Prototype Risk Early
A risky unknown may deserve an isolated spike before its feature reaches the front of the delivery queue.
The question
Which uncertainty could invalidate the plan?
Do this
- Write the single question the spike must answer.
- Time-box it and discard code that was built only to learn.
- Do not confuse a prototype with the production slice.
Example
- Checkout remains fourth in the dependency order, but a one-afternoon payment callback spike happens now because the answer affects order state and failure handling.
Ask next
- ?What decision could this invalidate?
- ?What is the cheapest convincing experiment?
- ?What code should not survive the spike?