1. Which forms of provider dependency do teams most often fail to assess?
2. What is strategic dependency, and what is the defence?
3. Two model providers both run on the same cloud, in the same region, on the same accelerator supply. What is this?
4. Which portability investment is described as the most valuable?
5. Why is the embedding model often a stickier dependency than the generation model?
6. Which model-API failure mode has no error attached and is detectable only by scheduled evaluation?
7. Why must retries be bounded and use backoff with jitter?
8. What can an AI product responsibly commit to in a service agreement?
9. According to the IEA's Energy and AI analysis, roughly what was global data centre electricity consumption in 2024 and what is projected for 2030?
10. How should a product team answer a procurement question about environmental impact?