This site is a Work In Progress

Site Navigation

NeoBooks In Progress

Creating a NeoBook (start here)

Admin and Help

Edit on GitHub


SJB Post: "What Karpathy's Big Idea Means for People Like Me"

Revised video outline — warmer, less technical, same excitement

The Hook (30–45 seconds)

One of the smartest people in AI just described something I've been trying to do for 28 years — and also something I think could change how we all think together.

Let me explain what he said, why it got me excited, and why it might matter to you even if you couldn't care less about AI.

Part 1: The Problem Everyone Has (2–3 minutes)

[No screen share needed here — just talking]

Here's a thing that happens to everyone who reads, listens, and learns.

You read a great article. You take notes. Maybe you even tell someone about it. And then — six months later — it's gone. Not the article. Your understanding of it. The connection it made to three other things you'd read. The question it raised that you never answered. Gone.

You start over every time.

Most people use AI the same way. You ask a question, it gives you an answer, you close the tab. Nothing builds. Nothing accumulates. It's like having a brilliant conversation at a party and then both of you developing amnesia.

Andrej Karpathy — one of the people who helped build modern AI — just published a short essay describing a different approach. A way to make knowledge actually stick and grow over time, with AI doing the boring maintenance work that humans always give up on.

I read it and immediately thought: this is what I've been reaching for.

Part 2: What I've Been Doing (2–3 minutes)

[Optional: briefly show Jerry's Brain graph view — just for a moment, don't explain it technically]

For 28 years, I've been keeping a public notebook of everything I think about. Not a diary — more like a map. Every idea, every connection, every person I find interesting, linked to everything else.

It's called Jerry's Brain, and you can actually look at it online. It has over 600,000 nodes at this point.

I love it. But it has a real problem: I can't easily ask it questions. I can't say "hey, what do all my ideas about trust have in common?" and get a useful answer. And every time I sit down with an AI like Claude, I have to re-explain myself from scratch. It doesn't know what I know. It doesn't remember what we figured out last time.

Karpathy's essay describes a way to fix that — not just for me, but potentially for anyone who thinks in public.

Part 3: The Simple Version of His Idea (3 minutes)

[Maybe a simple hand-drawn diagram or whiteboard — three boxes: Sources → Wiki → You]

Here's the idea, stripped of jargon:

Most of us collect things — articles, notes, highlights, bookmarks. They pile up. We rarely go back to them. When we ask an AI about a topic, it's reading from its general training, not from the specific things we've read and thought about.

What if instead, every time you read something important, an AI read it with you, pulled out what mattered, and wove it into a living document that kept getting richer?

Not a pile of files. A connected document. Where the AI has already done the cross-referencing. Already noticed that this new article confirms what you read six months ago. Already flagged the contradiction between what two sources say.

Karpathy calls this an LLM Wiki. You feed it things you care about. The AI maintains it. You read it, explore it, ask it questions. And crucially — it keeps getting smarter as you do.

The part that excited me most: the AI does the maintenance that humans always give up on. Cross-referencing is boring. Updating summaries is tedious. Nobody does it. AI doesn't get bored.

Part 4: Why This Is Personal (2–3 minutes)

I've spent 28 years doing this manually — building connections, maintaining my map, doing the cross-referencing myself. It's been worth it. But it's also been exhausting.

What Karpathy is describing is an AI-assisted version of exactly what I've been doing. Which means I'm in an interesting position: I'm not starting from scratch. I already have a 28-year head start on the thinking. I just need to figure out how to make it more alive, more queryable, more useful — for me and eventually for others.

I've already started. My notes in Obsidian are becoming the working layer. Claude is helping me maintain them. It's early and imperfect, but I can already feel the difference between a Claude session that knows my context and one that doesn't.

It feels like the difference between talking to a colleague who knows your work and talking to a stranger who's very smart but has no idea what you've been up to.

Part 5: The Bigger Thing (2–3 minutes)

[This is the SJB provocation — keep it short and open-ended]

Here's where it gets interesting beyond just me.

What if this isn't just about individual notes?

I've long believed that one of our biggest problems — as communities, as a society — is that we don't have good ways to think together. Everyone's building their own map in their own silo. We don't encounter each other's thinking. We don't build on each other's work.

I've been calling my version of a solution "Big Fungus" — after the way fungal networks connect trees in a forest, letting them share nutrients and signals. The idea: what if people using different thinking tools could share a commons layer? What if your notes and my notes could inform each other, even though we use different tools and have different styles?

Karpathy's maintenance insight changes the equation. The reason shared knowledge projects always die is that nobody wants to do the upkeep. It's the same reason Wikipedia took a small army of volunteers. But if AI handles the maintenance — the cross-referencing, the updating, the connecting — maybe shared thinking at scale becomes possible for the first time.

I don't know if that's true yet. But I'm going to find out in public.

The Close (45 seconds)

What Karpathy described is a tool. What I'm interested in is what the tool makes possible — not just for me, but for how we think together.

If you've been building your own map of ideas — privately or publicly — I want to hear from you. And if you've been meaning to start but never quite did, maybe now is the moment.

That's what Stump Jerry's Brains is really about: thinking in public, together, with stakes. Come find me.

Recording Notes