When I graduated from my MBA, I went to work at the intersection of people and technology doing interactive marketing and social media.

Technology Gaps

My own fascination with technology has led me to read the work of those at the forefront of technological change.

In my professional career, I’ve interacted with a lot of people with varying degrees of technology proficiency. A lot of my work has placed me in the position of being the most capable and skilled in any given room as it relates to the use of technology. I am often turned to as a resource by those who hope I can bring them current or give them an edge against their own contemporaries.

As we find ourselves thoroughly immersed in the era of artificial intelligence, I’m seeing a similar gap emerge as I have in every other cycle. Mass adoption of and fluency with a technology often lags far behind those at the forefront. While we see companies engaged in mass layoffs to allocate resources toward artificial intelligence, the truth is that the number of people on the inside of those organizations who deeply understand the technology are few in number. The number of those in the general public, among the labor class, or who work for themselves, that have the technological fluency to deeply understand these technologies are likewise few, and with even less access to capital.

The Missing Ownable Infrastructure

While all of the dominant large language model (LLM) tools are doing their best to keep things simple for the average user, the problem that surfaces is always in accuracy and personalization.

Sure, many of these tools will inevitably go the way of social networks and begin collecting information without our consent and then “personalizing” the experience. Either that or manipulating the direction that we go.

But from the standpoint of enabling us to do advanced work on our own terms, they’re sorely lacking because of a missing ownable infrastructure with which to operate on top of.

The Developer’s Toolkit Problem

When you look at advanced AI usage or capabilities, it typically requires a certain level of expertise. So even though tools like Claude Code are remarkably powerful, the truth is obvious: most people are not comfortable in the command line or even an IDE like VS Code. They will struggle to understand folder structures and glaze over as we explain the benefits of markdown.

The technical overhead is a barrier the majority of people will never overcome.

This is the foundational setting for why I believe Notion’s AI product is so powerful –– and why I believe it should be so attractive to most people as a method for going from beginner to intermediate or advanced in how they use AI tools.

Why Notion is an Attractive Solution for Advanced AI

I believe the technical hurdles to advanced AI are the very reason why Notion is such a strong option for most people. What it’s doing is providing access to large language models and then layering it on top of an interconnected information management system.

While you can plug Claude and ChatGPT and Gemini into other tools, Notion provides a singular ecosystem that is adaptable to a multitude of different functions simply by setting up a new type of template, and then allowing you to work within a single workspace for various types of tasks. With just a few additional types of modules, Notion could provide a platform for just about any type of work.

A Familiar Interface, A Powerful Backend

Notion sidesteps a lot of the technical problems because you can have a pre-setup workspace where all of the rules are already in place.

The way most users would interact with the tool is virtually the same as it would be in using Claude, ChatGPT, or Gemini: a chat interface. But behind the scenes, you have the ability to easily connect to any resource in your Notion workspace. You can point your AI to any resource and to any instructions. Rather than learning how to operate in a developer environment, you get to work within what feels like a chat interface combined with a note-taking tool.

Hiding in Plain Sight

I think this is grossly overlooked because Notion AI is not in and of itself its own AI, but rather utilizes the large language models that are out there, but in a structured environment. For the same cost as a ChatGPT or Claude subscription, you can get unlimited access to both, and Gemini, and more.

For those who wish to do advanced AI work, rather than learning how to manage all of the complexities of MCPs and IDEs and command lines and permission structures, working inside of the Notion environment is by comparison profoundly simple and yet equally as powerful.

The only hurdles that remain are at the truly advanced level and in the world of coding. But for most users –– those writing copy, keeping track of contacts, looking at their financials –– all of that can be managed within a single tool. And that tool is Notion AI.

Why I’m All In

One of the reasons I’m betting so heavily on Notion AI is because it’s the system that I use for myself. It would be hard to enumerate all of the ways in which Notion has helped me to be more productive, keep track of my priorities, and be able to support and help more people.

I believe the best technology should be powerful, but easy to use and widely accessible for people to take advantage of.

The people I seek to help with technology are less so Fortune 500 companies and more so those that want to just live a better life by keeping better track of what they’re working on, and automating away repetitive work.

I want people to have more time to spend with those they love and do more impactful work with less stress.

Sharing What Works

I’ve spent a lot of time developing my own system inside of Notion, and the benefit of Notion is that I can make that system widely available to people.

Any page you create, you can share as a template for people to download and duplicate into their own workspace. So as I develop my own organizational system and knowledge base, I can share it. As I develop my own rules for how my AI works, I can share it.

I’m continually updating and improving this system and plan to make available everything that winds up helping me, so that it can in turn help others.

I’m working to take the guesswork out of it and saving people from having to learn or build an entirely new system on their own.

The post I’m Betting on the Most Overlooked Power Platform in AI  appeared first on JGibbard.

There are a lot of people who would insist, via their reaction, that you should do one thing.

• You should be a lawyer, a doctor, or an accountant.
• You certainly should not be a multi-hyphenate, multi-passionate individual.

The market often demands that the mosaic of you be sharpened into a single pixel, so that you can be neatly organized and contained within one box.

But what about those of us for whom that is a near impossibility?

What about those who immediately recoil from the craft they once loved when it ceases to be an outlet and instead becomes a prison?

I am that person.

I have many interests and I cannot convince myself of the logic that I should not. Nor can I convince myself that it is worthwhile to limit my contributions to others by making myself smaller or hiding some of my gifts behind a single title or offer.

What I have chosen instead, is to simply work harder on organization and communication.

It’s not actually about choosing one thing. It’s about being deliberate in how you communicate.

• The problem isn’t how many interests I have but whether the connections between them are coherent.
• The problem isn’t that I have many things to offer, it’s whether I present it in a targeted and relevant drip or a powerful and overwhelming firehose blast.

For me, I need to understand the logic of how all of my endeavors connect. I only add something if it is congruent with the rest of the ecosystem and adds value.

For those I seek to work with, all that matters is showing them what is relevant to the conversation.

“You’re too much. You’re not enough.”

When we try to be all that we believe we’re capable of, we will inevitably hear a chorus of people telling us that we cannot. That we are too much and must tone it down. That we should not be all of those things. We cannot have all of those interests. We must be what they believe we should be.

Alternatively, they may see our ambition an scoff at our audacity. They will assure us, there is no way we can be good at all of those things. There is no way we can be enough to step into the pictures we paint.

For all of those who have heard that they are too much, or that they are not enough, please do not shape what you do and how you see yourself based on the opinions of others.

Consider that maybe you are the sole authority on you. Consider that you are the one who will live with the choices you make. And consider how much more vibrant life would be if you did that instead of living with the choices and constraints that were forced onto you by someone else.

The post On Doing Just One Thing appeared first on JGibbard.

He describes a knowledge base where AI doesn’t just retrieve information but actively maintains, cross-references, and compounds it.

I recognized two big problems:

1. A fully AI-maintained system can only organize in a way that is recognizable to the user after substantial setup by the user.
2. In the absence of information architecture the wiki will quickly grow past the point of efficiency it set out to achieve in the first place.

So, here is my contribution to the idea.

What Karpathy Gets Right

RAG is a significant improvement over giving your LLM every context document it might need at some point. I think of it as the AI-equivalent of “just-in-time manufacturing.” Less waste, faster output.

Karpathy correctly points out that most RAG systems lack an answer for how to replace old information, with new information, at scale.

Additionally, Karpathy is right to point out that the key bottleneck is the human user handling the organization tasks. A source not properly integrated as context is effectively invisible.

Karpathy’s wiki approach lets AI write back, integrating new information into existing understanding, filing strong answers for reuse, letting knowledge build on knowledge instead of starting from scratch with every new conversation.

The Missing Piece from Karpathy’s idea is Routing Architecture

As I understand it…

Karpathy’s index.md is the map of the namespace. It’s AI-maintained, which means it’s organized according to how the AI would organize it, not necessarily how the user would organize it. Most of the time this wouldn’t be a problem, but it’s also more than just edge cases where it does.

So, the first problem is the words.

The second problem is the lack of structure, which means this idea falls apart at scale. RAG is intended to be more efficient, but once your index gets too big, the LLM is still searching through far too many lines to find what it’s looking for.

I believe the missing pieces are routing tables and non-hierarchical taxonomies.

Let me explain with a simple example.

If you were walking into someone’s house and looking for a spatula, you would presumably first look for the kitchen. Next, you would scan the kitchen for the most likely places to find one, likely near the stove. Do this, and you would likely find it pretty quickly.

In this case you might have: 1st Floor > Kitchen > Drawers near stove > Spatulas

By contrast, imagine that instead you either need to walk in and begin searching in every room of the house, or look through a giant index listing every single item in the house.

Which is more efficient?

Additionally, back to the first problem, what if you have always called a spatula “the flat flippy thing” or “the pancake flipper?”

A good organizational system needs to assign multiple ways to move closer to the information you are looking for and needs to have enough slack in the system to support your unique use of language.

My Contribution

I’ve been using Notion as my knowledge management system for almost a decade now. And even before Notion AI, I knew that I needed a system where “everything has a home” so that I could easily find whatever I needed. This developed into what I call The SuperBrain OS.

This isn’t a pitch for my template (though you should be it because it’s awesome), I’m just using it to explain how it aligns with what Karpathy was eluding to while introducing an element that I feel is missing.

The SuperBrain OS doesn’t use one big wiki. It uses multiple databases, each with a defined purpose, connected to other databases through relational fields. Each database serves can serve as a taxonomy for the databases it is connected to.

When the Tags databases is connected to Notes, Tasks, and Articles, I’m then able to find any of those via the Tag. Add in other relations and now you can combine fields to find what you are looking for — just like my spatula example. But by having a non-hierarchical taxonomy, I can now quickly find bathroom towels, kitchen towels, and beach towels.

These relations, along with routing tables are the first layer of the index. This mean that my AI system can follow 2 or 3 layers before searching for the information, and is both more efficient in finding it, and more likely to find exactly what I was asking for. And this same system works in reverse for the AI to maintain my knowledge system.

Chunks > Piles

When developing my system, I chose a number of different categories that I felt best represented the big buckets that information falls into. Then, within those databases, I added some light refining categorization, as well as relations to some other big categories.

This creates a series of progressive chunks that shortens the distance from query to retrieval. Nothing ever gets so big that it can’t be grouped.

• Instead of an index with 100 lines, it’s 10 groups with 10 items each.
• Instead of an index with 1,000 lines, it might be 20 groups, with 50 items each broken into 5 groups of 10.

As the dataset grows, having a way finding system becomes critical. My approach has been to build that from the beginning and plug the AI into it.

You design the structure, the AI maintains it

Our knowledge is personal. It’s informed by our values, our decisions, our frameworks. Therefore, the organizing logic needs to be ours. My AI needs to understand how I connect ideas. That requires me to define the structure that enforces the logic I’ve chosen.

Karpathy’s whole idea works beautifully if you start with the information architecture.

Acknowledged Gaps

Because Karpathy is speaking about systems, running in an IDE or command line and organizing markdown files with a programs like Obsidian or simple folders, it’s a little different than Notion AI.

While I’ve been able to build some very clever instructions for my Notion AI to save transcripts, including decisions, ideas, and changes, I’m still relying on it to follow the instructions or have me manually invoke a command. In Claude Code, it can be built to happen every time. Similarly, the lint system can only happen in Notion if triggered manually or using a Notion Custom Agents (which would be prohibitively expensive).

This means that while the system I’m using has a more sophisticated organizational structure, it has a less powerful system to maintain it.

But somewhere in the space between what Karpathy, a research-focused AI engineer, and I came up with is a scalable managed knowledge.

The keys are going to be out ability to build it in a way that balances accuracy and efficiency with cost. Otherwise, we’re just wasting a lot of water so we don’t need to add tags to our notes.

The post The Missing Pieces from Andrej Karpathy LLM Wiki System appeared first on JGibbard.

I can’t even count how many website revisions I’ve gone through since I began working for myself in 2011. I’ve had personal homepages, company websites, landing pages for side hustles, and all sorts of content projects. Some have some and gone, others continue to change.

One thing that remains is the direction: FORWARD

Working for yourself is always an exercise in reinvention, in reframing, and in reflecting.

We’re all making this up as we go along. The hope is we keep getting closer to where we want to be.

The post One More Entry appeared first on JGibbard.