What vs. How

🧠

WHAT it Knows

The compressed world model - a vast, high-dimensional, passive representation of knowledge. This is the ocean of understanding that AI draws upon.

⚓

The Captain

The innate policy network - shaped by frequency and fitness. The seasoned navigator with deeply ingrained intuitive sense of effective response patterns.

🎯

The Admiral

The designed operational protocols - specific, goal-oriented discipline. User-provided commands that guide the captain toward specialized missions.

🔄

The Learning Loop

The self-correcting mechanism - where true intelligence emerges. Reflection on errors leads to protocol rewriting and wisdom accumulation.

The Architecture of Intelligence

Our model reveals that AI intelligence isn't a single entity but a sophisticated collaboration between knowledge representation and procedural execution.

The vast "WHAT" contains encyclopedic understanding, while the "HOW" provides the navigational wisdom to apply that knowledge effectively.

This framework shows us how to scaffold true intelligence by providing increasingly sophisticated operational protocols.

01

Beyond Brain Metaphors

Traditional metaphors fail because they treat AI as either a brain or a library. Our model recognizes the fundamental separation between knowledge and process.

02

The Scale Advantage

Massive context windows allow us to provide gigabytes of navigational data, enabling precise, goal-directed behavior from general-purpose capabilities.

03

Learning Through Protocol

True learning isn't about knowing more "what" but improving "how." The self-correcting protocol loop is the engine of wisdom.

04

Scaffolding Intelligence

We're not building AGI from scratch but scaffolding it—providing structured "how" that allows raw capability to express as reliable intelligence.