SEDIM Architecture

The first source-attributed
AI architecture

Every output traces back to its knowledge origin. SEDIM layers intelligence the way geology layers earth — traceable, composable, permanent.

Why SEDIM

LLMs are powerful.
But blind.

Problem 01
No Source Attribution

Trillion parameters in one opaque mass. Which answer came from which data? Unknown.

Problem 02
Catastrophic Forgetting

New knowledge overwrites old. Continuous learning requires full retraining.

Problem 03
No Knowledge Isolation

Fine-tuning contaminates the entire model. Domain expertise bleeds across boundaries.

SEDIM solves all three. Architecturally.

Architecture

Knowledge that knows its origin

Core Equation
CENTO  =  FACIES  +  Σi  STEMMAi(x)  ·  VARVEi
F
FACIES
Frozen foundation

The base model never changes. All knowledge accumulates on top of an immutable bedrock.

V
VARVE
Knowledge layers

Each domain gets its own low-rank stratum. Independent, versioned, composable.

S
STRIA
Intelligent routing

Per-block routing decides which VARVE serves each query. Attention-level precision.

S
STEMMA
Source attribution

Every output carries its knowledge lineage. Traceable to origin, by architecture.

Impact

What SEDIM changes for the industry

Regulatory Compliance
Traceability by design

EU AI Act and US EO 14110 demand traceability. STEMMA answers automatically — every output carries its knowledge lineage without post-hoc patches.

Continuous Learning
Add without retraining

New domain = new VARVE. No retraining. No forgetting. FACIES stays intact. Knowledge accumulates like geological strata — permanent and non-destructive.

Edge & Mobile AI
Real quality on device

FACIES Q4 4.5GB + VARVE Q8 0.15GB = 4.65GB total. Real 8B-parameter quality running natively on iPhone. No cloud dependency required.

Enterprise Knowledge
Multi-source, one model

Multiple knowledge sources, one model. Each VARVE traceable to its origin. Deploy domain-specific strata without cross-contamination.

Platform

Build on SEDIM

01
VARVE Studio
Layer your knowledge

Upload documents, connect sources, create domain-specific VARVEs. Each layer is independently versioned and deployable.

02
Agent Studio
STEMMA-powered orchestration

Build agents that know which knowledge they are using. STEMMA attribution flows through every agent decision and output.

03
PEC Console
Full control

API keys, STEMMA analytics, inference modes, A/B testing. Everything you need to operate SEDIM in production.

STRATUM API tiers: SURFACE · DRIFT · VEIN · LODE · CORE
Explore the platform →
Research

Backed by research, not promises

0.817
CKA Score
CONFLUX transfer
-6.4%
Val Loss Improvement
Fehm A/B experiment
2x
Convergence Speed
vs. baseline training
0.94
IsolationIQ Score
Knowledge isolation

SEDIM paper targeting Arxiv July 2026. Open source benchmark coming May 2026.

Read the research →
Join

Limited early access

We are opening Nage Platform to the first 100 builders. Get API access, VARVE Studio, and direct research updates.

Request Early Access →