Technology
Kernel Ontology — we set the broad frame, and the data itself organizes and generates the relationships, building up the network.
What we mean by structure
The structure kernalic builds is an explicit definition of how data relates. Which data supersedes which, which concept contains which — only once those relationships are defined can AI reason over the network instead of over scattered pieces of data, and feed what it finds back in to make it more precise.
A kernel a person sets, and the kernels the machine finds
We do not draw the whole structure by hand. What a person fixes is the broad frame: the kinds of things it handles, the names of the relationships that join them, and the citation rule that points at the evidence. We call that frame the kernel. Just as a kernel in deep learning sweeps one small window across an entire image to pull structure out of it, one broad frame applies identically across an entire body of data. The kernel is designed anew for each organization: even within one domain, every organization applies it differently and holds different data. The broad frame a person sets stays as it is; as the machine mines the data, it keeps finding further kernels and adds them on top. That is how the structure grows on its own.
Kernel Ontology
Once a person has set the broad frame, the data itself organizes and generates the relationships and builds up the network. AI does the finding — the things in the data, and the relationships that join them — but it does not stand in for judgment. Every connection carries where it came from: which data, which version, which location. Anything that cannot cite its source is not recorded. Whatever the machine adds is marked, so a person can review it later.
It gets denser as it grows
Every recorded connection stays in the ledger and becomes the starting point for the next analysis. The same data never has to be read twice, and as data accumulates so do the relationships already drawn. Structure is an asset that compounds rather than one that depletes.
One principle, applied differently per brand
Relationships between patent claims and standard specifications (SEPPi), relationships between patent claims and real products (IPSnap, planned) — the subject differs, but the principle is the same: define the kernel, and build up the relationships with their evidence attached.