Using software to create a "second brain" is a concept I only stumbled across a few months ago, although I have, in essence, been doing the same thing for years. Our brains are skilled at coming up with new ideas, but we have a finite amount of storage to work with. Sure, you might still remember your high school locker code, but it isn't always easy to retain the information that you need to know. That's where a "second brain" comes in — a way to keep notes that allow us to more easily access learned knowledge and create links between ideas.

Obsidian has become one of the most popular tools for building a second brain. Over the past few months, I've worked on transitioning my notes across various devices — including my phone, laptop, and a dozen notebooks scattered throughout my office — into a single Obsidian vault. However, syncing it with a local LM catapulted the project forward. Integrating Copilot into my vault granted me the ability to quickly summarize notes (and meant I could get the gist of long articles I'd saved but not gotten around to reading yet) while simultaneously granting the privacy control I sought for a system that contains personal information.

Copilot finds connections that I miss

It analyzes data faster than I can

I like to think of myself as a fast reader, but Copilot is significantly quicker than I am, even on my best day. It can scan the entirety of my vault and make connections between data that I might not see. For example, I use Obsidian to track my reading list (at least in part), although some data has been lost over the years as I've transitioned between Goodreads and other services. I can use the @Vault command in Obsidian to have it identify patterns in my reading habits. It pointed out that I have a preference for fantasy and a tendency to read full series versus standalone titles.

I then asked CoPilot to identify tags that could help me further organize the folder. It returned multiple different categories, broken down into genre, date, and more. Instead of just a "fantasy" tag, I could also include "epic fantasy" or "romantasy" depending on the title. The LLM's ability to analyze large amounts of data in a short time makes it a phenomenal tool, especially when you're trying to find threads between a huge amount of connected data.

I don't need to review the entire note anymore — I can just ask Copilot for a summary

I can spend my time on the research that matters

When I'm crafting an idea and conducting research, especially in the earliest stages, I typically skim the content and save it for later review. That's how I've always preferred to work: start out light and fast, like a scout surveying the landscape, and return later for a more in-depth analysis of what I've found. Of course, this can result in quite a few notes and saved URLs. Thanks to Obsidian, I can simply clip anything I find as a note and then ask Copilot to summarize it for me. Entire Wikipedia pages can be reduced to bullet points, with a focus on a given theme or subject. That isn't to say I don't do my due diligence while researching, but I spend less time reading things that ultimately turn out to have no relevance to the current project.

This might be my favorite Copilot feature of them all, actually. I can scan the bulleted summary of an article and get the big picture, then read deeper to find specific information. Here's a quick example. With a baby on the way, a huge amount of my time has been spent trying to learn everything I can about parenting. I stumbled across an alternative potty training method called elimination communication and saved the Wikipedia page to my vault, then asked Copilot to summarize it for me. It quickly listed out the proposed benefits of the method, potential downsides and concerns, and much more. Will I rely solely on that information? Absolutely not, but it does give me a starting point for further investigation.

Custom prompts are deceptively powerful

Your local LM lacking a feature? Tell it what to do

By default, Obsidian Copilot offers a host of different features and tools. However, you can create your own custom prompts that will tell it to do essentially anything you can. For example, I had a lot of notes that had been copied and pasted from other platforms, and the formatting was off in many of them. To remedy that problem without manually doing it myself, I created a "Clean Up" prompt that told my LM to add dividers between different sections, group relevant content together with a relatively generic header, and, most importantly, to ensure nothing was deleted in the process.

I have since used that prompt to organize pages of disjointed thoughts into one, and it has automatically grouped ideas together based on subject matter. I used it in a folder that contains all my different pitch ideas, and it has grouped those together based on what they're related to, such as travel, software, hardware, gaming, etc. It's not perfect by any means; some ideas get categorized in the wrong sections. However, replacing one or two misfires is a lot easier than trying to sort out all of the content myself.

The downside? The setup can be a bit technical for first-timers

Don't be afraid to seek out tutorials

To use a local LM with Obsidian, you'll need a third-party application to run it. In my case, I used LM Studio, but other options, such as Ollama, will also work. Here's the kicker: you'll need a relatively modern system to run LM Studio locally. My MacBook Pro was out of date, so I had to turn to my Windows machine instead. Running it locally is slower than using a cloud-based alternative, but it's far more private. That said, it's not the easiest thing to set up; it takes a bit of know-how, but LM Studio is approachable even for first-time users, and the most basic models are free to use.

Despite the setup, its functions are worth the work

Other AI tools can't draw connections like Copilot

While I have my concerns over Copilot's current capabilities, its ability to examine my entire vault and find connections can't be overstated. That's something that few other tools can do, making it a plus for power users. If you have a vault with thousands of notes, there's no feasible way to keep track of what you've included and what you haven't. Copilot can identify trends in your thoughts on a subject, such as whether you have leaned more in favor of or against a given idea over time. That's its main advantage over other AI tools: it doesn't treat each request as an isolated event, but sees it in the context of your overall vault.

Privacy + functionality is what I call a win

The initial technical setup can be an obstacle, but a combination of YouTube and Google can help clear that hurdle. After that, you'll find you get a lot of use from its various features. I can ask it to suggest specific tags for notes, automatically add them, and much more. Just the ability to organize and format walls of text is enough to make it a worthwhile tool. My only complaint is that it doesn't work as quickly as I would like, but that's likely due to my own dated hardware more than an operational issue.