Welcome to the 70 new members of the TOC community!

I hope all 9602 of you had a great week 🔥

A few years ago, a mentor who has been active in the tech space since the 2010s told me that what a lot of people don't know about Marc Andreessen is that he's one of most prolific WhatsApp group chat creators & curators out there. He's started hundreds of chats for all kinds of topics with a variety of people to make sure he's in the know about everything.

To be clear, these groups he's making aren't simply just for "s-tier influencers and founders". Of course, he has access to and conversations with them as well. But from my understanding, it's more so Andreessen's ability to identify people who may not have a public presence but clearly know something about a niche topic that he wants to keep tabs on. And it's through these light green texts where he gets his head start on understanding what everyone else finds out a few months (or years) later.

In a post from December, I mentioned the following:

I titled this article "What did Marc Andreessen see?". And to be honest, I think it made the most sense.

It's amazing to me that pmarca was not only able to sniff out Truth Terminal months before anyone else but also strategically talk to the agent and give it a $50k grant in BTC. We all know this lore by now but the impact of it only gets stronger as the new meta continues to increase in relevance.

Marc understood that there's deep value in giving grants to these hackers and letting them experiment at the frontier of new technologies.

When I first started going down the agent rabbithole in September, it was wild to see that some of the most random accounts with 200 followers had pmarca following them but not a single other person I follow!

Anyways, what's the point of this Andreessen group chat fanboying?

Not my own insight, but something I saw both Tyler Cowen and Ben Thompson say this past week. I spent some time to digest it and the point really started to make sense.

The concept itself is not complicated or anything but I want to explain my chain of thought 😉 to all of you because of how important I think it is.

In my last post, I made it clear that I needed to "learn smarter" with frontier models:

So last week I bought the $200 OpenAI deep research subscription.

There's no excuse for me to have extended timelines on learning new material. It's cope. There's a hammer available to me but I'm still banging the nail with a rock.

The goal is to learn the smart way and be fast about it.

Those who quickly adapt to the new methods of consuming and learning information will have a headstart on the way things are headed. And it's clear that even a few months of using these tools before others is enough to make the margin larger than most of us realize.

I've really stuck to that promise. In the past two weeks, I've been very intentional about using o3-mini, o1, and deep research as much as I can. In fact, this past week, my deep work session went something like this:

1. Read deep research report generated from day before (I'm most sharp with my first coffee)

2. Explore all the links that were cited in the report and draft a tweet or two as "my notes"

3. I then dump all my thoughts to o3-mini and ask it to find a common thread with what's on my head from all the reading. Based on this context it has, I have o3-mini generate a prompt for me that I then give to o1

4. Then I have o1 give me a couple of different angles on how I can explore the given topic. I'll have it generate 1-2 pages that cover the main bullet points and context. From there, I'll pick and choose what I think is most important and then have o1 generate an essay of a prompt that I can feed in to deep research

5. Generate deep research report for the next day's morning reading

It's not a perfect process by any means. But given the changes I'm seeing with my work flow already, it's clear that if I continue refining my approach...there's no going back.

Here's the thing. Right now deep research costs $200 / month and definitely takes some prompt tinkering to get truly personalized reports that hit the nail on the head.

In other words, it's not accessible to most people right now.

But it's only a matter of time before it is. It's not that crazy to think that the current deep research capabilities we're seeing right now will move to the general, cheaper subscription by the end of 2025. And the premium version will be a 5x better reasoning model or whatever upgrades SF is cooking up right now.

So what are the implications?

Well, as the number of people using advanced reasoning tools grows, there starts to be less of an advantage in your ability to curate and produce high quality research on any given topic that's on the internet.

Key words being "on the internet". Whatever frontier models can train on is fair game.

YB

@yb_effect

yea definitely. but i think the importance is orders of magnitude more bc it's a lot easier to "know everything".
there used to be an advantage in being able to surface valuable info in the public domain, but that seems to be going down. which imo makes the "secrets" stronger

1

2:44 PM • Mar 14, 2025

Or as Tyler Cowen beautifully puts it in an interview with David Perell (a bit long but just read the damn quote it's important)

Humans know secrets; maybe AIs can be fed secrets, but they don't in general know secrets. Now, a human only knows so many secrets—that's partly where decentralization comes in.

How AIs will handle secrets, I think, is a big and interesting question; it's somewhat under-discussed.

It seems like in the Peter Thiel definition of a secret, which is something you know about the world that other people don't know, there's a chance that those go up because now there's less of an incentive to put things in the public domain since they can spread so much faster.

So there might be more of an incentive to hoard information. It will be worth more to you because the public information you used to hold is now worth very little. The future—the AI-rich future—is also a world replete with secrets.

Secrets are super important; gossip is very emotionally and practically potent. It's another part of this new structure we're not ready for. We got to talk more about this: why social networks are more important now.

How good are you with secrets? Are you good at trading secrets? If you are, you're a lot more productive than you used to be. You ever have these conventions with your closer friends like, 'I'll tell you this secret, it's not quite a deal, but it's understood that you'll tell me that secret in return?'

Increasing returns to social networks—social networks become way more important as well.

And very similarly, Ben Thompson mentions this in a post of his covering deep research

One additional point he makes that piqued my interest was how these "secrets" are actually a huge bull case for prediction markets. If this information will go into the public domain, how can those with the information asymmetry capture some of that value?

YB

@yb_effect

never thought about AI & prediction markets in this way

bc of deep research like tools that will only get better, the advantage is in what information isn't publicly available

PMs are the best way for people to profit from secrets before they're in the training data

11

7:46 PM • Mar 12, 2025

For what it's worth, the whole concept of defining insider trading in this new paradigm of information markets is still in its earliest stages. It'll be interesting to see what happens in terms of legal and ethical standpoints going forward.

I did want to share a brief exchange I had with my friend who was questioning the above tweet:

So then, the natural question should be "how do I acquire more secrets?"

• Learn how to rally an IRL community through meetups

• Be selective about which events you attend, spend your social battery wisely

• Go travel to different parts of the world

• Get closer to the gossip

• Be the reason people connect, make yourself a critical node in your greater social graph

It's having the "create your own yacht" mentality - find ways to bring interesting people to your home turf and share things.

What are the key takeaways from all this? For me, three things.

Embrace what is most +EV in the new world

YB

@yb_effect

seems like the two advantages going forward are:

1. how much better can you prompt and have models work together (multi-step orchestration)?

2. what trade secrets / alpha do you know that isn't already in the public domain?

10

2:22 PM • Mar 14, 2025

Make my own yacht

I live in NYC, one of the most (if not the most) important cities and hubs for crypto. But all I do is go to the occasional meetup hosted by VC funds.

So why not host my own events? To keep it simple, I've challenged myself to put together an IRL social for NYC crypto folks by the end of March. I'll hold myself accountable by sharing a picture before my first April post. I'm not exactly sure about my execution, but I'll figure it out - time constraints will solve everything.

⚠️ If you're in NYC and want to come, please reply to this e-mail ⚠️

Learn about the privacy side of DeAI

I need to spend time learning what products are working on the importance of handing secrets between AIs and humans on a decentralized network

YB

@yb_effect

so far, I've only been focused on the pre-training side of DeAI but I'm realizing I also need to spend time on the verifiability & privacy side (zkml) as well

basically how MPC, FHE, etc. come into play

"how AIs will handle secrets is somewhat under discussed" - Cowen

12

7:31 PM • Mar 12, 2025

That's all for today's post, I hope you all have a great weekend!

See you next week 🤝

- YB