Effectiveness on a Real Production Codebase
A CRM platform (TypeScript/React/Node) before and after a Phase 4 rebuild. Every number traces to a neuralmind benchmark . command.
Token Reduction
48.8×
1,033 vs 50,000+ tokens/query
Wake-up Tokens
455
Per query, measured
Personal Edges
+275%
36 → 135 co-activations
Communities
810
Architectural boundaries
Phase 3 vs Phase 4
Full rebuild on a real production CRM codebase
| Metric | Before | After | Change |
|---|---|---|---|
| Total nodes | 9,190 | 9,293 | +103 |
| Communities | — | 810 | new |
| Personal edges | 36 | 135 | +275% |
| Shared edges | 10,079 | 10,183 | +104 |
| Shared edge weight | 2,774.73 | 2,924.66 | +5.4% |
| Wake-up tokens | — | 455 | — |
| Avg query tokens | — | 1,033 | — |
| Avg token reduction | — | 48.8× | — |
5.4 min full rebuild, then ~30s increments
Full Rebuild
326s
All nodes, edges, embeddings
Incremental (after)
~30s
Changed files + dependents only
What we cannot claim yet
Self-improving
No production fitness data. Architecture is complete but gains are unmeasured.
Team memory that learns
No co-view signal yet. Structural seeds exist, but synaptic learning needs weeks of sessions.
Zero code egress
Overclaim. The agent layer still talks to its model. NeuralMind makes no network calls of its own.
See it in 30 seconds
Clone the repo, run the demo, get numbers on YOUR codebase. Then email the output to [email protected] with your team size for a free full spend model.