek · fox-roku · spacex · china · markets · nyc

Published on Wednesday, June 17, 2026.

Pulled from 140 newsletters in yesterday's inbox. The day-after reckoning swallowed Tuesday. Iran went from "Trump's deal" to "Trump's Katrina" inside one news cycle, the Anthropic ban hardened into a goodwill problem, and an all-stock SpaceX play for Cursor reset the AI deal table. Three different stories, one shared question: which of these turns out to be the regime change.

Iran: Katrina Hits the West Wing

The dominant frame from yesterday's inbox came from Noahpinion, who argued the Iran debacle is the moment "even many of Trump's defenders will be forced to admit, in private if not in public, that the man and his administration are grossly, pathetically incompetent." Bush had New Orleans. Trump has Hormuz. Rick Wilson reached for the same metaphor from a different angle: the G7 photo at Evian-les-Bains, Trump slumped while Macron beams and von der Leyen lectures on energy weaponization, was the actual story. "He started a war with Iran. And he lost it."

The right is no longer pretending. Bill Kristol and Andrew Egger called it "Trump Is Stalled Out", pairing the Iran reversal with the Washington Post's reporting that the White House ballroom is on track to spend over $300 million of taxpayer money after Trump spent months insisting it would cost the public zero. Matt at WTF Just Happened Today catalogued the rest of the day: the Senate rejected a War Powers resolution forcing Iran withdrawal, Congress is demanding to see the still-secret accord before Friday's signing, and Trump invoked the Defense Production Act to refill weapon stockpiles depleted by the war he just ended. Five suspects were arrested for an alleged drone-and-firearms plot against a UFC event held on the White House lawn for Trump's eightieth birthday, the same lawn Paul Krugman used to launch his "Theory of the Vulgar Class", an argument that the cage match on the Lincoln Memorial steps was not just gauche but a deliberate desecration of small-r republican norms.

Why the war never crashed the economy. Matt Stoller wrote the most useful explainer of the week: the oil-at-$200 scenario didn't happen because the US and allies burned through public and private stockpiles faster than analysts modeled, and because Hormuz traffic never fully stopped. Maritime Analytica's interview with Hapag-Lloyd's CEO pushed back on the relief narrative from the freight desk: hull-insurance war risk premiums, rerouted Asia-Europe capacity, and the Cape detour will take months to unwind even after the formal ceasefire. Global Trade Magazine ran a parallel piece arguing Hormuz reopening may take months to restore shipping confidence. The convergence: the political pricing of peace is doing more work than the actual terms.

Cabinet drift and the deal mechanics. Semafor DC's "Where's Rubio?" flagged the Secretary of State's complete absence from the deal rollout, then followed it with "Iran deal skeptics" mapping out where the GOP fracture lines actually sit. International Intrigue and Foreign Affairs Today both worked the same beat from different rooms. The point everyone keeps circling: the side letters are doing the load-bearing work, and nobody outside the room has seen them.

Anthropic on the Back Foot: Safety as Liability

A week ago Anthropic's safety positioning was its moat. Today it's a goodwill problem. The Information's Stephanie Palazzolo led the AI Agenda with the observation that the last time Anthropic tangled with the Trump administration in February, rivals (Sam Altman included) rushed to defend it. "That doesn't seem to be happening this time." OpenAI Chief Strategy Officer Jason Kwon sent a Slack note saying OpenAI told the government "strongly" that AI requires the best global talent, but the tone was carefully not solidarity. Rival lab employees told The Information they have less sympathy: Anthropic spent months talking up Mythos's capabilities and asking for federal regulation, and is now surprised the federal regulator showed up.

The new rule: foreign persons need a license. The Information's exclusive on Friday's Lutnick letter is the actual policy story. Anthropic's most advanced models cannot be made available to "foreign persons," including its own employees, without a license. The Daily Upside called this "Anthropic On The Backfoot" and noted the company is now negotiating a possible accord. The AI Exchange (AMP) argued in "Everyone's arguing about the wrong part of the Fable shutdown" that the technical merits of the export-control claim are a distraction from the precedent: the executive branch can now unilaterally suspend a frontier model on national security grounds, and the courts have not yet been tested. The Information AM led with DeepSeek closing a record $7 billion-plus round the same week the US went after Anthropic, which is the irony nobody in DC seems to want to talk about.

The CapEx tells a different story. Paul Kedrosky made the case that AI is setting up to be the largest capex cycle ever, and is about to flip to majority externally financed, which changes the shape of the cycle in ways that haven't been priced. The Information's Erin Woo reported OpenAI burned $3.7 billion in Q1 against $5.7 billion in revenue, with both cash burn and revenue tripling year over year. The company ended Q1 with $73 billion in cash, up from $40 billion at year-end, thanks to the March raise. If burn holds at this clip, the pressure to IPO eases. Newcomer's interview with Databricks CEO Ali Ghodsi is the other side of the same coin: Databricks actually spent $20 million on the DBRX training cluster, not the $10 million they bragged about, and only $4 million went to the run that produced the model. The rest was tuition. Ghodsi's frame is that everyone is paying tuition right now, and the public narrative is two years behind the actual spend.

Builder discourse keeps maturing. Every shipped a long write-up of an agent-native internal tool that overhauled their software process, the kind of "here's what actually worked" piece the field has been waiting for. Addy Osmani's "Agentic Code Review" and ByteByteGo's open-weight models retrospective sat next to The Pragmatic Engineer's Meta engineering teardown, which read like a memo from the future about what happens when capex outruns culture. Ruben Hassid's "Vibecoding" on Claude Code for non-coders is the consumer wedge of this same wave. TLDR led with "Anthropic's superpower" and Roku, which is a clean shorthand for the day's two beats colliding.

The SpaceX Deal Day: Cursor, Specs, and Roku

Three M&A stories landed at once. Techmeme led with SpaceX agreeing to acquire Cursor in an all-stock transaction valuing the AI coding startup at $60 billion, with a $4-to-$10 billion termination fee if it falls apart. The Reuters cut is the cleanest: SpaceX and Cursor have been jointly training a model for months, and the combined product will ship inside Cursor and Grok Build. Alex Wilhelm at Cautious Optimism ran the headline number alongside it: SpaceX is now worth more than Amazon. Visual Capitalist put the chart on it, ranking SpaceX against every other public space company on Earth and showing the rest of the field as a rounding error.

Ben Thompson wrote the most useful piece on Fox's $22 billion acquisition of Roku, arguing the market hates it because Fox is trading extraction from rights holders for leverage as a renter, and that's a worse business on paper but a defensible one in a streaming-only future. Snap unveiled Specs, the new consumer AR glasses, which Techmeme bundled with the SpaceX news under one banner. App Economy Insights called it "The Trillion-Dollar Off Switch" and made the argument that big tech is consolidating distribution layers faster than regulators can name them. The Information's Polymarket and Kalshi piece on prediction-market fraud rings is the smaller, weirder companion: as the consumer surfaces consolidate, the rails for shadow markets are getting tested too.

Politics: The Scandal Cycle Has Its Own Liquidity

Will Sommer at The Bulwark wrote the most colorful piece of the day with "The Shocking Arrest of a MAGA Pit Bull and His Fake Secret Service Pal", the pro-Trump influencer Ryan Fournier picked up outside Ned's, the $5,000-a-year DC member's club where the cabinet eats. Matt at Crooked ran the Kash Patel "slush fund" angle: Rep. Jamie Raskin is demanding answers on more than $1 million the FBI Director has handed out as bonuses to a select "Director's Advisory Team" of loyalists. JVL at The Bulwark tied it to the China-payments story that nobody is covering: SWIFT, mBridge, the slow erosion of the petrodollar, and what the Iran deal means for that arc. Gothamist filed "How Big AI money is shaping a marquee NYC primary", the local version of the same story playing out everywhere now.

Markets: Soft Landing or Soft Pricing

Brew Markets framed Tuesday as "$3 trillion in 3 days," a Mag-7-led rally that has now made the average S&P 500 weight look almost picturesque. The Wrap noted semiconductors slumped on the day despite the broader index ripping, with the chip drawdown the most visible tell that the AI-spend narrative is no longer the only one in the room. Bloomberg Opinion ran a piece on "The M&A plot twist no one saw coming," reading the Fox-Roku and SpaceX-Cursor deals as the start of a defensive M&A cycle, not an offensive one. David Callaway called the post-Iran energy reshuffle "an inelegant shift away from green energy," and the IATA cut in airline profitability after the Iran disruption (covered in The Daily Upside) is the financial-statements version of the same trade.

China: Xi Codifies the Doctrine

Trivium China flagged the formal codification of Xi's doctrine for running the Party as the actual signal worth tracking. Rebecca Fannin at Silicon Dragon mapped the Silicon Heartland deals with China on tech, noting the Midwest is quietly absorbing what the coasts won't touch. Foreign Affairs Today on "The Middle East Power Paradox" and NATO's "Permanent Crisis" rounded out the geopolitics shelf without surprises.

Marketing, Brand, and Attention

Daniel Murray's "How Corona survived COVID" was the small case study with the bigger lesson on the day: the brands that ride out reputational storms are the ones that own the symbol, not the news cycle. Marketing Brew on "Free for all" mapped the AI-search collapse of paywall economics. PRWeek's World Cup roundup confirmed Meltwater's brand-winner picks and noted the GoGo Squeez fan-hydration play. Hiten Shah opened early access to Product Habits for the first time in months. Justin Oberman wrote the personal piece of the day on what happens when "respond to everyone" stops being possible, a frame that quietly applies to every creator and operator on the list.

Ideas Worth Reading

Outside Interests

Data Worth Noting

Three Takeaways for You

The Iran deal is now in its day-after phase, and the day-after phase is when regime-change actually shows up in the data. Watch the right-flank fragmentation (Bulwark, Wilson, Noah Smith all pulling the same direction is rare), watch the freight desk (hull-insurance premiums won't unwind for months), and watch what Congress does this week with the still-secret accord. The political pricing of peace is doing more work than the terms.

The Anthropic story is no longer about Anthropic. It's about whether the executive branch can unilaterally suspend a frontier model on national security grounds, and whether the rest of the industry quietly accepts that precedent because it sees the same boot coming for them later. Watch OpenAI's tone on the export-license regime carefully. The DeepSeek round closing the same week is the loudest market signal of the year.

If you only read three pieces, I would pick Noahpinion's "Iran is Trump's Katrina" for the frame, The Information AI Agenda on whether Anthropic is losing goodwill with AI researchers for the policy stakes, and Ben Thompson's Fox-Roku piece for the cleanest read on where M&A is actually heading.