Use this section when you are close enough to interviewing that concepts alone stop helping and you need reps.
The drills map to the signals an FDE loop tends to test: ambiguity, customer judgment, code learning, system design, and company-specific motivation.
Recommended Order
- The Coding Screen - clear the gate that opens before any signal is graded.
- Decomposition Drills - structure ambiguity before proposing a solution.
- Client-Simulation Drills - handle pushback and customer constraints.
- Codebase / Learning Drills - get useful quickly in unfamiliar code.
- AI System Design Cases - design systems with evals, failure modes, and rollout in mind.
- Values & Hiring-Manager Drills - rehearse the failure story and the "why here" answer.
- Company Differences - tune your prep for different FDE-style loops.
Best Next Step
Start with Decomposition Drills if you want the broadest interview signal.
Jump to AI System Design Cases if your technical design answers still sound like generic SWE system design.
Next: The Coding Screen — the gate that opens before any signal is graded.
How to use the practice system
Reading the worked answers builds recognition; it does not build interview performance. Run each fresh prompt out loud on a clock, record the attempt, and grade the process rather than whether your final idea resembles the example. The strongest signal across FDE rounds is legible judgment under pressure.
Clear the coding floor first, then spend most practice time on decomposition, unfamiliar code, production system design, and customer probes. Behavioral preparation should use real failures and company-specific motivation, not memorized values language. Revisit the same competency with new prompts rather than repeating an answer you already know.
A balanced weekly practice loop
- Two practical coding reps with changing requirements and spoken narration.
- Three five-minute decomposition openings using unfamiliar, non-software customer problems.
- One unfamiliar-code exercise: reproduce, trace, hypothesize, change, and defend.
- Two production AI designs graded on evaluation, failure bounding, rollout, cost, and adoption.
- One live customer pushback session with a partner who does not accept your first answer.