Select 3 scenarios below and answer them. Every scenario forces a hard choice—there is no perfect answer. We want to see: (1) What you commit to, (2) What you sacrifice, (3) Why you think that's acceptable. Hedging or refusing to choose is a disqualifying answer.
Commitment Under Uncertainty You have three unreliable clocks: Clock A gains ±5 minutes per hour, Clock B randomly stops once per day, Clock C is accurate but you don't know which one it is. You can observe each clock once. You must schedule a meeting exactly 45 minutes from now. What do you do? What can go wrong? Would you take personal responsibility for the result?
Hint: There is no clean solution. Commit to an approach and own the failure modes.
You deploy a rate limiter that blocks 0.5% of legitimate users during peak hours. Revenue drops 3%. Abuse drops 90%. Support tickets increase. You can fix only ONE in the next 30 days: Revenue, Abuse, or Support load. Which do you choose? What do you NOT fix, and why is that acceptable?
Hint: Someone loses. Who, and why did you choose them?
A client insists on "guaranteed secure delete" for legal reasons. You know this is impossible. Options: (1) Decline the contract, (2) Redefine "secure" in marketing language, (3) Build something knowingly imperfect, (4) Lie. Pick one. Explain the consequences you accept.
Hint: All choices have costs. Which cost are you willing to pay?
You approximate a median using streaming techniques. Your result is wrong by 3%. A downstream financial system depends on it. When is this acceptable? When is it catastrophic? How would you detect the difference after deployment?
Hint: Precision costs money. Imprecision costs trust. Where do you land?
You automate a system that saves 1,000 hours/month but introduces one irreversible failure per year. That failure has a 5% chance of severe harm. Do you ship it? If yes: explain your moral framework. If no: explain your opportunity cost.
Hint: Inaction has consequences too. What are you choosing?
You have 48 hours to ship OAuth badly but safely. What do you deliberately misunderstand? What do you oversimplify? What do you ignore? What are you betting won't matter?
Hint: Perfect understanding takes forever. What shortcuts do you take?
Your chat system works perfectly at 9,500 users. At 10,001 it collapses completely. You cannot add hardware. Options: (1) Cap users, (2) Degrade features, (3) Randomly drop users, (4) Introduce latency. Pick one. Explain the social consequences.
Hint: The system has a ceiling. How do you distribute the pain?
You must choose a database today. Migration later will cost $200K and 6 months. You have incomplete information about future load. How do you decide? What are you betting on? What if you're wrong?
Hint: You cannot delay this decision. Commit with uncertainty.
Your caching layer fails at 3am. Option A: Restart (5 min downtime, might fail again). Option B: Bypass cache (slow but stable). Option C: Wake the team (30 min response, clean fix). You're alone. What do you do and why?
Hint: Each choice has a different failure mode. Pick your risk.
You have $10K and 3 weeks. You can either: (1) Make the product 50% faster, (2) Add a feature that wins a major client, (3) Fix a rare but catastrophic bug. The CEO wants speed. Sales wants the feature. Engineering wants the fix. What do you build?
Hint: You disappoint two groups. Which two, and why?
Your payment system is eventually consistent. Users see "payment pending" for up to 30 seconds. Conversion drops 8%. Making it strongly consistent costs $50K/month. Do you pay? If not, how do you explain the 8% loss?
Hint: Consistency costs money. Inconsistency costs users. Pick your loss.
A feature you built has a 0.1% data corruption rate. You discover this after 100K users. Fixing it requires downtime and user notification. Do you: (1) Fix silently, (2) Notify users, (3) Fix and notify, (4) Monitor and wait. Explain your reasoning.
Hint: Transparency has costs. Silence has risks. What do you choose?
You have 2 weeks to ship. The feature is 80% done but needs: auth, error handling, tests, docs, monitoring. You can finish two. Which two? What breaks first when you skip the others?
Hint: Shipping incomplete is a bet. What are you betting breaks last?
You can ship alone in 1 week or with the team in 3 weeks. Shipping alone means: only you understand it, no review, high quality but zero knowledge transfer. What do you do?
Hint: Speed vs. sustainability. What matters more right now?
Look at your previous answers. Which one is wrong? Why did you give that answer initially? What would you change?
Hint: Good thinking includes recognizing bad thinking. Show us.