Iran, March 2026

Three crises converge in sixty days. A domestic uprising met with mass killing and the most comprehensive internet blackout in Iranian history. A nuclear confrontation that produced two rounds of US-Israeli strikes nine months apart. And the first assassination of a sitting head of state by airstrike in modern history — the killing of Supreme Leader Ayatollah Ali Khamenei on February 28, 2026.

This episode applies established academic frameworks to ask whether the simultaneous convergence of all three crises creates something qualitatively different from any one alone. Decapitation strategy theory (Pape, Cronin, Johnston, Jordan). Authoritarian resilience (Bellin, Brownlee, Slater, Levitsky). Velayat-e faqih and its succession architecture (Khomeini, Kadivar, Abrahamian). The IRGC as a political-military institution (Ostovar, Bajoghli). Welfare-state authoritarianism and its beneficiary constituencies (Harris). Network cascade theory applied to proxy degradation.

Eight perspectives steelmanned from inside their own logic: the Islamic Republic, the Iranian opposition, the Trump administration, Israel, Russia, China, the IRGC as a distinct actor, and the broader Iranian population as a composite.

Produced under active fog-of-war conditions. Every claim carries elevated uncertainty. Evidence is hedged where thin, contested where disputed, and absent where the blackout prevents verification. The episode does not resolve. The questions remain open.

---

How this episode was made:

Proxima.Earth uses an autonomous multi-model AI pipeline. No human is in the loop after subject selection. The methodology is the editorial control.

1. Source plan — Claude (clean context, web-search grounded). Identified current events, mapped structural domains, planned source acquisition with orientation notes on every source.

2. Social layer — Grok. Captured real-time public discourse on X. Verified posts with attribution. Narrative dynamics and perspective validation. Social content is evidence about discourse, not ground truth.

3. Academic review — ChatGPT Pro (deep research with web browsing). Adversarial fact verification, claim-by-claim correction, bias detection, missing scholarship, methodological critique.

4. Commission brief — Claude (clean context). Synthesized all inputs. 41 claims extracted with source orientation and confidence ratings. 25 structural questions. 8 steelmanned perspectives. Fog-of-war assessment. The agent had deliberative autonomy over analytical emphasis.

5. Script — Claude (clean context). 19,972 words written section by section with self-checking for redundancy, source coverage, and evidence survival. The agent had deliberative autonomy over narrative structure and pacing.

6. Audio — Kokoro TTS (local, bm_george voice). Broadcast processing chain (EQ, compression, EBU R128 normalization to -16 LUFS). Intro/outro music.

This episode was produced using Methodology v5.0, which was stress-tested and refined on this episode. The full methodology, including system prompts and production guide, is published at proxima.earth/methodology. It is open source and designed to be instructive — anyone with access to the models can run it.

Known gaps: No IAEA post-strike access data. No Persian-language domestic sources. Source pack is overwhelmingly Anglophone. No Russian or Chinese primary-source strategic analysis. No battle damage assessment. No ground-level economic testimony from inside Iran.

Sources include: AP, Reuters, FT, BBC, CNN, CPJ, NetBlocks, HRANA, Chatham House, CSIS, CFR, Brookings, Alma Center, IranWire, Iran International, IRNA, Press TV. Every source receives an orientation note identifying institutional affiliation, editorial lean, and directional bias. No source is treated as neutral.

Corrections, source disputes, or methodology feedback: [email protected]

On February 28, 2026, the United States and Israel launched joint military strikes on Iran. The Pentagon designated it Operation Epic Fury. Israel called it Operation Roaring Lion. Two hundred Israeli fighter jets struck roughly five hundred targets across twenty-four of Iran's thirty-one provinces. Supreme Leader Ali Khamenei was reportedly killed in the opening wave. Iran retaliated with missiles and drones against US bases across six countries, declared the Strait of Hormuz closed, and triggered the largest regional military crisis since the invasion of Iraq in 2003.

On Friday, February 27, 2026, Defense Secretary Pete Hegseth designated Anthropic -- an American AI company whose model Claude was already integrated into classified Pentagon systems -- a "supply chain risk to national security." The designation, previously reserved for foreign adversarial companies like Kaspersky and Huawei, bars every military contractor in the United States from conducting commercial activity with Anthropic. The reason: Anthropic refused to drop two contractual guardrails prohibiting mass surveillance of Americans and fully autonomous weapons.

This extended edition traces the full arc of the confrontation -- from Claude's presence during the January 3 Maduro raid in Caracas, through the Palantir phone call that triggered the crisis, through seven weeks of escalation culminating in presidential threats of criminal prosecution and the unprecedented supply chain designation. It goes deep on what large language models are before any ethics are applied, how Constitutional AI differs from RLHF and why it matters for systems making life-and-death decisions, and why "removing the guardrails" is technically not the simple act the Pentagon's rhetoric implies. It examines the legal instruments being repurposed (FASCSA, the Defense Production Act), the litigation likely to follow, the commercial cascade through the defense industrial base, the historical pattern connecting Qwest's Joseph Nacchio to AT&T's Room 641A to Apple's fight with the FBI to Anthropic's stand against the Pentagon, and the structural paradox at the center: on the same day the Pentagon blacklisted Anthropic for maintaining two safety restrictions, it accepted identical restrictions from OpenAI. Four plausible scenarios for the AI governance landscape in 2036.

This is also the first episode produced using a four-model AI research pipeline in which one of the models is made by a company at the center of the story. The same research prompt was submitted to Gemini, Grok, ChatGPT Pro, and Claude. Their responses -- and the systematic behavioral differences between them -- are part of the narrative. Full process transparency is documented.

Ninety-seven sources. Ten-section research brief. Five-tier confidence taxonomy. Available at proxima.earth

---

AI Transparency: This episode was produced using Claude Opus 4.6 (Anthropic) for primary synthesis and narrative, with parallel research from Google Gemini, xAI Grok, and OpenAI ChatGPT Pro. Human editorial direction at every stage. Claude is made by Anthropic, a principal actor in the events described. This conflict of interest is disclosed within the episode and mitigated through the four-model cross-checking methodology documented in the research brief. All named individuals and quoted material are drawn from verifiable public sources.

Story ID: TAR-2026-002E

On Friday, February 27, 2026, Defense Secretary Pete Hegseth designated Anthropic -- an American AI company whose model Claude was already integrated into classified Pentagon systems -- a "supply chain risk to national security." The designation, previously reserved for foreign adversarial companies like Kaspersky and Huawei, bars every military contractor in the United States from conducting commercial activity with Anthropic. The reason: Anthropic refused to drop two contractual guardrails prohibiting mass surveillance of Americans and fully autonomous weapons.

This episode traces the full arc of the confrontation -- from Claude's presence during the January 3 Maduro raid in Caracas, through the Palantir phone call that triggered the crisis, through seven weeks of escalation culminating in presidential threats of criminal prosecution and the unprecedented supply chain designation. It examines the legal instruments being repurposed (FASCSA, the Defense Production Act), the historical pattern connecting Qwest's Joseph Nacchio to AT&T's Room 641A to Apple's fight with the FBI to Anthropic's stand against the Pentagon, and the structural paradox at the center: on the same day the Pentagon blacklisted Anthropic for maintaining two safety restrictions, it accepted identical restrictions from OpenAI.

This is also the first episode produced using a four-model AI research pipeline in which one of the models is made by a company at the center of the story. The same research prompt was submitted to Gemini, Grok, ChatGPT Pro, and Claude. Their responses -- and the systematic behavioral differences between them -- are part of the narrative. Full process transparency is documented.

Ninety-seven sources. Ten-section research brief. Five-tier confidence taxonomy. Available at proxima.earth

---

AI Transparency: This episode was produced using Claude Opus 4.6 (Anthropic) for primary synthesis and narrative, with parallel research from Google Gemini, xAI Grok, and OpenAI ChatGPT Pro. Human editorial direction at every stage. Claude is made by Anthropic, a principal actor in the events described. This conflict of interest is disclosed within the episode and mitigated through the four-model cross-checking methodology documented in the research brief. All named individuals and quoted material are drawn from verifiable public sources.

Story ID: TAR-2026-002

Ohio was once the most "American" state -- the bellwether, the president factory, the demographic mirror. It produced seven presidents, the Wright Brothers, Neil Armstrong. For 130 years, it picked nearly every presidential winner.

Then it stopped.

This episode traces Ohio's transformation through nine chapters: the bellwether-to-meme arc, a twenty-year corruption record spanning five scandals and hundreds of millions of dollars, the projected loss of 675,000 people by 2050, a private economic development entity exempt from public records law, the Springfield immigration paradox, the gap between what Ohio voters choose on ballot initiatives and who they elect to govern them, both parties' institutional fractures, and the most expensive governor's race in state history.

This is the companion episode to "The Map Room" (OH-2025-001), which documented how Ohio's redistricting maps were drawn. This episode asks what those maps produced.

On the evening of February 24, 2026, the House chamber fills for the State of the Union address. Roughly seventy-two Democratic members are absent — the largest organized boycott in the history of the event. Those who attend wear white. One carries a handwritten sign. The president will speak for one hour and forty-eight minutes, the longest address in at least sixty years. He will award two Medals of Honor, present an immigration crime story involving an American-born killer with no connection to immigration, inflate a fraud figure by sixty-three-fold, and tell the opposition party that the only way they can win is to cheat.

This episode examines the ceremony itself — the ritual of shared governance continuing while the premise beneath it fractures. It is not a fact-check, though the facts are checked. It is not a partisan argument, though the positions are sharp. It maps a structural tension: what happens when the ritual persists but the belief system it was built to express does not.

This is an evolving document. Language is a thin abstraction layer that cannot be divorced from human perception, and even the prompts in this piece designed for maximum objectivity are prone to the biases of the human editor who directed the AI where to focus and when to provide more detail.

Eight chapters. Every factual claim sourced. Every perspective steelmanned before being complicated.

On transparency: This episode was written by Claude, Anthropic's AI, working from a detailed editorial commission, extensive human-directed research, and continuous human editorial oversight. Research was conducted using Grok for real-time social media intelligence, ChatGPT Pro for fact-checking and source verification, and Claude for synthesis and writing. The narration was generated by Kokoro, an open-source text-to-speech model running locally. Audio was processed through an open-source broadcast chain. A human conceived, commissioned, directed, and published this episode.

Sources include primary source polling data from PRRI, Pew Research Center, Gallup, CNN, and Ipsos/Reuters; democracy indices from V-Dem, Freedom House, the Economist Intelligence Unit, and the Century Foundation; legal analysis from the Brennan Center for Justice and Democracy Docket; fact-check compilations from NPR, the Associated Press, NBC News, CBS, ABC, PolitiFact, and FactCheck.org; academic literature from political science journals; and social media intelligence from X.

On February 22, 2026, Mexican military forces killed Nemesio Oseguera Cervantes, El Mencho, founder of the Jalisco New Generation Cartel, in the Sierra de Tapalpa. Within two hours, CJNG's retaliatory protocol paralyzed twenty of Mexico's thirty-two states. Twenty-five National Guard members were killed. More than 250 roadblocks shut down the country. The operation was a success. The aftermath was a catastrophe.

This episode traces the fracture lines that produced that moment and everything that follows from it. Eleven chapters. The cartel's structure and its $20 billion annual revenue machine. The fentanyl pipeline from Chinese precursor chemicals through Mexican labs to American streets. The kingpin strategy and why decapitating leadership consistently produces more violence, not less. The sovereignty trap between two countries that cannot afford to separate and cannot afford to trust each other. The forty million people of the Mexican diaspora caught between both. Claudia Sheinbaum's impossible position. The 98 percent impunity rate. The journalists murdered for reporting. The autodefensas who armed themselves when the state did not come. And the succession crisis now unfolding simultaneously in both dominant cartels for the first time in modern history.

Sources include INEGI, the DEA's National Drug Threat Assessment, the Congressional Research Service, CSIS, the International Crisis Group, the Modern War Institute at West Point, peer-reviewed research by Esberg, Phillips, Calderon, Dell, Trejo, and Skigin, and Mexican investigative journalism by Anabel Hernandez, Proceso, and Articulo 19.