Data Wins. Not AI. (A Raw Lab Session)

00:00 — Intro: Why Data Is the New Battlefield

Hey, welcome to Slow Builds.

So this is gonna be like an unedited, just speak out loud, follow the script about who I believe, well I don’t know who’s gonna win the AI battle in the end, but based on what different people have and have access to and are obtaining and control, I think that’s what helps people win the AI race basically, or in a combination, different avenues of it.

So this really started as something totally different.

I was thinking about building a personal AI memory system.

And it’s something I started a long time ago. My friends told me I was kind of thinking too big, I guess, or it was not plausible because it was too much information.

It really started as just a memory dump. I think I called it Brain Dump, and it might actually be another video where I talk about how I use AI just to give it everything.

The idea around that was I want to give it all my notes, emails, ideas, chats, calendar.

The example I always give is like I run a lot, so I know what shoes, exactly which brand, which size, which model.

And then if I keep track of that and keep notes on it — treadmill, outdoors, wears down, all these different things — even ones that like, okay, this one’s no good, that one didn’t work.

Then when I go see a new one is on sale or I want to try it, I can quickly say:
“Hey, I’m looking at this shoe.”

And it can come back and give me specifically:
“Oh, based on your previous things, I suggest this, this, and this.”

That was really the idea, like a brain dump.

So it was a way to try to figure out how to connect it all together, keep track of it.

And I really didn’t know what I was getting at, but it was more or less a personal assistant, I suppose.

And the deeper I got into that and realized all the connectors and everything you had to do, I was like:
“Hm… based on everything I know about AI and models and stuff, it’s all about the data.”

It’s not about the model in general.

The generative intelligence obviously is what it’s built off of and how it makes decisions and recognizes patterns and uses memory and analyzes things to find the best possible answer.

It’s all probability.

But the way you do that is with data.

00:54 — How the “Brain Dump” Personal Memory System Shifted My Thinking

I’ve been reading a book — I think I finished it now — by a YouTuber I watch. I’m gonna say his name wrong, Farzad I think. The book is called Abundance or Collapse.

And in this one he talks about data moats.

Because the data is the moat that protects the LLM.

Whoever has the biggest moats usually wins.

Who has the most data.
Who has the most connections.

And it’s understanding the value of that data because of how their AI learns from it.

It has this context and this feedback.

It’s such a big knowledge system.

And I started looking at all the companies that hold this amount of data and human behavior and how that’s really gonna fuel the AI systems.

It’s almost like their food source in a way.

Search history.
Emails.
Videos.
Conversations.
Workflows.
Purchase history.
Location history.
Medical data.
Social interaction.
Financial data.

It’s not just random information.

It’s patterns of human life.

And the more you dig into it and think about it, you stop seeing AI as just:
“Who has the smartest model?”

You start seeing it as:
“Who owns the deepest understanding of humans and how we operate?”

What we need.
What we want.

02:37 — The Generative Fallacy: Probability vs. True Human Patterns

Most people look at AI really like:
Who has the smartest chatbot?
Who can analyze the deepest?
Write the best code?
Who’s fastest?
Who’s gonna get there first?

But it doesn’t really matter.

It feels more like each model is kind of replaceable in a way.

It’s not bad.
It’s not unimportant.

But it’s less of a moat because eventually everyone catches up to everyone else.

Building the model itself, eventually everyone figures it out.

They do things a little different.
Different speed.
Different processing.
Different algorithms.

But what you can’t change is the underlying data that it builds from.

You can synthesize it, but then you’re gonna get hallucinations and side effects and incomplete hypotheses and stuff like that.

You’re not getting true human input and data and reactions and observations and how we act and how we see things and what we expect.

And that changes the probability of what we’re looking for.

So then what actually matters?

The context.

The memory.

The conversations and observations.

The ecosystem around the model.

The human side of the equation.

04:12 — Data Moats: Why the Biggest Knowledge System Wins

When you look at the companies out there, think about Google.

Search.
Chrome.
YouTube.
Maps.
Gmail.
Google Drive.
Android.

All the phones and devices running Android — all that’s feeding back into the system.

That’s not just internet.

That’s behavior.

People searching.
Where they’re going.
What they’re watching.
What documents they’re writing.
Where they’re clicking.
What they’re looking for.

All that becomes a very big moat.

That’s why when people thought Google was behind in the race and then suddenly they release something new and leapfrog again, it shouldn’t really surprise people.

They’re leveraging the data they already have.

They don’t just need a smarter model.

They already have one of the biggest pools of human behavior in the world.

05:41 — Company Deep Dive: Google’s Behavioral Pool

Then you look at Microsoft.

And I think people really underestimate Microsoft.

They own Office.
Outlook.
Teams.
Azure.
Xbox.
LinkedIn.
Windows.
GitHub.

And GitHub is a big one.

That’s where all the code lives.

There’s not many companies that aren’t running on GitHub in some way.

Even if private code is private, there’s still ways to analyze patterns at a high level.

What languages are being used.
What frameworks are trending.
Developer habits.
What companies are building.

And when you’re that big, you can do a lot with patterns.

Then combine that with LinkedIn.

Who’s hiring.
Who’s applying.
What jobs are being posted.
What companies are growing.

It’s a mammoth amount of information.

Then add Teams and Office and Outlook and Windows on top of it.

Microsoft holds a lot of keys to the business world.

That’s a very big moat.

06:58 — Company Deep Dive: Microsoft’s Keys to the Business World

Then you turn into Meta.

Facebook.
Instagram.
WhatsApp.
Oculus.

Between WhatsApp and Facebook and Instagram, you’re talking billions and billions of active users constantly.

All that conversation.
All that interaction.
All that human behavior.

Relationships.
What people are posting.
What they’re interested in.

That’s insane.

Google has one type of data.

Microsoft has business data.

Meta has social behavior.

That’s a huge one.

08:16 — Company Deep Dive: Meta’s Social Behavioral Advantage

Then you throw in Amazon.

AWS powers a huge amount of the internet.

When AWS goes down, huge parts of the internet feel it.

There’s logistics.
Software.
Storage.
Infrastructure.

Then you add Alexa.

I use Alexa all the time.

Groceries.
Cars.
Family communication.
Smart home stuff.

Then you buy everything through Amazon.

Purchase history.
Watching habits.
Prime Video.
Twitch.
Smart devices.

Amazon understands buying behavior.

What people actually spend money on.

What startups are happening.
What people are storing.
What companies are building on AWS.

Amazon has a huge moat too.

Then there are the AI-first companies themselves.

OpenAI.
Anthropic.

What’s interesting about them is they don’t necessarily own the massive ecosystem layers of data like Google and the others.

But they consume the data users give them directly.

And that data is still extremely valuable.

Because people aren’t just searching with AI anymore.

They’re thinking with it.

Planning.
Brainstorming.
Debugging.
Learning.
Emotional questions.
Financial questions.
Life decisions.

Those conversations become very rich very quickly.

OpenAI especially feels like it’s replacing parts of search and research and brainstorming and decision making.

Then you have Anthropic with Claude.

We use Claude at work.

And honestly, I think most developers do.

But not just developers.

Researchers.
Writers.
Analysts.
Lawyers.
Business people.

People are feeding it contracts, taxes, financial analysis, legal documents, code.

That’s a huge amount of high-value interaction data.

09:30 — The “Connector Realization”: Partnerships Are Everything

But the really interesting one in this category is xAI.

Because unlike the others, it potentially has access to the entire Musk ecosystem.

X for real-time public conversations and sentiment.

Tesla for real-world driving and camera data.

Starlink for global internet infrastructure.

Tesla Solar and Powerwall.

Robotics.
Chips.
SpaceX.

It’s not just internet behavior.

It’s real-world data.

Driving.
Energy.
Communication.
Robotics.
Movement.

And all of it is becoming tightly woven together into one ecosystem.

That’s very different from just being a chatbot company.

But the strange thing is we really don’t know what’s happening behind the scenes.

These are corporations after all.

We see demos.
Announcements.
Benchmarks.

But we don’t know:
what breakthroughs they’re sitting on,
what projects failed,
what systems are almost ready,
what partnerships they’re building internally,
or what data-sharing agreements exist.

One breakthrough could suddenly reorder everything.

Better memory.
Cheaper inference.
Robotics.
Agent systems.
Synthetic data.
New chips.

Any one of those could suddenly change who’s leading.

11:05 — The Scary Side: Who Owns the Patterns of Your Identity?

But the thing that really hit me planning my own little Brain Dump system was the importance of connectors.

The connectors are everything.

Because right now all our lives are fragmented.

Work.
Personal.
Bank.
Phone.
Doctor.
Accountant.
Insurance.
Email.

Everything is spread everywhere.

So a huge part of this is gonna come down to partnerships and integrations and APIs and what systems are allowed to talk to each other.

What information gets shared.

What patterns can be understood.

And honestly, this is where it starts feeling both exciting and uncomfortable.

Because I want this.

That’s why I was trying to build my own.

I want to just open my phone and ask something and it already knows.

It knows I forgot to call about the car.

It knows when my dentist appointment is.

It knows what I’m talking about without me needing to explain it.

That sounds incredibly useful.

But at the same time:
who owns that data?

Who owns the behavior?
The patterns?
The identity?

And that’s where it starts getting scary.

12:20 — We Aren’t Building Chatbots; We Are Building Digital Operating Systems

The more I think about AI now, the less I think this becomes:
“Who builds the smartest model?”

It’s more:
“Who builds the best understanding of people and systems?”

And honestly, I don’t think there’s going to be one winner.

I think different companies are building different pieces of the same future.

Search.
Work.
Social.
Commerce.
Finance.
Robotics.
Infrastructure.
Transportation.

The real race may end up being who connects it all together.

And the weird thing is all this started snowballing in my brain because I just wanted a better note-taking system.

I just wanted somewhere to dump my thoughts and ask questions later and have it already understand the context.

Anyway, there’s a lot to unpack here.

I’m trying to move these videos more into raw, unedited conversations.

I don’t really edit much.

I want it to feel more real and more like thinking out loud.

Anyway, thanks for watching.