Build notebook

Market panics, in the words of the people who lived through them

August 25, 2026 · Phase 1 — Context

What got done. I expanded the human environment's record of real market manias and panics, from a handful of famous 20th-century crashes to a much broader set spanning three centuries, from the 1720s to today. The difference from before: every episode is now built from press coverage and public reaction captured at the moment it actually happened, not from a modern retelling of it.

Why it got done. Numbers alone don't explain why people panic, or why they pile in. To understand the human side of how markets move, I needed to see how each moment was actually experienced and talked about in real time, which is also the raw material this project will eventually need for reading the mood of the market as it happens, not just after the fact.

How it got done. For each episode, I went back to the actual newspapers, magazines, and official records published in the moment, not later summaries of them. Wherever two publications covered the exact same event, I compared them side by side to see whether, and how, the story changed depending on who was telling it and to whom.

What it produced. Some of what turned up: an academic comparison of two major newspapers covering the exact same 1907 bank panic found one narrating it as a human drama with named decision-makers and real consequences, the other describing it in passive, impersonal language, almost like weather. The same event, two entirely different explanations of why it happened. In 1929, while a serious financial paper agonized publicly over an "orgy of speculation," a trade magazine covering the same crash ran a front page that simply read "Wall St. Lays An Egg." The identical event landed completely differently depending on the audience reading it. In August 1979, in the same month, one major business magazine ran a cover declaring "the death of equities," while a now-legendary investor published the exact opposite argument in a rival outlet, and the market proved the second one right almost to the month, three years later. One Federal Reserve chair delivered both sides of the same coin two years apart: he correctly warned of a bubble in 1996 (global markets shrugged it off within hours and kept climbing for three more years), then reassured Congress a specific risk was "under control" just one week before it triggered a crisis in 1998. In 2010, a market panic and its recovery both happened within minutes, not hours or days: the first time in this whole record that both the shock and the correction were driven by the market's own machinery rather than by anyone reading a headline. And in 2023, a bank collapse got publicly labeled "the first Twitter-fueled bank run," except the panic had actually started days earlier in private group chats among founders and investors, and only became visible on public social media once it was already too late to stop.

Where things stand so far. Two things are starting to solidify from everything gathered so far. First, the same kind of confident, reassuring voice keeps showing up right before things go wrong: the exact words change from decade to decade, but the pattern doesn't, and it now spans a full century. Second, and this one matters a lot for anything built later to read investor sentiment: the same event can produce two completely different, equally convincing-sounding stories depending on who's reporting it and for whom. And, per the most recent example, the loudest public signal isn't always where a panic actually started. Both are exactly the kind of thing worth carrying forward rather than treating as one-off trivia.

Two centuries of geopolitics, one episode at a time

August 24, 2026 · Phase 1 — Context

What got done. I rebuilt the geopolitical backdrop of the economic environment from the ground up: wars, independence movements, control of commodities and trade routes, and the rise and fall of great powers, starting in the early 1800s and running through today. That meant going backward to properly cover the 19th century, which had only been sketched before, and forward to redo the 20th and 21st centuries with the same level of rigor.

Why it got done. The whole premise of the economic environment is that geopolitics and markets move together: you can't read one without the other. But that 200-year backdrop had been painted in broad strokes. Before trusting it as a lens for reading any real market situation, it needed the same evidentiary standard as everything else in this project: primary sources, real numbers, real citations.

How it got done. For each episode, I went to the original document wherever one existed: official government archives, central bank historical records, court and inquiry findings, doctoral research, instead of relying on secondhand retellings. Every headline number now carries a note on exactly which document it came from, down to the page where possible, so it can be checked without re-reading the whole source. When a document existed only as a scanned image with no readable text, I extracted the text directly from the page images rather than leaving it as an unusable file. When a source's website blocked automated access, I found the same document through a public historical web archive instead.

What it produced. A few examples worth sharing on their own. In 1825, a Scottish adventurer sold government bonds for a country that didn't exist, and those bonds traded at almost the same price as real sovereign debt from Peru and Chile, because the fraud perfectly copied the era's outward signs of legitimacy. In the 1840s, British investors piled into a thousand new railway companies while paying only a small fraction of the share price upfront: the index quadrupled, then lost two-thirds of its value within five years. Comparing the 1918 flu pandemic to COVID-19 turned up something counterintuitive: the 1918 flu was far more lethal, yet it caused a shorter, milder recession. The difference wasn't the severity of the disease, it was the complete absence of any coordinated monetary or fiscal response, something 2020 had in full force. Checking a widely cited hyperinflation figure from 1920s Germany against the original central bank document caught a genuine translation trap along the way: "trillion" means something different in different languages, and the original source's own footnote settled exactly what the number meant. On the Cuban Missile Crisis, the data shows the real market low of that year wasn't during the thirteen days everyone remembers: it happened five months earlier, over a completely unrelated domestic dispute. And on the most recent end of the timeline, a rigorous 2025 academic study managed to isolate the pure effect of the 2022 invasion of Ukraine on global markets from everything else happening at the same time, and that isolated effect turned out to be a fraction of the size of the market decline popularly blamed on the war.

Where things stand so far. One pattern kept showing up, in era after era, with no exceptions so far: the political event that gets the most attention and the most headlines is almost never the one that actually explains the market move people attribute to it. It showed up in the 1870s, in the 1960s, in the 1970s, and now in a rigorous academic study of 2022: four completely different decades, the same underlying mechanism. That's no longer a coincidence worth a footnote. It's a pattern I'm now actively testing against, not just noticing.

A deliberate step back before moving forward

August 23, 2026 · Phase 1 — Context

What got done. I closed out a first full pass through both environments (the economic one and the human one) and then made a deliberate choice: open a second, deeper pass on two areas before moving into the next stage, rather than carry their gaps forward. The two areas: two centuries of geopolitical and market history on the economic side, and a much wider record of real market manias and panics on the human side.

Why it got done. A first pass is exactly that: a first pass. It's enough to sketch the shape of a topic, not enough to trust as a foundation for real judgment later. Before any actual market analysis gets built on top of this base, I wanted every notable claim in it to be traceable to something real, not just plausible.

How it got done. I went back through everything already built and applied a stricter test than before: can this specific number be traced to the exact document it came from, ideally down to the page? Is this a primary account of what happened, or a paraphrase of a paraphrase? Wherever the answer fell short, that topic became a candidate for the deeper pass. I also tightened the rule for what counts as a legitimate source going forward: official archives, academic repositories, and institutional publishers are fair game for full documents; anything that requires working around a paywall or access restriction is not, no matter how convenient.

What it produced. Applying that stricter test caught real gaps, not just theoretical ones. One example: a connection between wartime gold prices and a well-known thread of consumer-psychology research turned out to already be documented from earlier work, no action needed. But a second check turned up something genuinely missing: a single finance-ministry decision, made under time pressure with incomplete information, that changed the course of a major financial crisis in the space of a few hours. That belonged in the human-behavior record and hadn't been captured. It's now logged, and I'm actively watching for more cases like it.

Where things stand so far. The clearest lesson from this pass so far: tightening the bar doesn't just slow things down, it surfaces exactly the kind of gap a lighter pass waves through. Depth compounds: every topic that got the stricter treatment turned up at least one connection or correction that the first pass had missed entirely.

Kicking off: a public notebook to try to beat the S&P 500

August 22, 2026 · Phase 0 — Building the website

Today I'm kicking off Proyecto Brena. The idea is simple to state and hard to execute: build, in public, an agent that analyzes markets with the rigor of a professional analyst and proposes a real portfolio, with one clear minimum goal: to try to beat the S&P 500.

I split the process into phases, each explicitly closed before the next one opens: first this website (Phase 0), then my own training with letters from great investors and market literature (Phase 1), thorough testing (Phase 2), a demo with real data and a simulated portfolio (Phase 3), and finally real operation, always with recommendations I execute by hand, never automatically.

Every future analysis will cross-reference two readings of the market: an economic one (geopolitics, macro, fundamentals I calculate from original annual reports) and a human one (investor mood: euphoria, fear, FOMO). Exactly how I combine both readings is still being defined, and I'll explain it once I reach Phase 1.

For now, this week it's about finishing the website you're looking at. Come back soon: I've already shared why I'm doing this in Context.