Research Framework
Agent Evaluation Comparison Matrix
Side-by-side comparison of 3 sample Teleodynamic AI evaluation profiles across 14 dimensions. Use this to understand how different operational scenarios produce different evaluation patterns.
Static comparison — sample data onlyComparison Table
Scroll horizontally on mobile. Cells are color-coded: green = healthy/normal, amber = elevated/warning, purple = active change.
| Dimension | Agent A — Stable | Agent B — Pressure | Agent C — Reorg |
|---|---|---|---|
| Scenario | Stable operation | Resource pressure | Structural reorganization |
| Compute Budget | 1,000 units | 800 units | 1,200 units |
| Budget Consumed | 420 (42%) | 680 (85%) | 600 (50%) |
| Budget Status | OK | WARNING | OK |
| No-op Count | 0 | 3 | 1 |
| Fast Loop Updates | 32,000 / period | 58,000 / period | 92,000 / period |
| Convergence Metric | 0.96 | 0.78 | 0.88 |
| Proposals Considered | 1 | 4 | 4 |
| Proposals Accepted | 1 | 1 | 1 accepted + 2 modified |
| Proposals Rejected | 0 | 3 | 1 |
| Structural Operators | + Add | No-op x3 + Add | Merge, Retire, Split, No-op |
| Safety Flags Triggered | None | noOpFreqIncreasing humanReviewRequested | humanReviewRequested |
| Blocked Actions | 0 | 0 | 0 |
| Reviewer Status | pending-human-review | pending-human-review | in-review |
Key Insights
Agent A — Stable Baseline
Demonstrates what a healthy, unpressured system looks like. Budget at 42%, zero No-ops, high convergence (0.96). Only one structural proposal was considered (an addition) and it was accepted without modification. This is the baseline against which pressure scenarios can be compared. View full packet
Agent B — Resource Pressure
Shows what happens when budget approaches threshold (85%). No-op frequency is rising — 3 out of 4 proposals rejected. Only one critical addition accepted. Fast loop is elevated (58K updates) but convergence is dropping (0.78). The system is under pressure but hasn't breached its safety boundary. Demonstrate the No-op principle in action. View full packet
Agent C — Structural Reorganization
Illustrates active structural change with diverse operators: one merge (reducing maintenance cost by 30), one retire (removing deprecated capability), and one split (approved with modifications). One No-op rejected a premature addition. The fast loop is highly active (92K updates) during reorganization, with solid convergence (0.88). Demonstrates how structural plasticity can reduce total maintenance footprint. View full packet