Intelligence
The Intelligence group turns raw telemetry into actionable signals — what’s breaking your flow, where you’re getting throttled, and how to act on it.
| Item | Route | What it shows | Read more |
|---|---|---|---|
| Throttle | /insights/throttle | Rate-limiting events, latency spikes, and recurring drop-off detection — plus hours lost | — |
| Rightsizing | /insights/rightsizing | Model selection and tier recommendations based on your usage patterns | — |
| Tokens | /insights/tokens | Token usage breakdown — cached read/write, output, per model | — |
| Report | /my/report | Chapter-based personal review with trends, strengths/weaknesses, and worst prompts | My Report |
Skill input pages
Section titled “Skill input pages”These are grouped here because they’re behavioral signals, not top-line scores. They also appear from the PRISM Score page as drilldowns.
| Item | Route | What it shows |
|---|---|---|
| Iteration Efficiency | /insights/ie | Convergence + Recovery trend |
| Context Reset Rate | /insights/crr | How often you /compact or /clear |
| Flow Continuity | /insights/fc | Median minutes of uninterrupted coding per session |
| Quality Retention | /insights/qr | Fraction of AI output that survived review |
| Rework & Loop Rate | /insights/rlr | How often recent work needed redoing |
How insights are generated
Section titled “How insights are generated”Once your session data lands, the Prism Engine’s background workers run sub-session aggregation, rubric scoring, and Layer 3 assembly. Most insights here update within a few minutes of a session ending — no manual refresh needed.