Agile Mythology
Autonomy-Washing Warning Chews
For capability claims that need to be muzzled before publication.
Plain English: Autonomy-washing checks keep public language from implying sentience, AGI, exact private translation, runtime authority, or automatic certification.
What this snack represents
For capability claims that need to be muzzled before publication. A red-team chew for copy that confuses tool chaining, memory continuity, benchmarks, or schema conformance with autonomy.
Plain English: Autonomy-washing checks keep public language from implying sentience, AGI, exact private translation, runtime authority, or automatic certification.
Formula: 20m forbidden-claim scan + 20m context classification + 20m copy downgrade + 20m reviewer note
What this teaches: Autonomy-Washing, Claim Boundary, Human Review Gate.
Why teams underestimate it
The shelf label works because the cost is usually hidden across clarification, review, context switching, QA, and follow-up decisions. BarkBowl makes those ingredients visible before the work is treated as free.
Time formula
20m forbidden-claim scan + 20m context classification + 20m copy downgrade + 20m reviewer note
Worked example
One serving of Autonomy-Washing Warning Chews is currently modeled at 1h 20m, or 1.33 person-hours. Multiply servings when the work repeats, spreads across more people, or needs another review cycle.
Warning label
Must not be marketed as proof of safety.
Evidence needed
Before committing this snack, collect an owner, a reason, a time formula, a rollback or no-op alternative, and a short explanation of what the receipt does and does not prove.
How to reduce the cost
Claim Boundary Kibble
No-op alternative
If evidence, authority, or capacity is missing, the mature move may be to record a no-op decision and revisit when the missing ingredient exists.
What this teaches
- Autonomy-Washing
- Claim Boundary
- Human Review Gate
Related lesson
autonomy-washing
Related field report
autonomy-washing
Related glossary terms
Autonomy-Washing, Claim Boundary, Human Review Gate
Related blueprint
Decision record starter ยท Open Blueprint Gallery
Copy decision-record starter
Decision: Consider Autonomy-Washing Warning ChewsTime formula: 20m forbidden-claim scan + 20m context classification + 20m copy downgrade + 20m reviewer noteEvidence: [link or owner] No-op alternative: [what happens if we do not act] Boundary: This is a draft for human review, not an approval. Reviewed at UTC: [timestamp]
Boundary note
This snack produces draft learning and time-burn evidence only; it does not approve fixes, validate credentials, train models, or certify safety.
Dogfooding notes
Exercises product snapshotting, time totals, TTL refresh, telemetry, learning translation, and checkout record creation.