Made for Engineers
Run a Time-Boxed Technical Investigation
Also useful for: people in Technology
120 min
Advanced
Research
Research Something
Need to Research Something
Resolve a technical unknown without letting open-ended research consume the schedule or outrun the decision it needs to support.
Create a free account to save your progress and use this Playbook.
Use this when
You need to answer an unclear technical question, compare competing explanations, validate an assumption, or reduce uncertainty before a design or implementation decision.
You'll need
- Investigation question
- Decision context
- Time box
- Current evidence
- Systems and access
- Constraints
- Experiment risk
- Stopping conditions
You'll get
- The technical question narrowed to a testable form
- Scope and stopping conditions documented
- Material unknowns converted into ranked hypotheses
- High-value hypotheses tested with traceable evidence
- Contradictory findings and limitations recorded
- A conclusion stated with an explicit confidence level
By the end of this playbook
You will have:
- The technical question narrowed to a testable form
- Scope and stopping conditions documented
- Material unknowns converted into ranked hypotheses
- High-value hypotheses tested with traceable evidence
- Contradictory findings and limitations recorded
- A conclusion stated with an explicit confidence level
Sample
This is sample data, not your prompt. Pro unlocks this higher-stakes Playbook. On Free Playbooks you can still copy the assembled prompt into any AI.
Sample input
Investigation question: I need to determine whether new request batching is causing higher API latency under sustained load. Decision context: Keep, revise, or roll back the change before release approval. Time box: 120 minutes. Current evidence: Latency rose 20 percent after deployment to staging, database CPU stayed flat, and only one load test has been run. Systems and access: Staging, load generator, traces, application metrics, deployment history, and code access. Constraints: No production testing, no schema changes, no customer data, and no more than two staging deployments. Experiment risk: Isolated test environment only. Stopping conditions: Stop after repeatable evidence separates the leading explanations, after 120 minutes, or before any production access would be required.
Sample assembled prompt
Investigation conclusion: Batching is the leading... Evidence: - With batching disabled, p95... - With batching enabled, p95 latency... - Database CPU remained stable, weakening... - Traces showed added queue wait... Confidence: High for the staging workload... Medium for production behavior because production... Recommendation: Hold the current batching configuration... Reduce the batch wait threshold and... Remaining risk: The staging workload may... Reopen trigger: New trace evidence showing...