Narada across comparator-optimized agent-system spaces

Narada is not a single score on a universal benchmark. Its shape depends on the question being asked of an agent system — especially the question of how a system handles the two failure modes of long-range agency: entropy at the language interface and fallibility in reasoning.

This visualization places fifteen systems in a three-dimensional space, then changes the meaning of the three axes when a different system becomes the comparator. Each comparator defines the virtues it optimizes for. Narada is shown alongside Pi, Claude Code, Codex, OpenCode, Goose, Cline, Aider, OpenHands, Letta, LangGraph, Temporal, CrewAI, Agent Framework, and Herdr so that its relative strengths and tradeoffs remain visible rather than collapsing into one ranking.

The coordinates are ordinal architectural judgments on a 0 to 10 scale. They are a thinking instrument, not empirical measurements. Read the argument behind them in Entropy & Fallibility.

How to read it

Select a perspective to make its comparator the canonical point at (10,10,10). The other points show how each system is judged relative to the comparator's chosen virtues. Select a point to edit its coordinates locally, then use the reset controls to return to the canonical values.

The graph is intentionally comparative. It does not claim that Narada is universally better than the other systems. It shows that Narada's distinctive claim is strongest where continuity, execution boundaries, and authority are treated as separate architectural concerns. Under another comparator's priorities, a different system can occupy the optimum.

Residual

The coordinates remain ordinal architectural judgments, not empirical measurements. They are now partially grounded by Async Message Responder: the latest L0–L6 run reports for Narada, Temporal, LangGraph, and Pi feed directly into the feature matrix. The 3D graph itself is still a disciplined hypothesis — useful because it exposes which virtues are being optimized, provisional because the judgments remain contestable.