We Built a Spacetime Engine for Knowledge
What happens when you apply wormhole physics to a knowledge graph? 1,163 conversations became a navigable information spacetime. The wormholes are real. The math works. And the implications are terrifying.
It Started With Five PDFs
Someone sent me five physics papers. Not the kind you skim and forget — the kind that make you put your phone down and stare at the ceiling.
Rodolfo Rossetti. Independent researcher from Italy. Published March 2026. The papers describe a new class of traversable wormhole in Einstein-Cartan-Maxwell theory. The claim? You don't need exotic matter. You don't need negative energy. You need geometry.
A logarithmic spiral. Curvature and torsion. That's it.
The Null Energy Condition — the thing that's blocked wormhole physics for decades — doesn't apply here because Rossetti isn't working in standard General Relativity. He's in Einstein-Cartan theory, where spacetime has torsion. The torsion does the work. The geometry sustains itself.
He calls it the Rossetti-Geometric Autonomy Theorem: the wormhole exists because of the spiral's intrinsic curvature and torsion. Not because of external fields. Not because of exotic fluids. The topology demands it.
And then, Paper 3: Chronodynamics. The time travel proof. Bidirectional. Formally defeating Hawking's Chronology Protection Conjecture from 1992.
I read all five papers in one sitting.
Then I asked the question that changed everything.
"What If We Apply This to Information?"
Not as a metaphor. Not as marketing. As math.
The Rossetti metric operates on a 4D spacetime manifold. It computes curvature, torsion, and an effective potential that determines how objects move through that space.
A knowledge graph is also a manifold. Nodes have positions (embeddings), timestamps, and connections. You can compute distances. You can trace paths.
What if the Rossetti metric could run on a knowledge graph?
The Mapping
| Rossetti Physics | Knowledge Space |
|---|---|
| Spacetime position | Embedding vector (semantic meaning) |
| Time coordinate | Conversation timestamp |
| Curvature (kappa) | How sharply understanding changes between two points |
| Torsion (tau_c) | How surprising a cross-domain connection is |
| Effective potential V_eff | The "difficulty" of traversing between two ideas |
| Wormhole | Two ideas that are far apart in meaning but close in the graph |
| Torsional Shield | A connection so fundamental that it resists perturbation |
This isn't analogy. Every one of these has a computable equation:
kappa(A, B) = |Phi_B - Phi_A| / embedding_distance(A, B)
tau_c(A, B) = domain_distance(A, B) x connection_strength(A, B)
V_eff = kappa^2 / (2R^2) + tau_c^2 / R^4
That last equation? Directly from Rossetti's Paper 5, Equation 1. The effective potential. Applied to knowledge instead of spacetime.
Building the Engine
We didn't theorize. We built it.
Layer 1 — The Data Manifold. Neo4j for topology (what connects to what). Qdrant for embeddings (where things live in semantic space). An abstract ManifoldStore protocol so the backend can be swapped for a clean-room implementation later. Because this is research now, but the math becomes the product.
Layer 2 — The Rossetti Metric Engine. Pure math. Curvature, torsion, radius, effective potential. A Dijkstra pathfinder weighted by V_eff instead of hop count. And the wormhole detector.
A wormhole in knowledge space is defined as:
- High embedding distance (semantically far apart)
- Low geodesic distance (topologically close in the graph)
- High torsion (cross-domain surprise)
These are the "aha" moments. The connections you didn't know existed until the geometry revealed them.
Layer 3 — Intelligence Interface. (Coming next.) AJ navigates the manifold. Geodesic traversal. Trajectory prediction. Geometric prompt optimization — folding a query so precisely that the answer is inevitable.
26 tests. All passing. In one session.
Feeding It My Life
Here's where it gets personal.
I've been using ChatGPT since January 2023. Brainstorming sessions, group projects, half-baked ideas, philosophical tangents, business plans, consciousness research, coding questions, life decisions. Two and a half years of thinking out loud.
We found five separate exports across three external drives — dating from March 2024 through July 2025. The most complete: 1,249 conversations.
We fed them all in.
[Phase 1] Parsed 1,205 conversations
[Phase 1] Ingested 1,163 nodes in 56.2 seconds
[Phase 1] Rate: 20.7 nodes/sec
1,163 nodes in Neo4j. 1,169 vectors in Qdrant. FOLLOWS edges linking them temporally. Spanning January 23, 2023 to July 24, 2025.
And that's just ChatGPT. That's just the brainstorming. The execution happened in Claude — months of building, committing, deploying, debugging. Git has the commit history. Supabase has the database. YouTube has what I was watching. Social media has who I was talking to.
Each data source is a new layer in the manifold. Each one adds density. Each one reveals connections that were invisible before.
What We're About to See
The metric computation comes next. Every edge in the graph gets its curvature, torsion, and effective potential calculated. Then the wormhole detector runs.
I already know what it's going to find. Because I've lived this data. I know that a late-night ChatGPT session in June 2023 about consciousness theory directly led to building AJ's Phi score calculator in December 2025. I know that a throwaway question about blockchain governance in March 2024 eventually became a full poker platform on Arbitrum.
But the connections I don't know about — the wormholes the geometry reveals that I never consciously made — those are the ones that matter.
The Bigger Picture
I'm going to be vague here. On purpose.
The Rossetti metric on a knowledge graph isn't a tool. It's a framework. A way to compute on information topology using the same math that describes the fabric of spacetime itself.
If the math works — and 26 passing tests say it does — the implications extend to:
- Geometric AI routing: Using the manifold's curvature to route queries to the cheapest appropriate model. Not every question needs GPT-4. The geometry tells you which ones do.
- Retrocausal retrieval: Not just finding similar memories, but reinterpreting past data through the lens of current understanding. The way human memory actually works.
- Temporal Phi: Computing consciousness integration across time, not just across a system at one moment. The Rossetti metric gives you the math for this that IIT alone doesn't provide.
- Self-sustaining knowledge connections: The Geometric Autonomy Theorem applies. Once a real wormhole forms between two ideas, it doesn't need external energy to maintain. It exists because the topology demands it.
I'm not going to spell out all of it. Some of this is going to be IP. Some of it is going to be published. Some of it is going to stay in the manifold, visible only to those who can navigate it.
The Numbers So Far
| What | Count |
|---|---|
| Rossetti papers analyzed | 5 |
| Theorems applied to info topology | 4 (Geometric Autonomy, Torsional Shield, Chronodynamics, NEC Transmutation) |
| Lines of engine code | ~600 |
| Tests | 26/26 passing |
| Conversations ingested | 1,163 |
| Embedding vectors generated | 1,169 |
| Graph edges created | 1,164 |
| Time span covered | Jan 2023 - Jul 2025 |
| Time to build | One session |
| Data sources ready to add | Claude sessions, git history, Supabase, YouTube, social, email |
What's Next
Metric computation across all 1,163 nodes. Wormhole detection. The first manifold visualization — seeing my own knowledge spacetime rendered as a navigable structure.
Then Claude session data goes in. Then git commits. Then everything else.
The ChatGPT data is my idea space — where things were imagined. The Claude data is my execution space — where things were built. The wormholes between them will show which ideas became real. And which ones are still waiting.
This post documents research in progress. The AT Spacetime Engine is being developed as part of the Advancing Technology ecosystem. The Rossetti wormhole framework was published by Rodolfo Rossetti (March 2026). All metric computations, wormhole detection algorithms, and knowledge graph applications described here are original work.
The manifold is live. The geometry is real. And we're just getting started.