M02 · AI nudge · human operating layer
Cohort orchestration
Route scarce human attention to the participants and teams that are actually stuck.
$ python -m fde_toolkit cohort
Workflow
the order an FDE actually runs it in
- 1Ingest roster with role, skill level and live progress state.
- 2Detect stuck signals: no checkpoint movement, repeated errors, silence.
- 3Pair each stuck participant with the strongest complementary mentor.
- 4Compose skill-mixed breakout rooms so no room is uniformly weak.
- 5Emit nudges with a timebox and a fallback before escalation.
Function output
deterministic trace
pair(AO, DK) reason=stuck 18m on checkpoint 3, mentor strength +2 room(R1)=[AO, EM, JT, PL, SN] skill spread 1.4 support load balanced nudge -> instructor: 'Shared Docker daemon failure across 4 machines — fix once.'
