Monday, March 2, 2026 · 87 newsletters

The Day After Khamenei

iran-war · anthropic-pentagon · ai-labor · tariffs · voting-rights · fintech-charters · multipolar-world

Published on Monday, March 2, 2026.

Pulled from 85 newsletters sent to read@madho.net over the past day. Sunday volume runs lighter, but the signal is unusually heavy: a single dominant geopolitical event swallowed almost every category writer, and the secondary story (Anthropic's break with the Pentagon) bled into nearly every business and AI newsletter on the list.

The Big Story: A War, A Dead Supreme Leader, and the First Twenty-Four Hours

This was the only story that mattered. Saturday morning, US and Israeli forces launched a coordinated air campaign across Iran. By Sunday, Trump had confirmed that Ayatollah Ali Khamenei was killed alongside much of Iran's defense council. News Items led with the Institute for the Study of War's tally: nearly 900 strikes in the first twelve hours, with the regime imposing an internet blackout to limit what gets out. Lincoln Square had retired Navy Commander Bobby Jones on by Sunday morning reporting three US service members killed, Iranian retaliation hitting US bases in the Persian Gulf as well as Abu Dhabi and Dubai, and at least 148 dead at a girls' school in southern Iran. Jones's most memorable line on Trump launching the war from Mar-a-Lago instead of the White House: "He has been desperately trying to recreate the moment that Barack Obama walked down the East Room and talked about killing Osama bin Laden."

The legal and intelligence backstory came out almost in real time. SpyTalk's Jeff Stein reported that the CIA had been tracking Khamenei's movements for months, feeding Israeli warplanes the location of a Saturday morning leadership meeting in central Tehran. Michael Isikoff, in a separate SpyTalk piece, traced the legal cover for the assassination back thirteen years to a secret Obama-era Justice Department memo arguing that targeted killings during "imminent threat" did not constitute assassinations under the executive-order ban. Matt Stoller framed the Monopoly Round-Up around the conflict's political economy, noting that this time Iran's response is a "drip attack" of small barrages of rockets and drones spread across Bahrain, Iraq, Kuwait, Jordan, the UAE, and Saudi Arabia, rather than the single choreographed barrage of June.

The macro and analytical responses converged fast. Foreign Affairs ran three same-day pieces (Ali Vaez on the Pandora's Box risk, Nate Swanson on Iran's escalation logic, Akbar Ganji on Khamenei's succession) plus a Karim Sadjadpour Q&A. Noah Smith at Noahpinion read the strikes as a genuine inflection point because killing a head of state crosses a Rubicon, and called out China and Russia's conspicuous non-response as evidence that Iran was never a core member of either alliance. Gov Brief Today flagged a parallel war that almost slipped notice: the US backing Pakistan's open warfare with Taliban-ruled Afghanistan, with hundreds killed along the 1,600-mile border. 1440 Sunday tied it to a stark structural fact: the last US-Russia strategic arms treaty expired in February, North Korea is vowing to expand its arsenal, and the nuclear non-proliferation regime is now functionally unanchored.

AI and Defense: Anthropic's Line in the Sand

While Iran burned, the AI industry was processing its own rupture. Techmeme led Sunday with Anthropic's statement that it would challenge any "supply chain risk designation" in court and that any such designation would only affect contractors' use of Claude on DoD work. Dario Amodei's framing: "We haven't received any formal information whatsoever. All we have seen are tweets from the President and tweets from Secretary Hegseth." The Atlantic's Ross Andersen got the scoop on what actually broke the talks: the Pentagon wanted Anthropic's AI to "analyze bulk data collected about Americans," and Anthropic refused. The Wall Street Journal then reported that the Pentagon used Claude in its major air attack in Iran hours after Trump declared the federal government would end its use of Anthropic's tools.

OpenAI took the deal Anthropic walked away from. The Neuron broke down OpenAI's $110B headline raise, arguing that only about $15B is actually deployable cash now: Amazon's $50B is mostly conditional on AGI milestones, Nvidia's $30B is compute-capacity commitments on Vera Rubin chips, and SoftBank's $30B is still a letter of intent. The round values OpenAI at $730B pre-money. Paul Graham's response was characteristically dry: don't be deterred from using Anthropic because you might one day sell to the DoD. Kevin Roose pushed back on the framing of Amodei as a Silicon Valley naif, noting that his favorite book is "The Making of the Atomic Bomb." The Social Juice noted that a 'QuitGPT' movement is gaining steam on the back of the OpenAI Department of War deal, with users calling to cancel ChatGPT.

The AI Productivity Reckoning Is Now Quantified

Two threads ran in parallel: more skepticism, and bigger consolidation. Skepticism is now load-bearing. Rich Turrin at Cashless led with an IBM survey: 84% of execs call AI transformative, only 26% use it effectively, and just 5% have agentic success metrics. He paired that with the NBER finding that 80% of companies report no productivity gain from AI despite billions invested. Sent Items' Matthew Hertz connected this back to Block's 40% layoff announcement, explicitly attributed to AI; Jack Dorsey wrote the memo in lowercase. The stock went up 23%. From Every, Mike Taylor argued the case for turning off ChatGPT's memory, naming "context rot" as the slow degradation that happens when stale preferences and contradictory signals pile up. Kate Lee separately pulled back the curtain on Every's editorial AI workflow.

Consolidation continued under the radar. Sacra estimated Mistral hit $400M in ARR in January 2026, up ~20x year-over-year, riding the European sovereign-AI thesis to a $13.8B valuation with a $2B Series C led by ASML. Guillermo Flor at Product Market Fit is bullish on Perplexity Computer's 19-model orchestration approach as "AI as execution coordination." The Signal's Alex Banks flagged Google's Nano Banana 2 image model, now default across Gemini, AI Studio, and Search, with text rendering accuracy jumping from ~85% to over 95%. Lenny Rachitsky interviewed Anthropic's design lead Jenny Wen on why the classic discovery-mock-iterate process is becoming obsolete.

Politics and Democracy: The Plan to Seize the Midterms

Trump pivoted off the war within a news cycle to election machinery. Marc Elias at Democracy Docket titled his Sunday note "Trump's attack on Iran and the plot against your vote," arguing the war is the pretext for a midterm power grab. Adrian Carrasquillo at The Bulwark reported on a private call where DHS official Heather Honey, an architect of 2020 stolen-election claims, promised election officials ICE would not be at the polls in November. Arizona's Adrian Fontes and Maine's Shenna Bellows both told reporters they did not believe her.

The more alarming story came from George Bounacos. Gov Brief Today flagged a ProPublica report that Michael Flynn convened six federal officials and a room of 2020 election deniers ten days ago to discuss declaring a national emergency to seize control of the midterms, with photos and a LinkedIn post documenting the dinner with Cleta Mitchell. Lincoln Square had Joe Trippi calling the AI election-interference threat "Agents of Chaos." Zach Everson, who spent a decade reporting on Trump's business conflicts at Forbes, announced he is joining Public Citizen as research director for its Trump accountability project, citing his finding that Binance now holds 87% of Trump's USD1 stablecoin supply. And Jennifer Schulze at Lincoln Square detailed FCC Chairman Brendan Carr's Pledge America Campaign asking broadcasters to commit to "patriotic, pro-America" content for the 250th.

Trade and Tariffs: The $130 Billion Refund Scramble

The Iran story buried what should have been the week's headline business story. Sent Items and Dexter Roberts at Trade War both led with the Supreme Court's ruling striking down Trump's sweeping IEEPA tariffs, triggering a $130B refund scramble across importers. China, ironically, came out a winner: Morgan Stanley estimates weighted tariffs on China dropped from 32% to 24% pending the next move. USTR Jamieson Greer has already announced that China tariffs should stay in the 35-50% range, and the administration is considering revoking Permanent Normal Trading status. Xi Jinping used a Beijing meeting with Friedrich Merz to declare that China and Germany should jointly defend "free trade," a sentence that would have been unimaginable in either capital a decade ago.

Fintech: The Bank Charter Rush Accelerates

Jason Mikula at Fintech Business Weekly catalogued an extraordinary recent wave of charter applications: Morgan Stanley Digital Trust, Nomura's Laser Digital National Trust Bank, Payoneer's PAYO Digital Bank, and Stripe's Bridge entity receiving conditional approval. Sam Boboev at Fintech Wrap Up framed it as buying a "different physics model": cheaper insured-deposit funding, federal preemption over state-by-state licensing, and direct payment rail access. Mikula also flagged that Basis raised a $100M round at a $1.15B valuation, led by Accel with Lloyd Blankfein participating. Tearsheet had PayPal positioning itself as the infrastructure layer connecting merchants to AI shopping surfaces (ChatGPT, Gemini), with EVP Michelle Gill arguing 40% of Americans will start product discovery on AI surfaces this holiday season. Samora Kariuki at Frontier Fintech ran a longer essay on African DFS leadership search dynamics, and Linas reported Revolut is building a "valuation staircase" toward a $150B IPO.

Marketing, Media, and the Wartime Ad Question

Justin Oberman at Advertising History Today wrote the most timely piece of the day: a long essay on whether brands and individuals should advertise during wartime, anchored in the February 2022 split-screen of CNN broadcasting from Kyiv beside an Applebee's chicken-wing ad. Steve Bryant wrote about "vibe wording" and how phones reorganize human attachment, framing it through Walter Ong's theory of orality. Marketing Letter flagged Google's February core update finishing, with ChatGPT traffic now converting better than organic. PRWeek UK reported WPP confirming the Burson restructure and Omnicom/IPG jointly cutting 8,200 jobs in 2025. Daniel Murray at Marketing Millennials wrote a practical guide to hiring a marketing AI agent that reads less like a hype piece and more like a brief-writing tutorial.

Lifestyle and Grace Notes

Om Malik returned to his weekend reads format with a unifying thread: who's really in control? His picks included VoteMyAI on Suno hitting $300M while artists declared war, Hanna Horvath at Your Brain on Money on the economics of low trust, and 404 Media on Ring's "Search Party" AI dragnet. Nautilus reminded readers that Tuesday brings one of the last total lunar eclipses before late 2028. Superhuman's Zain Kahn covered China's UFO-shaped eVTOL test flight in Wuhan. The Liber had its usual upper-tier list: Maze Art Gstaad wrapping, WatchHouse opening on the Millennium Bridge, Bryan Johnson's $1M Immortals program, no finishers at the Barkley. Polina Pompliano used C.S. Lewis to argue that armies and institutions matter only insofar as they prolong "a household laughing together over a meal," then linked a profile of Palantir's CTO. The juxtaposition felt earned today. Anand Giridharadas at The Ink wrote a short Sunday rally cry on the institutional press folding under pressure.

Three Takeaways for You

The geopolitical regime has changed, and not slowly. Killing the head of state of a country of 90 million is not a "perfunctory strike" like the June one Noah Smith described, and the markets, the alliances, and the legal architecture around drone warfare were all rewritten over a single weekend. Watch for what China and Russia do next; the silence so far is a real data point.

The Anthropic-OpenAI-Pentagon triangle is going to be one of the defining stories of 2026. Anthropic chose to fight a supply-chain designation in court rather than analyze bulk data on Americans. OpenAI took the deal. That divergence will price into recruiting, into enterprise procurement, and (per Paul Graham) into early-stage startup tooling decisions for years.

If you only read three pieces today, I would suggest: Noah Smith on the shape of the multipolar world (the cleanest geopolitical frame), Ross Andersen's Atlantic piece on the Anthropic-Pentagon dispute (the inside story on what actually broke the deal), and Adrian Carrasquillo at The Bulwark on the ICE-at-polls plan (the domestic stakes most people will miss while watching the war).