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. What NeuralMind itself does: transmits no repository content, and sends no telemetry.
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.