Skip to content

AI pilot licence

Every figure here is computed from the session logs. None of it is self-assessed.

AIP-2026-0136class ATP · issued 2026-03-08
Sessions136
Messages373,258
Projects70
Models flown8
Active days40
Longest streak10 d

ai-pilot-data.json, generated Sep 15, 2026

Ratings

Instrument ratings are keyword coverage over the session transcripts, so they measure what the work touched rather than how well it went. Competency scores come from the same logs and are capped at 100.

lowerhigher

Ratings0 to 100
By instrument
AI / ML79
Backend75
Security64
Systems58
Frontend48
Data Engineering47
IoT / Edge47
DevOps39
By competency
Tool Mastery95
Debugging95
Multi-Project95
Iteration75
Architecture71
Planning49

Fleet

Share of output tokens by model. Balanced piloting style: 81% collaborative, 55% plan first.

Type ratingsshare of cost
By model
opus-549.6%
fable-525.5%
opus-4-812.1%
fable-5-111.5%
opus-4-71.0%
sonnet-50.1%
opus-4-60.1%

Token economy

Tokenscumulative
Input4.3M
Output281.2M
Cache read88.5B
Cache efficiency98.3%

Token counts are abbreviated. Hover any of them for the exact figure. Pilot analytics cuts the same record by project, by hour and by day.