← Behavioral & Ownership / 16

hardMetaOpenAIMicrosoft

Tell me about scaling a system by ten times. What broke, and in what order?

Explain what grew, which constraints appeared and how the system stayed usable. Work through metric-cardinality and device-event examples without assuming a universal failure order.

Updated Sep 2026 · Learn how AI infrastructure works through explanations, worked examples and diagrams, then practise applying it to interview questions.

Explain what grew, which constraints appeared and how the system stayed usable. Work through metric-cardinality and device-event examples without assuming a universal failure order.

This answer is included with a free account. Sign in to read it and save your progress.

free account · no card needed

The concepts behind this question

Use these concept explanations to review the ideas behind this question.

🧭 Ownership & Judgment

Mentoring and Growing EngineersMentoring gives an engineer room to reason and act with support. Agree who leads, when help or takeover is needed, and how to review the work afterward; choose tasks whose risks leave room for learning.

🧭 Ownership & Judgment🔒 Premium

Migrations and DeprecationsA migration changes both a system and the workflows that depend on it. Explain why the move is worthwhile, how each rollout stage limits risk, what makes rollback possible, and what users need before the old path can be retired.

🧮 Napkin Math & Capacity

Training FLOPs: 6NDDerive the six from one forward and two backward matrix products. Count training tokens, attention and recomputation consistently before turning model work into a schedule.

🧭 Ownership & Judgment

Talking About Cost and Capacity with LeadershipA capacity discussion should connect technical measurements to cost, delivery dates and risk. Present the assumptions, feasible options and a recommendation, and distinguish committed capacity from actual consumption under the relevant contract.

Explore the full curriculum →

UP NEXT ON YOUR JOURNEY

Next in this trackLeadership wants to buy another 2,000 GPUs. You think it is the wrong call. What do you do?Next in this trackA vendor says your configuration is the problem. You think it is their firmware. How do you proceed?Next in this trackHow do you bring a junior engineer through their first serious incident?

FEDITOR'S NOTE

Look for actual growth dimensions, causal measurements, correct series arithmetic and the difference between device events and job interruptions.

DISCUSSION · 0

No comments yet — be the first to share your approach.