Storage is not trust.
A stored entry is not automatically reliable. AVSM distinguishes between what exists and what has been reviewed, challenged and scoped.
Technology
Knowledge that keeps evolving
Knowledge is not static. Most AI systems treat it like a drawer: once stored, always equally correct. That is not how knowledge works in an organisation. It emerges, is confirmed, contradicts itself and changes.
AVSM stands for Evolutionary Artifact Vector Synthesis Model. In short: infrastructure for the evolution of understanding.
The idea behind AVSM
A statement is more than a sentence
AVSM works with individual statements — for example: “The contract runs until March.”
The point is not simply to store that statement as truth. What matters is why it applies, what supports it and whether it is still current.
AVSM therefore connects a statement with everything that belongs to it. This makes it possible to understand why a statement applies and when it needs to change.
What matters
Four distinctions that make the difference
Retrieval is not governance.
Finding a document is one step. It also matters who reviewed it, what applies when sources conflict and why it should be relevant now.
A document is not yet memory.
Memory includes the history behind it: how a statement gained or lost trust and when its scope changed.
A missing source is not proof.
If a search finds nothing, that does not mean nothing exists.
The architecture
Six layers
This is how the system is structured. The layers separate interpretation, review, decision and execution — because linguistic probability alone grants neither truth nor authority.
Where we stand
What exists today. What comes next.
Anyone who cannot answer the question “How do you know that?” may have built a search engine — but not yet a traceable knowledge system. That is why we clearly separate what already works today from what we are developing next.
Available today
- Tenant separation, proven in operation
- Memory with provenance for every entry
- Authentication, memberships and roles
- Knowledge per tenant, connected to Google Drive and other storage
- The Hale assistant as the way in
Next
- Freely configurable roles and permissions instead of hard-coded defaults
- Areas within a tenant
- A toolkit for building custom assistants
- An expanded administration interface