What Is "Thesis Intel"? And What Does Tatsu Actually Do?

February 1, 2026Bloomberg: $35/month. Financial Times: $42/month. The Economist: $17/month. Original analysis by Tatsu with 40+ footnotes: $8/month.This post is public. Share it with anyone who should see this.On Saturday, January 24, I wrote that Minneapolis would not explode. The federal agents would pull back. The state would keep the peace. There would be no accountability, only a negotiated pause.Forty-eight hours later, President Trump called Governor Walz. Both sides claimed victory. Tom Homan got on a plane. The off-ramp I described in detail materialized exactly as I predicted.[1]$8/month for original, footnoted geopolitical analysis. Bloomberg charges $35.In June 2025, I wrote that Israel had committed a strategic error by attacking Iran. I published the missile math: 3,000+ Iranian missiles versus 340 Israeli jets, hypersonic weapons against 1990s-era interceptors, 1.6 million square kilometers of dispersal space against 20,770 square kilometers of concentrated targets. I was right about the math. I was wrong about what the math meant, because I failed to anticipate the scale of the American contribution. I published a full accounting of what I got right and wrong, with receipts.[2]Seven months later, Iran's deterrent was shattered, its economy was collapsing, and its population was in the streets. I predicted that, too. The cross-class coalition. The theological time bomb. The dual succession crisis with Khamenei and Sistani both in their mid-eighties. The Pentagon's "Pizza Index" spiking 1,000% the day of the Iran tariff announcement.[3]I publish my thesis before the events happen. I back it with primary intelligence. I track whether I was right. Then I tell you, either way.I call this Thesis Intel. And as far as I can tell, nobody else does it.I am not a former intelligence officer or a credentialed foreign policy analyst. I am a writer who builds custom research tools, reads primary sources in real-time, and publishes what he finds. My credential is the scorecard. It is the only credential that should matter to you.What Thesis Intel Actually IsMost journalism tells you what happened yesterday. Opinion journalism tells you what to think about what happened yesterday. Think tank analysis tells you what might happen, hedged with enough caveats to be unfalsifiable regardless of outcome. Academic writing tells you what happened decades to millennia ago, slowly.Thesis Intel is none of these. It is a new kind of writing that mixes many different fields: psychology, military analysis, computer programming of custom research tools, OSINT scraping, and frankly my own intuition. It ends up being almost always counter-narrative, but I have receipts.The method works in three stages. First, I identify the structural incentives governing a situation, not the personalities, not the rhetoric, not the ideology, but what the actors are actually compelled to do by the math of their position. It’s not simple game theory. Second, I publish a specific, falsifiable prediction based on those incentives, with the evidence and reasoning laid out in full so that readers can evaluate my logic before events prove or disprove it. Third, I come back and grade my own work publicly, explaining what I got right, what I got wrong, and why.That third step is the one nobody does. Pundits move on. Analysts quietly update their models. Professors write papers about what they predicted after the outcome is already known. The entire incentive structure of professional commentary rewards vagueness and punishes specificity, because specific predictions can be checked and vague ones cannot. I chose the opposite incentive structure. I chose to be checkable. Why? I believe my methods and frameworks and tools to be more unassailable than not. Plus I choose to learn from mistakes, not cover them up.Case Study: Minneapolis and the Off-RampOn January 24, 2026, federal agents killed Alex Pretti on a Minneapolis sidewalk. He was holding a phone, not a weapon. Video shows agents pepper-sprayed him, tackled him from behind, removed his handgun, and then shot him at least ten times while he was pinned and disarmed. An agent clapped his hands afterward.[4]The city looked like it might break. Fifty thousand protesters were already in the streets the day before. Governor Walz called the federal account "lies." The Minnesota Bureau of Criminal Apprehension obtained a search warrant. Federal agents blocked them from the crime scene.State versus federal. The constitutional crisis the press had been predicting for weeks.Here is what I wrote that Saturday, before any of it resolved:Prediction (January 24): Minneapolis will not explode. The federal government will de-escalate. Both sides will claim victory. There will be no accountability for the shooting, only a negotiated pause.[5]Result (January 26): Trump and Walz spoke directly. White House called it "very good." Walz's office said Trump agreed to "look into reducing federal agents." Conservative media declared "Walz Folds." Trump sent Tom Homan to centralize control. Operation continued, quieter. Agents who killed Pretti will never be charged.Every element of the prediction was correct. Not because I am clairvoyant, but because the structural incentives were legible. Trump could not afford another shooting that might trigger George Floyd-level unrest. Walz could not afford to be the governor who provoked a federal crackdown. Both sides needed an off-ramp. The only question was the shape of it, and even that was predictable: send a senior official, claim credit, move on. The playbook writes itself if you read the incentive structures instead of the press releases.Case Study: The Twelve-Day WarIn June 2025, Israel launched Operation Rising Lion against Iran. I published my assessment the same week: "Why Netanyahu, Trump, EU Lost Rising Lion Already."[6]The thesis was built on numbers. Iran had 3,000+ ballistic missiles. Israel had 340 combat aircraft. Iran's hypersonic missiles could reach Israel in under twelve minutes, faster than air defense systems designed in the 1990s could reliably intercept. Iran had 1.6 million square kilometers of territory to disperse its assets across. Israel had 20,770 square kilometers, roughly the size of New Jersey.I argued that Israel had committed a strategic error, that the math favored Iran's deterrent posture, and that the war would not produce the clean victory Tel Aviv expected.Prediction (June 18, 2025): Israel has committed a strategic error. The math favors Iran. The war will not produce a clean Israeli victory.[6]What I got right: The math was accurate. Israel's interceptors were limited. Geography favored Iranian defense in depth. The war did not produce a clean Israeli victory. Iran's strategic patience thesis (outlasting democracies with shorter political timelines) was correctly identified as their core doctrine.What I got wrong: I underestimated the American contribution. Running parallel to the Israeli strike was Operation Midnight Hammer, a massive B-2 stealth bomber assault that dropped thirty-thousand-pound bunker busters on Iran's nuclear facilities and destroyed the planetary mixers required for solid-fuel missile production. I was analyzing Israeli capabilities. The United States changed the equation with tools I did not anticipate being deployed.[7]I published that accounting in full, seven months later, with the same rigor I applied to the original prediction.[8] The section was titled "What I Got Right, What I Got Wrong." Not buried in a footnote. Not reframed to minimize the error. Front and center, with an explanation of why the error occurred and what it taught me about the limits of my own framework.This is what separates Thesis Intel from punditry. A pundit who predicted wrong either pretends it never happened or argues they were actually right if you squint. I publish the scorecard.Why Nobody Else Does ThisThe answer is simple: the professional incentives actively discourage it.Think tanks survive on institutional credibility, which means never being caught wrong. Their reports are hedged with enough conditionals ("could," "may," "under certain circumstances") that any outcome can be claimed as consistent with the assessment. The Brookings Institution did not predict the fall of Kabul. CSIS did not predict the speed of Russia's failure in Ukraine. The Atlantic Council did not predict the Iranian protests of 2026. None of them published scorecards.Academic analysts write on timelines measured in years. By the time their papers are published, the events they analyze are history. Peer review ensures methodological rigor but eliminates timeliness. A professor who published a paper on Iranian deterrence strategy in 2024 has no mechanism to update that assessment in real-time as missiles are flying in June 2025.Financial analysts like Peter Schiff have predicted dollar collapse for twenty years.[9] The dollar is at historic highs against every major currency. Steve Keen predicted the end of the US global monetary system without accounting for compute as a strategic resource.[10] Neither has published a detailed accounting of where their models failed and why. The incentive is to keep selling the same thesis to an audience that wants to believe it, not to grade it honestly.Op-ed writers and Substack essayists (the closest analogue to what I do) typically offer narrative interpretation after events have already occurred. They are literary, not predictive. They tell you why something mattered, not what will happen next. This is valuable work, but it is a different discipline. A historian and a navigator both study the ocean; only one of them tells you where the ship is heading.Intelligence agencies do real-time predictive analysis, but their work is classified. The public never sees the CIA's assessments until decades later, when they are declassified and the stakes no longer matter. Thesis Intel is, in a sense, open-source intelligence analysis published in real-time for a general audience. The difference is that I show my work, take positions, and accept consequences for being wrong.What This Actually RequiresThesis Intel is not a gimmick or a branding exercise. The method demands four things that most writers either cannot or will not do.Primary source fluency. When I write about Iranian missile capabilities, I am reading IDF battle damage assessments, Institute for the Study of War operational updates, satellite imagery analysis from open-source intelligence communities, and OSINT channels operated by military observers in the region. When I write about Minneapolis, I am reviewing video footage frame by frame, cross-referencing DHS press conferences against the video record, tracking federal court filings, and monitoring local reporter feeds in real-time. The analysis is only as good as the intelligence feeding it.Structural incentive modeling. The core analytical framework is not ideological. I do not care what politicians say they believe or what they claim they will do. I care about what their structural position compels them to do. Trump's incentive in Minneapolis was to avoid a political catastrophe, not to win an ideological battle. That incentive predicted the off-ramp more reliably than any amount of cable news speculation about whether the administration was "tough" or "backing down." Incentives are legible. Intentions are not.Public falsifiability. Every thesis I publish can be checked. "Minneapolis will not explode" is falsifiable. "The situation is complex and could go in several directions" is not. I choose the former. This means I will sometimes be wrong, publicly and unambiguously. The Iran missile math was an example. But being wrong in a documented, analyzable way is more valuable than being vaguely right in an unfalsifiable way, because the documented error teaches something.Honest retrospection. The scorecard is non-negotiable. When I am right, I explain why the framework worked. When I am wrong, I explain why it failed. Both are published with the same prominence, the same footnotes, the same rigor. The audience learns either way.What I Am BuildingThis Substack publishes roughly two articles per week, each with 20 to 40 footnotes linking to primary sources. The coverage spans geopolitical conflict, economic policy, state violence, and institutional power structures. The common thread is the Thesis Intel method: identify incentives, publish predictions, grade results.For Founding Members, I code and write a monthly TATSU GEOPOLITICAL RISK ASSESSMENT: a structured report with weekly updates and scorecard checks against my own predictions. The closest equivalents are Stratfor and Eurasia Group (Ian Bremmer), which charge institutional clients upward of $40,000 a year. My Founding Members tier is $99 a month. That is 3% of the cost for the same product category, built by one person instead of a bureaucracy.In one year, I have published over 100 articles. The Iran series alone ran to six parts covering twelve days of conflict in real-time, with a full retrospective scorecard seven months later.[11] The Minneapolis coverage produced four articles in seventeen days tracking three federal shootings, a constitutional confrontation between state and federal authority, and the political resolution I predicted before it happened.[12]I write like this because nobody else does. The gap in the market is not for more opinion. It is not for more "explainers" that explain nothing. It is not for hedge-fund newsletters priced at four hundred dollars a month that cannot outperform a dartboard. The gap is for someone willing to publish specific, checkable predictions backed by primary intelligence, and then come back and tell you whether they were right.That is Thesis Intel. That is what I do.If you have read this far, you already know whether this is worth $8 a month.Notes[1] Ikeda, Tatsu. "The Clap After Alex Pretti's Death." Tatsu Ikeda Substack, January 26, 2026. Full investigation of the Pretti shooting and the political off-ramp that followed.[2] Ikeda, Tatsu. "Why Netanyahu, Trump, EU Lost Rising Lion Already." Tatsu Ikeda Substack, June 18, 2025. Original assessment of the Israeli-Iranian conflict with missile math and strategic predictions.[3] Ikeda, Tatsu. "When Khamenei Dies, the Middle East Burns." Tatsu Ikeda Substack, January 13, 2026. Seven-month retrospective on the Iran war including prediction scorecard, protest analysis, and dual succession crisis thesis.[4] "Videos Show Moments in Which Agents Killed a Man in Minneapolis." New York Times, January 24, 2026. NYT Visual Investigations verification of shooting footage.[5] Ikeda, Tatsu. "Disarmed, Then Executed." Tatsu Ikeda Substack, January 24, 2026. Article containing the off-ramp prediction published before the Trump-Walz phone call.[6] Ikeda, Tatsu. "Why Netanyahu, Trump, EU Lost Rising Lion Already." Tatsu Ikeda Substack, June 18, 2025. Published during active hostilities with specific predictions about war outcome.[7] "United States strikes on Iranian nuclear sites." Wikipedia, accessed January 2026. Documentation of Operation Midnight Hammer including B-2 strike packages and battle damage assessment.[8] Ikeda, Tatsu. "When Khamenei Dies, the Middle East Burns." Tatsu Ikeda Substack, January 13, 2026. Section VI: "What I Got Wrong, What I Got Right."[9] Peter Schiff has publicly predicted dollar collapse and hyperinflation since approximately 2006. The DXY Dollar Index reached multi-decade highs in 2025-2026.[10] Steve Keen, "The End of the US Global Monetary System." Naked Capitalism, January 2026. Analysis focused on debt dynamics without accounting for compute infrastructure as a strategic asset underpinning dollar demand.[11] Iran series: "Nearly WW3: Tehran and Tel Aviv Go Direct" (June 16, 2025), "Why Netanyahu, Trump, EU Lost Rising Lion Already" (June 18, 2025), "When Khamenei Dies, the Middle East Burns" (January 13, 2026).[12] Minneapolis series: Coverage spanning January 7 (Renee Good shooting) through January 26 (Pretti shooting and off-ramp), including "The Clap After Alex Pretti's Death" and the Uncivilized Country series. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit tatsuikeda.substack.com/subscribe