Third Gulf War
Schuyler's LLM-powered hypothesis interrogation machine, built as a wiki tracking a hypothetical (or not) 2026 US-Iran war. This is not a prediction market or a news aggregator -- it's a structured analytical framework that tracks hidden variables, maintains time-stamped hypotheses with confidence levels, and recalibrates scenario probabilities daily. It's one of the most sophisticated examples of using LLMs for real-time geopolitical analysis.
Key Quotes
IC warnings systematically overridden at every stage by "one person's psychology filtered through an ideological gatekeeper."
Key Themes
The analytical architecture is the real contribution. Four hidden variables (P1-P4) drive the entire framework: Trump's political tolerance as a time constraint, the interceptor depletion race, Iran's infrastructure strike intent, and economic transmission mechanisms. Three content layers (narratives, hypotheses, trends) organize different types of analytical output. Three scenario bands (B1/B2/B3) with daily probability updates provide the probabilistic framing.
The hypothesis tracking with confidence levels and validation against emerging events is rigorous in a way that most geopolitical commentary isn't. This is closer to intelligence analysis tradecraft than journalism.
The "oil price fiction" analysis -- four reinforcing elements (reporting infrastructure, political convenience, cognitive intolerance, model miscalibration) sustaining a systematic disconnect between market pricing and actual supply disruption -- is the kind of structural insight that emerges from sustained analytical attention rather than hot takes.
Critical Analysis
The war profiteering documentation ($920M crude oil short placed 70 minutes before ceasefire reports, $125M profit by market open) is the kind of detail that makes this wiki genuinely valuable as a historical record, regardless of your views on the underlying analysis.
The framework is impressive but raises the question: how much of the analytical rigor is the LLM's contribution versus the human analyst's? If Schuyler is providing the hidden variables, the scenario structure, and the daily interpretive judgment, the LLM is mainly helping with information processing and organization. That's still valuable, but it's a tool for an expert analyst, not a replacement for one.
The daily update protocol with probability recalibration is the right approach for evolving situations, but it requires sustained human attention. A wiki like this is only as current as the analyst maintaining it.
This is a unique artifact -- there's nothing else quite like it in the AI tooling space. It's not an agent framework or a developer tool; it's an example of what happens when a skilled analyst uses LLMs as a thinking infrastructure for real-world problems.
Sources: Third Gulf War Last updated: 2026-05-14