NTFS → NoFS
The file system
had a good run.
Knowledge lives in graphs. Files are just observations.
NoFS is a research initiative exploring what happens when the graph becomes the source of truth.
The Wave Function Analogy
A quantum particle doesn't have a definite state until it's observed. Before observation, it exists as a wave function: a superposition of all possible states. The moment you observe it, the wave function collapses into a single, concrete particle.
But the particle isn't the truth. The wave function is.
Knowledge in the graph is the wave function. It exists as relationships, entities, connections — a living, evolving superposition of everything the system knows. When you ask "show me config.ts" — the wave function collapses. The graph projects a file: concrete, editable, a snapshot of what that knowledge looks like from one angle at one moment.
Traditional filesystems froze the particle and threw away the wave function. NoFS keeps the wave function alive.
Core Thesis
Files are projections
The file system is no longer the source of truth — the graph is. Files become temporary materializations of intent.
Structure is emergent
Directories, modules, packages — all emergent from relationships in the graph, not predefined by human convention.
Instance = Drive
Each project is a cognitive drive. One project, one instance, one graph. Cross-project relationships are handled by federation.
Evolution Phases
| Phase | Filesystem Role | Graph Role | Status |
|---|---|---|---|
| Phase 0 — Traditional | Primary (source of truth) | Does not exist | Past |
| Phase 1 — DreamGraph | Primary (files on disk) | Augments (indexes, enriches) | Now |
| Phase 2 — NoFS Emerging | Secondary (still exists) | Primary (source of truth) | Research |
| Phase 3 — NoFS Mature | Cache + compatibility layer | Sole truth | Future |
Already Proven in DreamGraph
NoFS is not a new project — it is an extraction and formalization of what already works in DreamGraph.
Identity model that replaces file paths (feature://, workflow://, data_model://)
Typed, weighted, directional edges — structure as first-class data
Graph → Markdown projection — proof that files can be generated from graph state
Query-based entity resolution and subgraph extraction
Cross-instance knowledge transfer with anonymization
Each project is its own cognitive universe — the drive model
Primitive Mapping
What old concepts become in the NoFS model.
| Old Concept | NoFS Replacement | Scope |
|---|---|---|
| File | Projection | Instance |
| Folder | Context / View | Instance |
| Path | Query / URI address | Instance |
| Save | Graph mutation | Instance |
| Package / dependency | Federated instance reference | Cross-instance |
| Import statement | Graph edge | Instance or cross-instance |
| Symlink / workspace | Federation edge | Cross-instance |
node_modules | Resolved graph edges (no duplication) | Cross-instance |
| Monorepo | Instance federation | Cross-instance |
The Research
NoFS research is structured as testable hypotheses — 9 test suites covering primitives, projection, identity, reconciliation, determinism, performance, compatibility, mental model, and federation.
Primitives
Do the replacement primitives — projection, context, query, mutation — actually hold?
Suite 1Projection Lifecycle
Can files be generated on demand, edited, and reconciled back into the graph?
Suite 2Identity Model
Can entity URIs replace file paths as the primary identifier?
Suite 3Reconciliation
Can line-level diffs be mapped to entity-level graph operations?
Suite 4Determinism
Are projections stable and reproducible? How much variance is acceptable?
Suite 5–6Performance
Latency targets for projection, reconciliation, and full materialization.
Suite 7Compatibility
Can existing tools — VS Code, Git, compilers, linters — work with NoFS?
Suite 8Mental Model
Will developers accept this? Explainability, escape hatches, and trust.
Suite 9Federation
Instance isolation, cross-instance resolution, versioning, and namespacing.
Built on DreamGraph
NoFS is a research initiative by Mika Jussila, extracting and formalizing the graph-native patterns already proven in DreamGraph — a cognitive development system that treats knowledge as living, relational, and emergent.