547: Citadel & Enron, OpenAI vs DeepSeek, Amazon Prime, Short-Sellers, Sonos, Duolingo, Anthropic, Black Holes, Vitamin D, and AlphaGo

"We’re like reverse Jenga towers."

What separates people is the return on luck, what you do with it when you get it.

—Brent Schlender

🏗️🧱 My colleague Dylan (🇮🇪🇪🇸) recently posted a reflection on a C.S Lewis quote:

Thinking about that C. S. Lewis quote about how courage is not just one of the virtues but the spine of every virtue, and how brilliantly/brutally it frames how virtuous we like to think we are

You're kind, but are you kind when it's hard to be kind? Or just when it's easy

This sparked a metaphor in my mind:

We’re like reverse Jenga towers.

Instead of starting complete, we begin with missing blocks. They represent undeveloped virtues, and we have to spend our lives trying to add them back to make the tower stronger.

You could think of the blocks as the old-fashioned virtues like:

• Wisdom (Prudence), Justice, Courage (Fortitude), Temperance (Self-Control), Curiosity, Critical thinking, Discernment, Open-mindedness, Rationality, Creativity, Intellectual honesty, Integrity, Honesty, Truthfulness, Loyalty, Honor, Responsibility, Accountability, Reliability, Trustworthiness, Kindness, Compassion, Empathy, Generosity, Hospitality, Courtesy, Respect, Tolerance, Forgiveness, Gratitude, Humility, Gentleness, Perseverance, Resilience, Determination, Discipline, Patience, Diligence, Industry (Work ethic), Self-improvement, Adaptability, Growth-mindedness, Imagination, Frugality, Authenticity, Eloquence, Clarity, Professionalism, Excellence, Mindfulness, Wonder, etc

How many of these blocks are in your tower, and which ones are you working on adding next?

🤖🤔🛀💭🚢⚓️ Recent advancements in AI, such as DeepSeek and OpenAI's Deep research, have made me consider the implications of AI for what I’m doing here.

Here's where I've landed, so far: I want to be *augmented* by AI tools (in the same way that writers have historically benefited from word processors and Google), but that the output of the process has to be as differentiated as possible from vanilla AI outputs. I'm doubling down on being my authentic human self and letting my personality shine through.

There’s a real risk that a lot of writers will all start sounding the same all of a sudden because they rely too much on AI tools, and that’ll create a “blandification” of content (more than already is the case 😅). All spiky idiosyncracies 🦔 and rough edges will be sanded off until most people appear to be polished pool balls. 🎱

But you know what? To paraphrase John Blackthorne from Shōgun… screw that! I’ll be the guy over here doing my thing, and if it resonates with you, awesome, welcome on board.

I’ll be your Anjin (あんじん)! 💚 🥃

🛀💭🤖 I've been using Gemini 2.0 Pro Experimental for a few days and I'm impressed.

For the first time, I'm considering using a Google model instead of Claude 3.5 Sonnet as my default (this may change if Anthropic ever releases Opus 3.5 or Sonnet 4.0 or a reasoning model..).

It’s very good. The “flavor” is a bit bland but doesn’t get in the way, and it’s remarkably fast considering how big it is likely to be — that’s not public information, but I’m guessing it must be a Mixture of Experts (MoE) model that adds up to somewhere in the ballpark of 1-2 trillion parameters.

In parallel, I've been testing DeepSeek R1 (versions hosted in the US, not in the DS app) and it’s also very good and quite fast considering how many tokens it generates.

---

🏦 💰 Business & Investing 💳 💴

🛒🇺🇸 Incredible: Almost 200m shoppers are using Amazon Prime… just in the US

I mean, at this point, who’s left..?

The Prime membership turned 20 this year, and while it is based on the Costco membership — which, depending on what you count as the beginning, is 40+ years old, and has around 130 million members worldwide — it has morphed over time into a lot more than just a paid shopping club.

Amazon disclosed:

[In 2024,] Prime members worldwide saved nearly $95 billion on fast, free delivery. In the U.S., that means Prime members saved on average over $500 on their deliveries last year—nearly four times the cost of an annual membership fee.

They’re going to use assumptions to make that number as big as they can. I’m pretty sure that without Prime, the average shopper wouldn’t buy as much, so they wouldn’t have spent $500 on shipping…. But that’s the point! Prime members shop more because they have Prime!

🦴💰☠🔍🕵 The Tale of How Citadel Picked the Bones of Enron and made $30bn + John Arnold’s Ken Griffin Story

This is an incredible two-part story 🤯