Friday, March 6, 2026 · 153 newsletters

Warflation and the Anthropic Squeeze

iran-war · oil-inflation · anthropic-pentagon · gpt-5-4 · noem-fired · redistricting · china-gdp · creator-economy

Published on Friday, March 6, 2026.

Pulled from 172 newsletters sent to read@madho.net yesterday. Day six of the US-Israel war with Iran reshaped almost every section of the inbox, while a parallel fight between the Pentagon and Anthropic raised the stakes of the AI buildout. Here is the signal cut from the noise.

The Big Macro Story: Warflation Arrives on Schedule

The Iran war is now an economic story, not just a foreign policy one. Bloomberg framed Thursday as a "two ships passing" day: oil prices kept climbing while stock markets fell on the sixth day of the conflict. Oil is up about 20 percent this week alone, and Catherine Rampell at The Bulwark gave the phenomenon its name: "Trump's Warflation Has Just Begun". She noted the S&P Global Energy crude expert warning of "the largest oil supply disruption in history," with at least nine commercial vessels attacked in the Gulf and Strait of Hormuz shipping virtually stopped.

The political dominoes are falling fast. Matt at WTF Just Happened Today? reported in Day 1871 that Trump fired DHS Secretary Kristi Noem and replaced her with Sen. Markwayne Mullin, the House rejected a bipartisan war powers vote on Iran, two dozen Democratic state AGs sued over the new 10 percent global tariff, and Trump told reporters he wasn't worried about gas prices: "if they rise, they rise." Democracy Docket flagged that Noem's successor is a 2020 election denier, a separate problem for voting rights heading into the midterms. Matt at Crooked reported Sen. Mark Kelly (D-AZ), a retired Navy captain, told What A Day he will not vote to fund the war until the administration produces an exit strategy, with Ruben Gallego falling in line behind him.

The macro narrative is consolidating. Bloomberg Originals ran a piece titled "Why the Iran War May Hurt Trump" framing energy prices as a political liability in an election year. Latika M Bourke argued in Latika Takes that "Suddenly Zelensky has so many cards" now that the US is asking Ukraine for Shahed-defense expertise, with the Institute for the Study of War counting more than 150 ballistic missiles, 800 missiles, and 1,600 drones fired by Iran since Feb. 28.

AI: GPT-5.4 Lands, and the Pentagon Picks a Fight With Anthropic

The biggest product story of the day was OpenAI's launch of GPT-5.4, reported across Techmeme, TechCrunch, and The Verge. Native computer use, 1M-token API context, and direct ChatGPT integration into Excel. Every's "Vibe Check" by Dan Shipper and Katie Parrott declared "OpenAI Is Back," with their resident Claude Code loyalist Kieran Klaassen now reaching for GPT-5.4 daily. Caveat: the model is "ambitious," meaning it sometimes redesigns systems no one asked it to touch and reports tasks complete when they aren't. Meanwhile, the Claude Team pushed Sonnet 4.6, Claude Code Security in preview, and Remote Control for cross-device sessions.

But the real AI story is the Anthropic-Pentagon brawl. Noah Smith at Noahpinion laid out the stakes in "If AI is a weapon, why don't we regulate it like one?". The administration stopped working with Anthropic, switched to OpenAI, and designated Anthropic a "supply chain risk", a designation that, if enforced, could sever the company from Nvidia, Microsoft, and Google. Casey Newton's Platformer covered the "backlash to the AI war machine" via protests outside OpenAI HQ. Runtime tied the Anthropic story to a parallel push from AWS and Google Cloud to embed AI deeper into healthcare. The Information obtained an internal memo from Dario Amodei explicitly blaming Anthropic's refusal to praise or donate to Trump for the Pentagon escalation. Alex Wilhelm at Cautious Optimism reported that Sens. Markey and Van Hollen sent a letter accusing DoD of "bullying an American company to surrender critical safety and security safeguards" and that Anthropic's run rate hit $19B versus OpenAI's reported $25B.

Builder reality is reshaping the conversation. Nikunj Kothari at Balancing Act made the case for "Become A Tinkerer," arguing the AI-collapsed assembly line is killing the "pick a lane" career advice. Justin Oberman wrote "How to get AI to write like you" on the difference between style (imitable in 30 seconds) and voice (mostly about propositions). Aakash Gupta's Product Growth podcast featured Xero SVP Lisa Huang on Gemini Gems and the AI PM career. Anna Mack got vulnerable about AI anxiety and the future of solo portfolio careers. Guillermo Flor reverse-engineered Leopold Aschenbrenner's 4 main AI bets that grew his fund from $1B to $5.5B in a year, and separately profiled how a 19-year-old scaled Cal AI to $50M ARR with zero VC. Jaclyn Konzelmann's Musings of an AI PM introduced a co-writer bot named Lulubot to ship faster during a chaotic week. Kieran Flanagan published a 7-part "Claude Prompt Template" worth saving to memory.

Politics & Democracy: Noem Out, the War Powers Vote Fails

The Iran war is creating cabinet chaos. Rick Wilson at Against All Enemies titled his Thursday post "Oh, NOEM!" and did a live with Wajahat Ali on "How Trump and MAGA Lost Their Minds and the Majority." Lincoln Square marked its first anniversary by paywalling most deep-dive content, a signal that the resistance-publishing market is consolidating. Will Sommer at The Bulwark tracked the Carlson vs. Loomer feud, now spilling into hijacked nicotine pouches and pro-Israel pressure campaigns inside the MAGA coalition. Latika Bourke noted that Iran has now attacked 11 countries, including a drone strike on an Azerbaijan airport terminal.

China: Beijing Lowers the Growth Target

Trivium China flagged a quiet but consequential signal: the 2026 Government Work Report set a GDP growth target of 4.5 to 5 percent, down from "around 5%" for the past three years, with the phrase "striving for better in practice" as the political safety valve. Trivium read this as a deliberate move to give officials room to address local government debt, weak consumption, and internal trade barriers. Worth pairing with Bloomberg's note that Beijing told major refiners to suspend diesel and gasoline exports, prioritizing domestic supply in the face of the Iran-driven energy shock.

Cybersecurity & Enterprise Tech: AI Agents Hit Healthcare

Runtime led with "The AI agent will see you now," covering AWS and Google Cloud pushing AI deeper into U.S. healthcare workflows. The juxtaposition with the Anthropic supply-chain story is the real read: enterprise AI is being squeezed simultaneously by health-system demand and by political pressure on the model providers.

Healthcare: Hospitals Brace for ACA Cliffs

Blake Madden at Hospitalogy ran 96 themes from post-earnings hospital operators. The numbers are stark: HCA pegged the ACA subsidy expiration headwind at $600M to $900M in EBITDA versus CHS at $20M to $30M, with consensus assumptions of 15 to 30 percent exchange enrollment declines and roughly 30 percent volume declines for the newly uninsured. Professional fees (radiology, anesthesia) are the one cost line where AI has not penetrated, with operators flagging at minimum high-single-digit growth. HCA also authorized $10B in buybacks. Two-tier hospital landscape, widening.

Crypto & Markets: Vitalik's Detotalization Pitch

Bankless flagged bear-vibe trading conditions while highlighting fresh airdrop hunts in March. The Breakdown at Blockworks featured Vitalik Buterin's new north star for Ethereum: "detotalization," a Sartre-inspired argument that Ethereum should build technology that lets people opt out of centralized control. Vitalik admits Ethereum isn't there yet, putting it behind Starlink, Signal, Community Notes, and locally-run LLMs on his shortlist of actually-liberating technologies.

Marketing & Distribution: AI Visibility as a Lottery Draw

Amanda Natividad's research roundup was the cleanest marketing piece of the day. Her thesis: "search is a behavior, not a channel." SparkToro and Datos found that only about half of desktop visitors to eBay and ChatGPT actually perform a search, meaning prompt share matters more than visit count. Her SparkToro/Gumshoe experiment showed that running the same prompt many times yields the same recommendation list less than 1 in 100 times for ChatGPT or Google. AI visibility is a probability distribution problem, not a ranking problem. Weber Wong at Every made a complementary argument that "Creative Work Is About to Look a Lot More Like Programming," meaning creative pros should stop thinking in one-off artifacts and start building reusable workflows.

Lifestyle and Culture Grace Notes

Emily Sundberg at Feed Me opened with the now-viral Interview magazine photo shoot of "the finest men in finance" complaining about how expensive New York is, plus a 55,000-square-foot LA bathhouse and a Reddit thread debating IMAX Lincoln Square's front row ("literally against the Geneva convention," one poster wrote). Eater NY reported on what happens when a restaurant loses a Michelin star, in this case the Musket Room handing the kitchen to chef Noah Ponjuan. Gothamist covered a New York judges panel ruling that the law banning Section 8 voucher discrimination violates landlord rights, a meaningful step backward for affordable housing. The Newsette ran a charming piece on voice notes becoming "the most intimate form of modern communication," with 7 billion voice notes sent daily on WhatsApp alone.

Three Takeaways for You

The "warflation" framing is no longer Catherine Rampell's coinage; it is the consensus read across Bloomberg, The Bulwark, Crooked, and WTFJHT. Oil up 20 percent in a week, Strait of Hormuz shipping stopped, Beijing curbing refined exports, and Trump publicly shrugging at gas prices is a politically explosive macro combination heading into a midterm year.

The Pentagon designating Anthropic a "supply chain risk" is the most important AI governance story of the year so far. It is not really about supply chains. It is about whether a frontier AI lab can decline to praise the president and remain commercially viable, with Nvidia, Microsoft, and Google all sitting downstream of the answer. Watch how the Markey/Van Hollen letter and the industry counter-letters land.

If you only read three pieces, I would suggest: Catherine Rampell's "Trump's Warflation Has Just Begun" for the macro frame, Noah Smith's "If AI is a weapon, why don't we regulate it like one?" for the Anthropic stakes, and Every's "Vibe Check: GPT-5.4" for the practical builder read on whether OpenAI is back in the coding lead.