Aviana Le · Creator + Founder · Building in publicInstagram @avianaleMogul Mastery waitlist

Mogul Maxxing  /  Blog  /  AI Second Brain

02 · AI Second Brain

How to use AI as a second brain without letting it do your thinking

An AI second brain should remember for you, not decide for you. Here is what the research says about where AI helps and where it quietly makes your work worse, and a simple setup for founders with limited work windows.

By Aviana LeOctober 202618 min read
Woman founder at a cream desk with a laptop and an open notebook, reviewing notes in warm morning light
A second brain is a place to put context down, so the next work window doesn't start from zero.

To use AI as a second brain, give it the job of holding and retrieving your context, your priorities, open loops, notes, decisions and rules, and keep the judgment for yourself. The AI drafts, sorts, summarizes and reminds. You decide what matters, check what it gives you, and make the final call. Used that way, it gives your limited work windows back. Used as a replacement for thinking, the research suggests it can make you faster and worse at the same time.

The short answer

A "second brain" is an outside system that holds what you would otherwise try to keep in your head. AI makes that system easier to search and talk to. The best evidence so far says AI raises speed and quality on tasks it is good at, and lowers accuracy on tasks just outside its reach, especially when people trust it without checking. So: let AI remember, retrieve and draft. Don't let it be the one who decides.

In the first article in this series, I wrote about decision fatigue and why a work window can vanish before the work starts. The last rule in that piece was to keep your decision system outside your head. This article is about that outside system, and specifically about the role AI can play in it, because "AI second brain" has become one of those phrases that means everything and nothing.

I'm building Mogul Maxxing in public, and an AI second brain is one of the core systems I'm designing for it. I'm not going to tell you it changed my life or saved me a number of hours a week, because I haven't measured that and I won't invent it. What I can do is lay out what the research actually shows, where the risks are, and the setup I'm testing, so you can try it yourself and judge it on your own results.

What is a "second brain", and where did the idea come from?

The idea of a second brain is older than AI. People have always kept commonplace books, notebooks, card indexes and filing systems so they didn't have to hold everything in memory. The phrase "building a second brain" was popularized by the productivity writer Tiago Forte, whose 2022 book of the same name describes a personal knowledge system built on two frameworks.

The first is CODE, a four-step process:

  • Capture: save the ideas, notes and information that resonate, without over-filtering.
  • Organize: sort them by where they'll be used, not by topic.
  • Distill: pull out the essential points so a note is useful at a glance.
  • Express: use the material to make something: a post, an offer, a proposal, a decision.

The second is PARA, a way to organize everything into four buckets: Projects (things with a goal and an end date), Areas (ongoing responsibilities like finances or health), Resources (topics you're interested in) and Archives (anything inactive).

The underlying idea is useful for founders because it shifts the purpose of notes from "storing knowledge" to "making things". A second brain isn't a library. It's a workshop. And that is exactly the frame that matters once AI enters the picture.

What AI adds to a second brain (and what it doesn't)

A traditional second brain has one big weakness: retrieval. You can capture hundreds of notes and still not find the one you need when a work window opens. AI changes that. Instead of searching folders, you can ask a question in plain language, "what did I decide about pricing last month?", and get an answer drawn from your own material. AI can also do the tedious middle steps of CODE: summarizing a long voice memo, turning scattered notes into a first draft, or sorting a messy capture list into projects.

What AI doesn't add is judgment about your business. It doesn't know which customer conversation mattered most, which idea fits who you are, or which trade-off you're willing to make, unless you've told it, and even then it's predicting, not deciding. That distinction sounds obvious. The research suggests it's easy to lose in practice.

An AI second brain is most useful as memory and a first draft. It's most dangerous as an authority.

What the research says about AI and knowledge work

There is now a small but solid body of experimental research on generative AI at work. It's worth knowing, because it points to a clear pattern: big gains on the right tasks, real losses on the wrong ones.

The gains are real on tasks AI is good at

In a 2023 experiment published in Science, MIT economists Shakked Noy and Whitney Zhang gave 453 college-educated professionals writing tasks typical of their jobs. Those given access to ChatGPT finished about 40% faster, and independent evaluators rated their work about 18% higher in quality.

A large field study by Erik Brynjolfsson, Danielle Li and Lindsey Raymond looked at 5,179 customer support agents who were given an AI assistant. Issues resolved per hour rose by 14% on average, and by 34% for novice and lower-skilled workers, with minimal impact on the most experienced and skilled agents.

The most-cited study is the Harvard Business School and Boston Consulting Group experiment led by Fabrizio Dell'Acqua, with 758 consultants. On tasks inside what the authors call AI's "capability frontier", consultants using AI completed 12.2% more tasks, finished 25.1% faster and produced work rated more than 40% higher in quality. Below-average performers improved the most.

The losses are real on tasks just outside its reach

The same consultant study included a task deliberately designed to sit outside the AI's frontier: a problem where the model would give a convincing but wrong answer. On that task, consultants using AI were 19 percentage points less likely to reach the correct solution than those working without it. The authors call this the "jagged technological frontier": AI is excellent at some tasks and poor at others that look similar, and you can't always tell which side of the line you're on.

That is the most important finding for anyone building an AI second brain. The risk isn't that AI is useless. It's that it's useful enough that you stop checking.

Trusting the tool more can mean thinking less

A 2025 study from Microsoft Research and Carnegie Mellon, presented at the CHI conference, surveyed 319 knowledge workers about 936 real examples of using generative AI at work. Workers who had more confidence in the AI reported doing less critical thinking. Workers who had more confidence in their own ability at the task reported doing more. The authors also found that critical thinking doesn't disappear with AI so much as shift, toward verifying information, integrating responses and stewarding the task.

There's also a widely shared 2025 preprint from the MIT Media Lab, "Your Brain on ChatGPT", which used EEG to compare 54 participants writing essays with an AI model, with a search engine, or with no tools. The AI group showed the weakest brain connectivity and had more trouble quoting their own essays. It's an interesting signal, but it's a small study, it was released as a preprint rather than peer-reviewed research, and it measured essay writing in a lab, not running a business. Treat it as a reason for caution, not a verdict.

Offloading to an outside system isn't new, and isn't bad

Psychologists call the habit of using the outside world to reduce mental load cognitive offloading. In a 2016 review in Trends in Cognitive Sciences, Evan Risko and Sam Gilbert describe it as using physical action to change a task's information-processing demands so it takes less mental effort, like writing a reminder or setting an alarm. Offloading is normal and often smart. The question is what you offload. Offloading memory frees attention for thinking. Offloading the thinking itself is a different trade.

And there's evidence that writing things down outside your head does more than save effort. In a 2011 series of experiments, E. J. Masicampo and Roy Baumeister found that unfinished goals tend to intrude on unrelated tasks, but that making a specific plan for the goal reduced those intrusions. You didn't need to finish the task. You needed a plan for it, written somewhere you trusted. That's a large part of what a second brain is for.

Overhead view of a notebook with handwritten project lists beside a closed laptop and a ceramic coffee cup on a linen tablecloth
Capture first, then let AI help you sort and retrieve. The system lives outside your head either way.

The rule I'm building around: AI remembers, you decide

Putting that research together, the principle I'm designing the Mogul Maxxing system around is simple: AI holds context and drafts; I hold judgment and approve. In practice, that splits the work like this.

Give the AIKeep for yourself
Remembering where you stopped and what's nextChoosing what the next window is for
Summarizing notes, calls, voice memos and researchDeciding which insight actually matters
First drafts of posts, emails, outlines and offersThe final version and the claims in it
Sorting captures into projects and areasWhat gets dropped, delayed or killed
Finding past decisions and the reasons behind themWhether those reasons still hold
Spotting gaps, questions and counter-argumentsThe trade-offs you're willing to make
Checklists and templates for repeat workAnything involving money, legal or health decisions

If a task sits in the right-hand column, AI can still help you think about it. It can lay out options, argue the other side or point out what you've missed. But the decision is written down by you, in your words, with your reason next to it.

How to set up an AI second brain in six steps

This is the setup I'm testing in the build. It's deliberately simple, because a second brain you have to maintain for an hour a week will eat the windows it's supposed to protect. You can do it with almost any combination of a notes app and an AI assistant that can read documents you give it.

1. Start with one home for everything

Pick one place where your notes live: a notes app, a folder of documents, or a workspace tool. The tool matters less than the rule that there is only one. A second brain split across five apps isn't a brain; it's a scavenger hunt. If your AI assistant can connect to that place or accept uploaded files, even better, but a copy-and-paste workflow is fine to begin with.

2. Create four core documents

Before adding hundreds of notes, create the four documents that matter most in a limited work window:

The four core documents

  1. Priorities: the two or three outcomes that matter this quarter, and why.
  2. Open loops: every unfinished commitment, idea or question, one line each.
  3. Decision log: date, decision, reason, and what would make you revisit it.
  4. Rules: your if-then rules for repeat decisions (when you check messages, what happens to new ideas mid-window, what you say no to).

These four documents are what you'll feed the AI most often. They give it the context to be useful instead of generic.

3. Capture fast, sort later

Capture should take seconds: a voice memo after a call, a quick note when an idea arrives, a saved link. Don't organize while you capture. Once or twice a week, paste the raw captures into your AI assistant and ask it to sort them into your projects and areas, flag anything that looks like an open loop, and list questions you'll need to answer. Then you review the sort and correct it. The AI does the tedious part; you make the calls.

4. Write a ready-to-resume note at the end of every window

This one comes directly from the decision fatigue article. Before you stop, write two or three lines: where you stopped, the next physical step, any open question. Then, when the next window opens, ask the AI one question with your notes attached: "Based on my priorities and my last resume note, what's the one task for this window, and what's the first step?" The answer is a suggestion. You confirm it, change it, or override it, and then you start.

5. Use AI to distill, not to replace reading

It's tempting to have AI summarize everything so you never read the original. For low-stakes material, that's fine. For anything that will shape a decision, such as customer feedback, a contract, research you'll cite, or numbers you'll act on, read the source, or at least spot-check it against the summary. This is exactly where the "outside the frontier" problem shows up: a fluent summary that's subtly wrong.

6. Ask it to argue with you

One of the most useful roles for AI in a second brain isn't answering, it's questioning. Before you log a significant decision, paste your reasoning and ask: "What am I missing? What's the strongest case against this?" The Microsoft and Carnegie Mellon research suggests that critical thinking with AI shifts toward verifying and stewarding. Asking for counter-arguments is a way of keeping your own thinking switched on.

Mogul Mastery · founding waitlist

Want a second brain built for limited work windows?

Mogul Mastery is the implementation program I'm building in public: a founder operating system and AI second brain that holds your context so each window starts on the work. It's in development and shaped by founding members.

Prompts that keep you in charge

The way you phrase a request decides whether AI is supporting your thinking or replacing it. Here are the kinds of prompts I'm using in the build. Notice that each one gives the AI your context and asks for something you'll then judge.

Second brain prompts, by job

  • Start a window: "Here are my priorities and my last resume note. Suggest the one task for this 60-minute window and the first step. Explain why in two sentences."
  • Sort captures: "Sort these notes into my projects and areas. List anything that sounds like a commitment as an open loop. Don't delete anything."
  • Find a past decision: "From my decision log, what did I decide about [topic], when, and what reason did I give?"
  • Stress-test a decision: "Here's what I'm planning and why. Give me the three strongest objections and what evidence would change my mind."
  • Draft, don't finish: "Turn these notes into a rough outline. Mark anything you're unsure about with [check]."
  • Close a window: "Here's what I did this session. Help me write a three-line resume note: where I stopped, next step, open question."

The "[check]" instruction in the drafting prompt is small but useful. It asks the AI to show where it's guessing, which makes it easier for you to catch the parts that need verification.

Woman entrepreneur writing in a leather notebook at a wooden table with a tablet beside her, in a warm minimal studio
The judgment stays on paper, in your words. The AI just makes the context easy to reach.

What an AI second brain should do for a founder

If you're deciding whether this is worth setting up, here is the test I'd use. A good AI second brain, for someone with limited work windows, should do four things:

  • Shorten the start. The gap between opening the laptop and doing real work gets smaller, because "where was I?" has an instant answer.
  • Close open loops. Unfinished commitments live somewhere trusted, so they stop interrupting other work. That's the Masicampo and Baumeister finding in practice.
  • Keep decisions consistent. You can see what you decided and why, so you stop re-deciding the same questions every few weeks.
  • Turn notes into output. Captures become drafts, outlines and offers, the "Express" step that most note systems never reach.

If your setup isn't doing at least one of those within a couple of weeks, it's probably too complicated, or it's being used to collect rather than to make.

Common mistakes with AI second brains

Building the system instead of using it

Second brains attract tinkering. New templates, new tags, a new app every month. Every hour spent rebuilding the system is an hour not spent on the business. Start with the four core documents and add structure only when you feel a specific pain.

Capturing everything and expressing nothing

AI makes capture almost free, which makes it easy to save hundreds of articles and transcripts you'll never use. Capture should serve projects you're actually working on. If a note doesn't connect to a project or area, it can go straight to the archive.

Trusting fluent answers

The consultant study's 19-point drop came from people accepting confident, wrong answers. Fluency isn't accuracy. For facts, figures, quotes and anything you'll publish or act on, check the source.

Letting the AI choose your priorities

Asking "what should I focus on?" with no context gets you generic advice. Asking it with full context can get you a reasonable suggestion, but it's still a suggestion. Priorities are the most important decisions in a small business, and they're the ones to keep firmly in your own hands.

Putting sensitive information in without thinking

Before you paste customer data, financial details or anything confidential into an AI tool, check the tool's privacy and data-use settings and your obligations to the people involved. Some information shouldn't go into a second brain at all, or only into one you control.

A simple weekly rhythm

Here's how the pieces fit into a week without becoming another job.

WhenWhat you doWhat the AI does
Every window, first 2 minutesConfirm the one task and startSuggests the task from your priorities and resume note
Every window, last 2 minutesWrite the resume noteHelps tighten it into three lines
During the dayCapture quickly, don't sortNothing yet
Once or twice a weekReview and correct the sort; update open loopsSorts captures, flags open loops and questions
Before big decisionsWrite your reasoning; log the decisionArgues the other side; finds related past decisions
Once a monthRe-read priorities; archive what's doneSummarizes the month's decisions and loose ends

Frequently asked questions

What does it mean to use AI as a second brain?

It means using AI as part of an outside system that holds your notes, priorities, open loops and decisions, so you can retrieve and work with them quickly. The AI helps with remembering, sorting, summarizing and drafting. You keep the judgment about what matters and what to do.

Who created the "second brain" idea?

People have used notebooks and filing systems as outside memory for centuries. The phrase "building a second brain" was popularized by Tiago Forte, whose 2022 book describes the CODE method (Capture, Organize, Distill, Express) and the PARA system (Projects, Areas, Resources, Archives).

Is using AI as a second brain bad for your thinking?

It depends on what you hand over. A 2025 Microsoft Research and Carnegie Mellon survey found that people with more confidence in AI reported less critical thinking. Offloading memory and first drafts frees attention; offloading decisions and verification is where the risk sits. Keeping a "you decide" rule helps.

What is the "jagged technological frontier"?

It's a term from a 2023 Harvard Business School and BCG study of 758 consultants. AI improved speed and quality on tasks within its capabilities, but on a task just outside them, consultants using AI were 19 percentage points less likely to get the right answer. The frontier is "jagged" because similar-looking tasks can fall on different sides.

What should I put in an AI second brain first?

Start with four documents: your current priorities, a list of open loops, a decision log, and your rules for repeat decisions. They give the AI the context it needs to be useful, and they're the documents you'll use most at the start of a work window.

Which app is best for an AI second brain?

There isn't one right app. What matters more is having a single home for your notes and an AI assistant that can read the documents you give it. Choose something you'll actually open every day, and avoid switching tools often.

Can AI help me decide what to work on first?

It can suggest a task if you give it your priorities and your last ready-to-resume note, and that can shorten the start of a work window. Treat the answer as a suggestion: confirm it, change it, or override it. The decision about priorities should stay yours.

Is cognitive offloading good or bad?

Usually good. Psychologists Evan Risko and Sam Gilbert describe cognitive offloading as using the outside world, like a reminder or a note, to reduce mental effort. Writing things down frees attention. The trade-off changes when you offload the thinking itself rather than the remembering.

How much time does an AI second brain take to maintain?

It should take very little: a couple of minutes at the start and end of each work window, and a short weekly review. If maintaining the system is taking more time than it saves, simplify it.

Where this goes next

An AI second brain is one piece of a bigger system. The next article in the series is about founder operating systems: how priorities, rules, rhythms and tools fit together so the business runs on decisions you've already made, not on whatever is loudest that day. After that come multifaceted focus, the move from creator to creator-founder, and validating an offer before you build it. The follow-up to this article, what an AI second brain should do for a founder, turns this setup into a job description: seven jobs to hand it and the ones to keep for yourself.

If you want the build notes as they happen, including which parts of this setup hold up, join the Mogul Movement. And if you want to see the program taking shape around it, the Mogul Mastery founding waitlist is open. You can also start from the beginning with decision fatigue for entrepreneurs.

Sources

This article is general information about productivity research and AI tools, not professional, legal or business advice. AI tools can be wrong; verify important information before acting on it, and check a tool's privacy settings before sharing sensitive data.