# If You Use AI for Work, You Need a Second Brain

https://www.youtube.com/watch?v=nTiMbqFwv4c

[00:00] The number one trend in AI right now is the rise of the AI second brains.
[00:04] People building personal operating systems for themselves, which are helping them to do better in their jobs and accelerate their careers.
[00:12] In this episode, I'm going to give you the quick introduction to what the AI second brain is and actually give you a demo of my AI second brain, so you can figure out if it's something you want to build for yourself and I will give you some tips on how to get started.
[00:26] When Andrej Karpathy posted about building an LLM wiki all the way back in April, it got 20 million views on X and ever since then he has set off the number one craze in AI.
[00:43] Today, something is really happening in tech.
[00:45] The smartest people are racing to build personal AI systems that remember everything they know and today I'm going to show you why that's happening, give you the best and most simple explanation of what the AI brain
[01:01] Is and give you a demo of my personal AI brain, so you can see how this could work for your life and to optimize your entire job.
[01:14] Okay, so what's going on?
[01:14] In the past couple of months, there has been a pattern that has emerged and it's really not just for AI geeks like myself or just AI developers.
[01:22] It's people [snorts] running real companies, making real decisions, CEOs, founders across the board.
[01:26] It is the AI second brain.
[01:30] So, Andrej Karpathy, the former director of AI at Tesla, he's the co-founder, former co-founder of OpenAI, now at Anthropic, built what he called an LLM wiki.
[01:39] So, this was his explanation where you would have a simple folder where you could drop sources of information into that folder.
[01:47] The AI would read it all and it would build a little wiki and anytime you talk to the AI, it could access that wiki.
[01:54] So, you give it some real intelligence, you give it some persistent memory.
[01:58] At its kind of most basic level, this is
[02:01] What he did.
[02:02] He had a raw folder, and inside that folder, you could drop lots of information.
[02:07] The AI would process that information, and would create a wiki for itself.
[02:13] The LLM wiki kind of extrapolate in all of the intelligence or knowledge you would need to know.
[02:20] And anytime he talked to his AI assistant, it has access to that knowledge, and so you're continually talking to your AI assistant, and it's getting smarter and smarter over time.
[02:30] It's gathering more and more knowledge over time.
[02:34] And this is not just confined to people like Andrej Karpathy.
[02:37] There was a really good post from Brian Halligan, the former CEO and founder of HubSpot.
[02:44] Um, obviously know Brian really well, and he built — he had an X thread, where he talked about Hal.
[02:52] Wait, who is this?
[02:54] Uh, that's Hal.
[02:55] I told you about him before, but he's sort of my assistant clone second brain.
[03:02] Hal is actually a real step up in terms of AI second brains.
[03:05] He has built an AI agent that will not just have access to all of the things that Brian has access to, but can actually run his day.
[03:15] It sends emails.
[03:18] It organizes projects.
[03:20] It goes to Zoom meetings.
[03:22] I hear it is here looking like a Terminator.
[03:24] I'm not sure what Brian decided to make his look like a Terminator, but there it is.
[03:27] It's on the meeting, and can talk to you and participate in that meeting.
[03:31] Brian's system is actually pretty wild.
[03:34] It's given me some inspiration for my V2 of the second brain that I'm building.
[03:38] I covered it on my Substack here.
[03:41] This is a little graphic of what Brian has built.
[03:45] So, it reads a bunch of information, your Gmail, your calendar, your drive, again, granola, all of the information it needs.
[03:51] And it has a core brain.
[03:54] It memory, it has tools that it can use like web and research.
[03:56] Can do a bunch of different things, can join a meeting, and then produces things for Brian.
[03:59] Very similar to my brain.
[04:01] An example that I'm going to show you.
[04:03] That mine's not cool enough to join meetings just yet.
[04:04] It is getting there.
[04:06] If you're enjoying this episode and you want all of the resources, you can click the link in the description or scan the QR code to get them all delivered to your inbox.
[04:15] Another well-known influencer, Tiago Forte, you might know him.
[04:18] He literally built a book called The Second Brain.
[04:22] Now, originally that was a note system.
[04:25] He actually was doing that with a bunch of post-it notes.
[04:27] Now, he has retrofitted and re-released that as a course on how you can create an AI second brain.
[04:32] And this goes on and on and on.
[04:35] There is just a ton of momentum around Jason Lemkin of SaaStr.
[04:39] He built an AI version of himself.
[04:41] So, here's Jason Lemkin.
[04:42] This was kind of more the traditional version of a brain where it captures all of the intelligence that Jason knows and you can say, "Hey, should I use a SaaS product or an AI agent?"
[04:52] And then it will research all of the things that Jason knows and give you some answers.
[04:57] Now, I think this is kind of the original product that what people were building around second brains, which is it would scrape all your external data and then kind of
[05:04] Create a version of you and answer questions like you.
[05:08] I think what I'm going to get into is like the much more interesting version of second brain where it really becomes your personal operating system.
[05:16] That's what I'm fascinated by.
[05:19] I'm not as fascinated by AI that kind of acts like you online because it's retrieved all of the information.
[05:25] So, why is this becoming so popular so quickly?
[05:27] Of course, it takes someone like Andre Karpathy to showcase different things that are possible with AI.
[05:33] We're still really in the infancy of how we use and understand AI.
[05:35] The idea of a second brain is not new.
[05:38] It is something that we have all dreamt about.
[05:41] That's why Forte's book he sold over 500,000 copies of the book A Second Brain.
[05:46] Before that, people tried wikis, they tried Notion, they tried all different types of things to kind of keep notes and jot down what you've learned and put it into a system that makes it easy to retrieve.
[05:58] And the problem is that nobody does that.
[06:00] No one really maintains it, it decays, your system goes out of date, it's too problematic.
[06:03] It takes too much.
[06:05] Effort.
[06:08] Again, to showcase this is not a new concept.
[06:10] The actual concept isn't new.
[06:13] In the 1980s, Ray Dalio actually started writing down every decision he made and the reasoning behind it.
[06:17] Ray Dalio is an incredibly famous investor and founder of Bridgewater, the investment firm.
[06:25] Over decades, that became his principal system.
[06:27] It was really a codified repository of how he thinks.
[06:33] And when new employees joined, they could study it and they could really learn pretty rapidly from how Ray Dalio thought and made decisions.
[06:39] And any decisions that someone wanted to make in the company, they could cross-reference it against what would Ray do in this situation.
[06:46] And it worked, right?
[06:48] But it also required decades of manual effort and a whole team to maintain it.
[06:52] And he could do that because he runs a $150 billion fund.
[06:55] So, a lot easier for someone like that to maintain it, but for the ordinary person, much much harder.
[07:00] So, what changed is really AI has removed the maintenance to ingest and enrich and
[07:08] Keep track of all of the knowledge and intelligence you're gathering across your lifetime and all of the different jobs that you do.
[07:18] The AI reads, writes, organizes, links, and retrieves.
[07:20] It does all of that things.
[07:22] And the other thing that happened is AI context windows started to get big enough to be able to do this.
[07:28] Agents got capable enough, tools like Claude Code made it easy to point to a folder, make real sense of it.
[07:34] So, you have context windows getting bigger, agents getting more and more capable, better reasoning, and you have tools like Claude Code work and Claude Code and Codex that could basically attach itself to a folder and start to make real sense of how all of these things should work.
[07:50] But, that isn't the best thing about having an AI second brain.
[07:52] It's not just can I ingest all of this information, can I keep track of it, can I build links between it so the AI knows that this thing is related to another thing, it's that over time it compounds and it gets much much more intelligence.
[08:07] Anyone I talk to talks.
[08:10] About the same feeling.
[08:11] It's like having a version of yourself that never forgets things and gets smarter over time.
[08:16] Like, if you were ingesting all of the information that you come across on a day-to-day basis, remembering it and able to reuse it, you're just going to be much much smarter.
[08:26] And so, if you are a marketing leader running a team or a founder running a business, it is a place to keep all the intelligence you gain through the years of work, what worked, what hasn't worked, why you made certain decisions, why certain things were successful.
[08:39] It's all of that context and it lives there in your AI second brain.
[08:44] It compounds the more that you use it, gets much more smarter, and then anytime you're using your AI assistant, your AI assistant has access to that intelligence and can use it to actually help you do the things you want to do.
[08:56] Let me give you a quick hit on some of the things to think about that other people may not be talking about when you are reading about AI second brains online.
[09:04] You're probably reading a lot about should I use Obsidian or should I use Notion or should I use Google Docs?
[09:11] Well, there's a couple of other things that I would just point out that are good notes to take if you are truly interested in building an AI second brain that really helps you, that becomes the personal operating system for how you do work, right?
[09:25] It's really your personal operating system for how you do work.
[09:27] A lot of people think of it as just storage and search, right?
[09:32] What most people are building is dumping everything into a folder, you can search notes with the AI, you call it a second brain, and the knowledge really decays over time, right?
[09:42] It doesn't stay up to date, it's not dynamic, it's not learning over time.
[09:46] The hard part really is the retrieval and evolution, and I'm going to show you how this looks in my demo.
[09:52] The reason you would want to build a second brain is not for storage and search.
[09:55] You can really do that through a cloud project.
[09:57] You can point it at a bunch of docs.
[10:00] You can query those docs.
[10:02] And it's pretty good at doing that.
[10:03] The reason you build a second brain is because the AI is being enriched over time with the things that you're learning across your teams, across your work, and it's able to connect ideas.
[10:12] that you have never thought of yourself.
[10:14] It's able to actually look at contradiction in that you're making across the work that you do.
[10:21] It's always up to date.
[10:23] And I'm going to show you how that works.
[10:25] It in it enriches every single day.
[10:27] And has the latest information.
[10:29] And it really is a great system for telling you where ideas and decisions and things like that originated from.
[10:36] The first thing I will say is people probably are not talking about the value of the AI second brain as much in terms of being the retrieval and how it compounds over time.
[10:45] And the second part is the really interesting in that I'm talking uh really about your personal OS.
[10:49] How it is mapped to all of the work that you do.
[10:55] But there's also the concept that there is a team brain where you want your team connected to the same system of intelligence.
[11:03] You want to connect it to the same quote unquote brain because all of those learnings all of the things that you're capturing are as applicable to your colleague as they are to you.
[11:14] A company org brain.
[11:16] Like that is really how companies will likely differentiate themselves and have leverage in the future in that they, their in raw intelligence of the company, the things that the company know and have figured out that no one else has is really their asset that they can plug and play into AI and bring to life.
[11:35] And so, the hard thing that no one has figured out, and I've spent a lot of time talking to people in the space, is how do you build a system that easily helps you navigate between your personal brain, your team brain, and your company brain.
[11:49] So, there are two interesting concepts to consider when you are building your own AI second brain in that you want it to be really good at ensuring it keeps up to date with the latest information, that it's really gathering real intelligence and knows how to make connections across all of the things that it's learning, that it's able to provide strategic insights and show you where contradictions are.
[12:15] Being made across all of the things that it's capturing, and that I would start with my personal brain, and then if you are get really into this, you're going to start to figure out, "Hey, how do I build something for my team and my company?"
[12:30] So, let's get into a quick demo.
[12:34] Now, this is a prototype because I can't show my second brain right now because it really has access to everything, and I could not figure out a way to show it to you all without disclosing a lot of information that I don't want to put on a YouTube channel.
[12:45] So, instead I've replicated it with fake data, and luckily the AI tools are pretty great at doing that.
[12:51] But, I can show it to you so you can see concepts and start to like figure out how this works.
[12:56] Now, the cool thing is the AI second brain is mapped to how you want to work.
[13:01] So, when you are going to create it, it can be customized for how you want to work.
[13:05] So, this would be an example where I come in and I can see today what are important things for me.
[13:09] I like to look at any kind of blockers across my different teams, decisions that are still open.
[13:14] It tells me how many projects it's tracking, and then
[13:16] It's going to start to give me ranked priorities, right?
[13:17] So, anything that's a blocker, hey, this is high urgency, tells me why, tells me who's dependent upon this, and what should be my action, and then I can tell the AI to take this action.
[13:27] So, you know, first of all comes in gives me a quick overview of like where am I?
[13:32] My personal operating system, what is happening in my life?
[13:36] Again, this can be as customized as you want.
[13:38] You can have all of your personal stuff in here that is capturing from Google that I need to go to the shops, all of these things.
[13:44] I'm specifically focused on a work version, but then it will come in, I'll look at the different projects I'm tracking, right?
[13:51] So, I may have a variety of different projects.
[13:52] If you're a marketer, you might be ingesting what the AI search team is doing, what the paid marketing team is doing, what the content team is doing.
[14:00] And so, it's integrating all of that information and figuring out how to make sense of it.
[14:07] So, in this example, I have these projects that are being synthesized and included in my AI second brain.
[14:13] The reality is for my brain second brain, it has everything, everything that is applicable to me.
[14:17] Then we have Karpathy's wiki.
[14:19] And so, what it's doing is over time learning different concepts and then creating wiki files for them so that it doesn't them.
[14:28] So, it may pick up over time that hey, like this concept of an internal AI SDR.
[14:34] So, you can actually have an AI agent help your sales team do prospecting, book meetings, all of these different things.
[14:40] They keep coming up and so, I'm going to start building wiki pages for them to make sure we store this intelligence and then store what we're learning about that concept.
[14:47] So, that's the LLM wiki for your agent, your pricing strategy, your partner program, whatever it may be, it starts to ingest it into this wiki and build singular files for it.
[15:00] I will say that in synthetic data I've used, it's used some of the most boring examples of all time.
[15:05] So, that's on me.
[15:07] I just told it to use synthetic go-to-market data.
[15:09] I should have actually told it to use interesting projects.
[15:13] So, you can see here, it will pull in a bunch of information, build a wiki.
[15:14] So, the thing to know here is
[15:17] You have to like set it up in terms of how to route stuff into the second brain.
[15:21] So, I will tell it I always want to keep track of experiments for a singular project.
[15:25] I always want to keep track of blockers.
[15:29] I always want to know what open decisions there are.
[15:32] And it's basically understands who are the different stakeholders that are related to this core project, who are the cross-team dependencies.
[15:39] It figures out what are all the documents that are applicable to this update.
[15:44] So, like where am I grabbing all of this information from?
[15:48] Build you a little knowledge graph.
[15:50] So, pretty cool, right?
[15:52] So, how is it capturing all of this?
[15:55] Well, we run ingest.
[15:57] And so, this again is a prototype, but this is how it really works.
[15:59] It will basically go through all of those sources that you've connected, Slack, email, docs, everything.
[16:03] And because it has routing logic, because in the files I tell it what to look for, and it then starts to understand where to update those different things.
[16:14] And so, it's keeping, remember we talked about the second brain decay, knowledge gets old over
[16:18] Time.
[16:20] This is how you keep it really current.
[16:23] I just run ingest.
[16:23] It captures everything for me across every singular file, across every singular comms channel that's applicable to me, and then starts to write it to the proper files.
[16:34] So, I think the main thing I would say here is the second brain can be customized to how you work, right?
[16:37] This is somewhat how I work.
[16:39] You may work very differently.
[16:40] And that's what is kind of really interesting about the second brain is it's a very personal operating model.
[16:47] It's basically going to be very much based upon how you work.
[16:51] So, why would you build this?
[16:52] And I think the main thing is that when you have an AI second brain, if you're using AI assistants and tools, you are connecting them to your second brain, and it's compounding over time.
[17:02] If you think about that ingest command, the more I learn, the better the system gets.
[17:06] The more I use the system, the more knowledge it's acquiring.
[17:10] Whereas most people are connecting to Claude or OpenAI and starting from scratch each and every time.
[17:15] That is the number one reason to use it.
[17:17] So, how do you get
[17:19] Started, right?
[17:20] I'm going to give you the quick starter pack and then my recommendation is just to get the transcript of this YouTube video and give it to Claude Code and build out a plan for you.
[17:31] And so, what do people have?
[17:33] There's some core building blocks.
[17:33] The first one is a vault, right?
[17:34] It's just a folder of plain text files.
[17:37] Obsidian is a tool that most folks are using.
[17:39] It's free, but it can be any folder on your desktop with Markdown files in it, and that's where your knowledge lives and that's what the AI system is writing to.
[17:48] When it's learning things, it has to be right to a knowledge vault, and that's your personal knowledge vault.
[17:54] An AI that can read the knowledge vault is the second building block.
[17:58] So, I use Claude Code.
[18:01] You can use Codex.
[18:01] You can use Claude desktop with MCP connections to your different files, right?
[18:05] You just need a tool, an AI assistant that can actually connect to that vault.
[18:11] Then, your third part is you really just need two basic folders.
[18:13] You need one where you can store things, and then you need one where you can write that intelligence from the stuff that you're
[18:20] Storing into your wiki folder.
[18:23] What you're going to need to create is the writing logic.
[18:25] What do you want the AI to look for in all those files?
[18:28] You may say, "Hey, look for blockers.
[18:32] Look for updated experiments.
[18:33] Look for decisions that were made.
[18:35] Look for places where people seem to be struggling to hit their goals.
[18:38] Look for places where people seem to have real clear articulation on the opportunities that they believe can help accelerate the business."
[18:47] You can basically put in some writing logic, and so when the AI assistant reads from the raw folder and takes all of that information, it understands what to look for and then update into your wiki.
[18:57] Then, you need to get into the habit of continually updating it cuz it gets better the more that you use it.
[19:05] And the connections, so what are you going to connect it to?
[19:06] So, you basically where are you going to read that information?
[19:10] Putting things in a raw folder is the most basic level.
[19:12] I just put a bunch of files in there and it reads from those files, but then it can connect to your Slack.
[19:16] It can connect to your Google Docs.
[19:18] It can connect to all different forms of comms.
[19:20] And docs and basically auto ingest those things and then wicks them into your vault.
[19:27] So again, you have a knowledge vault, you have an AI system that can read that vault, you have a place where you can store data that you want to enrich into the wiki and then eventually you have connectors that go and grab information from elsewhere and actually auto populates it into the wiki.
[19:43] And so that is your AI second brain.
[19:46] And if you want to go and start to build one, simply take this transcript, put it into Cloud Code and start to ask it to build one for you and go back and forth to build one in your style and you too can join the community, the AI second brainers.
[20:00] Hope you enjoyed this episode and see you on the next one.
[20:04] This data is wrong every freaking time.
[20:07] Have you heard of HubSpot?
[20:10] HubSpot is a CRM platform where everything is fully integrated.
[20:13] Wow, I can see the client's whole history, calls, support tickets, emails, and here's a task from 3 days ago I totally missed.
[20:21] HubSpot, [music] grow better.