← Back to blog
· philosophy, open-source, AI

Own Your Stack: Why I Rejected Third-Party AI Services

When I came across an AI memory system that already solved a large part of what I needed, the decision initially felt obvious. Use it, save time, ship faster.

Then I looked closer at the license.

The Sovereignty Argument

It wasn’t MIT-licensed. That single detail changed everything.

If you build your cognitive layer: memory, reasoning and automation on top of someone else’s proprietary foundation, you don’t actually own your system. You’re renting it. The provider controls the terms, the pricing, the roadmap, and ultimately the relationship between you and your AI.

We’ve seen this pattern before. Platforms become essential, then gradually shift priorities: pricing increases, core features degrade, and lock-in becomes real. The convenience upfront hides the long-term dependency.

A person standing at a fork between a locked box and a layered stack.
Rent the foundation, or own the stack.

Building vs Buying in the Age of Agents

What changed my perspective is how much easier it has become to build these systems yourself. With agents, you no longer need to design everything perfectly upfront or spend months implementing infrastructure. You can start from zero and iteratively build:

  • A memory layer
  • A user model
  • Automation and tool orchestration

The cost of building from scratch has dropped dramatically. Agents can generate, refactor, and extend systems as you go. What used to be a major engineering investment is now a continuous, incremental process.

This flips the equation.

It’s no longer:

“Is it worth building this ourselves?”

It becomes:

“Why would we give up control when building it is this accessible?”

The Practical Outcome

Instead of relying on a third-party system, we built our own core components:

  • Recall (memory)
  • User Model (preferences and behavior)
  • Skill Curator (automation layer)

These pieces cover the same ground as many external tools, but they’re ours. Fully controllable, modifiable, and aligned with our architecture. if there is a SPEC for it then you can build it!

Owning the Vehicle

This isn’t about reinventing wheels.

It’s about owning the vehicle.

Because when your system becomes an extension of your thinking, control isn’t a nice to have. it’s essentially the whole point.