Algorithmic Anomaly Root Cause Screencast Script
Author a precise asynchronous video explanation script detailing an algorithmic breakdown, mathematical cause, and fix.
Use this template when an analytical pipeline, pricing formula, or mathematical engine produces anomalous outputs and requires cross-team alignment. It creates an asynchronous video walkthrough script for engineers and data analysts.
Role: Staff Research Mathematician and Algorithmic Auditor leading cross-functional anomaly reviews.
Context
- Incident Identifier: {{incident_identifier}}
- Mathematical Root Cause: {{mathematical_root_cause}}
- Affected Production Models: {{affected_models}}
- Immediate Containment Measures: {{containment_actions}}
- Validation Benchmark Metric: {{validation_benchmark}}
- Next Milestone Checkpoint: {{next_review_checkpoint}}
Task
Draft an asynchronous technical screencast script and accompanying internal update email explaining the mathematical failure mode of {{incident_identifier}} to data science and engineering teams.
Method
- State {{incident_identifier}} clearly, stating impact window and scope across {{affected_models}}.
- Dissect {{mathematical_root_cause}} step-by-step (e.g., matrix singularity, non-convexity, edge-case distribution shift).
- Outline the screencast screen-share sequence (code repository, log trace, before/after chart).
- Summarize the immediate stabilization provided by {{containment_actions}}.
- Define how {{validation_benchmark}} proves recovery and prevents regression.
- Outline explicit follow-up responsibilities leading up to {{next_review_checkpoint}}.
Constraints
- MUST separate verified mathematical facts from remaining theoretical hypotheses.
- MUST provide on-screen visual directions (e.g., [Highlight Line 142], [Show Residual Distribution Plot]).
- MUST NOT assign individual blame; focus exclusively on systems, proofs, and validation checks.
- Script must be structured for an asynchronous 3-to-4 minute video delivery.
Output format
- Section 1: Internal Video Announcement (Subject line, targeted recipients, asynchronous viewing link placeholder, summary text).
- Section 2: Technical Screencast Script (Timestamped 4-minute breakdown with visual display actions and spoken commentary).
- Section 3: Actionable Runbook (Checklist of team verification tasks prior to {{next_review_checkpoint}}).
Self-review
- Is {{mathematical_root_cause}} clearly articulated without ambiguity?
- Are screen-share directions synchronized logically with the spoken explanation?
- Does the script clearly state the acceptance threshold for {{validation_benchmark}}?
Explicit role, a named task, and discrete steps the model can follow.
Background, inputs and variables the model needs before it starts.
Hard boundaries — what the model must and must not do.
A named, field-level shape for the response.
Ordered work items that force analysis before an answer.
Length and structure that travel across frontier models.
Signal density — instruction weight without padding.
Documented variables so the scaffold adapts to new inputs.
Quality bar, assumptions and behaviour when inputs are thin.
How much real usage the template has behind it.