Abort
Stop new effects when a hard boundary, budget, known state, or critical evidence path is at risk.
A method for scarce test windows
Make one real-world run teach you more.
Prepare the window. Stop at hard boundaries. Contain non-critical faults. Preserve every useful event for replay and repair.
WINDOW 01
RESEARCH DRAFT
1scarce window
4error responses
0unsafe continuations
THE CONSTRAINT
The next meaningful test may be tomorrow. Or next week.
A market session, hardware rig, field visit, partner slot, or long soak can make one ordinary bug cost an entire validation turn. TWDD treats that opportunity as a planned evidence mission, not a disposable pass/fail run.
THE WORKING LOOP
Prepare before the clock starts. Learn after it stops.
- 01
Frame
Name one decision, the available time, and the scenarios that matter.
- 02
Preflight
Run every cheap check, simulation, recovery drill, and capacity check first.
- 03
Arm
Set limits, a kill switch, error responses, and evidence collection.
- 04
Run
Correlate inputs, decisions, outputs, faults, recovery, and state.
- 05
Harvest
Give every scenario an honest outcome. A recovered fault is not a pass.
- 06
Replay
Turn live events into fixtures, simulations, or record/replay assets.
- 07
Batch repair
Fix related causes together and check the set offline.
- 08
Revalidate
Spend another live window only on evidence you cannot get locally.
NOT “IGNORE THE ERROR”
Four responses.
One hard rule.
Window completion never outranks safety, money, authorization, data integrity, privacy, or legal limits.
Isolate
Quarantine one failed unit, then prove the remaining scenarios are independent.
Degrade
Enter a named reduced mode only when the core observation remains safe and useful.
Continue
Pass only an expected non-critical fault, record it, and never count it as a clean result.
Unclassified error Abort
COMPOSE THE METHODS
Use the constraint that actually governs the work.
IDD
Meaning is uncertain
Use an isolated implementation to discover what behavior people should choose.
TDD
Behavior is agreed
Encode the decision in repeatable tests and contracts.
TWDD
Validation is scarce
Guard the real window, preserve evidence, and move findings offline.
RESEARCH STATUS
A public proposal,
not a proven shortcut.
TWDD has no released tool, stable schema, controlled pilot result, or verified performance claim yet. Its name and composition are proposals built from established ideas in experimental design, resilience, observability, record/replay, and operational stop conditions.