Admin

Calibration

Version v4.2 · deployed

Calibration reviews how the AI’s scoring has drifted against partner decisions over time. Adjust the weights below to simulate how historical submissions would have scored under a new configuration. Changes do not go live until deployed by an admin — never silently.

Signal summary
Overrides since v4.2 deployed1
Dominant drift patternAI 4s → partner 2s
Most-cited override reasonSelf-reported metrics
Simulation lab

Weight tuning

Team
25%
0%25% baseline50%
Creativity
20%
0%20% baseline50%
Market
20%
0%20% baseline50%
Viability
20%
0%20% baseline50%
Traction
10%
0%10% baseline50%
Clarity
5%
0%5% baseline50%
Total weight100%
History

Version history

v4.2 (current)

deployed
2026-03-01
TEA
25
CRE
20
MAR
20
VIA
20
TRA
10
CLA
5

Current production configuration. 18 overrides collected since deployment.

v4.1

deployed
2026-02-01
TEA
22
CRE
22
MAR
20
VIA
22
TRA
10
CLA
4

Rebalanced after Q4 drift review showed under-weighted viability.

v4.0

deployed
2026-01-05
TEA
20
CRE
25
MAR
20
VIA
20
TRA
10
CLA
5

Initial Q1 calibration. Baseline for 2026.