The Quick Screen.
A short, sequential diagnostic that confirms your clinical judgement lines up with expert consensus — before you commit to the full calibration. It's private to you: nothing here ever reaches a buyer.
At a glance
- Duration
- ~20 min
- Tasks
- 11
- Type
- Diagnostic
- Gates
- Phase 2
- Price
- Free
A diagnostic, not a verdict.
Phase 1 tells you — and only you — where your judgement is strong and where it needs work, before the procurement-grade Phase 2 measurement. It deliberately produces nothing a buyer will see. Clear it and Phase 2 unlocks.
Eleven tasks, scored honestly.
The screen is built to measure real agreement with expert consensus — fast when you’re clearly ready, rigorous when you’re on the line.
Sequential stopping
You don't always work all 11. Once your agreement with the gold standards is consistently strong, the screen stops early and passes you — no grinding through items you've clearly cleared.
Per-item agreement thresholds
Every task is scored against expert gold-standard answers, with a lower-confidence-bound (LCB) gate. You pass on demonstrated agreement, not on a lucky run.
Anti-luck guardrails
The stopping rule is designed so you can't guess your way through — borderline performance keeps going until there's enough evidence either way.
Cool-down on retry
If you don't pass, there's a short wait before you can try again — enough to prevent over-fitting to the specific items you saw.
A 5-minute warm-up first
Before anything counts, a quick warm-up runs three practice items and one fully-worked example, so the runner UI and the scoring are familiar from the start.
A clear path to Phase 2.
Pass the screen and the Full Calibration opens — the assessment that produces your reliability score and your buyer-facing Reliability Report. Phase 1 is the warm-up lap; Phase 2 is the one on the record.
Ready to screen in?
Finish Foundation, take the Quick Screen, and unlock the full calibration — all free.
Verified against
Clinicians powering AI alignment, training & safety.