AI

Where does AI play into the whole strategy?

For a long time I've known that the biggest bottleneck wasn't the software—it was my ability to explain complex ideas quickly enough.

The original concept site was an attempt to solve that problem, but I was constrained by the tools available. Systems like Confluence were slow, difficult to navigate, and encouraged large collections of disconnected documents rather than a coherent body of knowledge.

Every day we're improving the way this site is generated. The goal is to make knowledge progressively easier to create, organize, and understand.

A key part of that vision is using AI to extract important concepts—such as OpenSSL, TLS, or HTTP—and automatically generate explanations and documentation. AI gives us an excellent starting point, but it still requires human review. We routinely refine, correct, and simplify what it produces.

One of AI's biggest weaknesses is information pollution. It learns from the internet, where complexity is often mistaken for quality. Technologies such as HTTP/2, for example, accumulated enormous complexity while being heavily promoted by large organizations. More features do not necessarily produce better engineering. In many cases, they simply create larger attack surfaces and systems that are harder to understand.

Our goal is different.

We want to distill knowledge down to its essential ideas. The philosophy is much closer to the reduced instruction set approach that made ARM processors so successful, or the simplicity that made JSON widely adopted. Remove unnecessary complexity. Keep the concepts that matter. Make every layer easier to understand.

Over time, this growing body of curated knowledge becomes training material—not just for people, but for AI itself.

Our long-term vision is to build smaller, sovereign AI models trained on carefully distilled engineering knowledge rather than the noise of the public internet. We believe those models will ultimately be faster, easier to run, require fewer resources, and produce higher-quality code because they will reason from clear first principles instead of trying to imitate everything ever written.

In many ways, this project is about capturing the accumulated experience of decades of software engineering: reducing problems to their fundamentals, questioning unnecessary complexity, and communicating ideas clearly enough that both humans and AI can build upon them.

That won't happen overnight. It's a long-term project.

But every improvement we make to how we communicate, structure knowledge, and simplify software moves us one step closer to that goal.