Effectiveness Dashboard
Live, anonymized metrics from all projects using the Bastion Security Framework. This data proves the system works — or shows where it doesn't.
Loading dashboard...
What This Measures
| Metric | What it means | Healthy target |
|---|---|---|
| Projects reporting | How many projects are actively syncing | Growing over time |
| Avg coverage | What % of known vectors have mitigations | >70% across projects |
| Accept rate | What % of AI proposals get accepted by humans | >40% (signal, not noise) |
| Blind spot rate | What % of findings are unique to one model | >10% (multi-model adds value) |
| Composites accepted | Cross-domain chains that were real | >0 (composition works) |
| Latent activations | Regressions that activated dormant chains | ~0 (mitigations hold) |
Where the Data Comes From
Every project that runs /bastion:sync automatically submits anonymized metrics. No vector names, IDs, or project-identifying information leaves the project. Only aggregate counts and rates.
Projects can opt out by adding metrics: { opt_out: true } to bastion.yaml.
Data Architecture
Consumer projects → /bastion:sync → anonymized YAML → proposal PR →
GitHub Action → metrics/submissions/ (immutable) →
metrics/aggregate.yaml (derived) → Supabase (dashboard view)
Raw data lives in the GitHub repo (metrics/submissions/). Supabase is a read-optimized projection that can be rebuilt from source at any time.