What you will be able to do#
- Run optional 30-, 45-, 60-minute, paused, or untimed rehearsals under declared conditions.
- Recover from a wrong direction while preserving a correct simpler solution.
- Debrief observed errors, verify environment and permissions, and choose one evidence-based next session.
Use time as an editable rehearsal constraint#
Engineering Foundry supports timed and untimed work. A 45-minute preset is one practice shape: 0–5 clarify; 5–12 baseline and plan; 12–30 implement; 30–38 test and debug; 38–43 complexity and follow-up; 43–45 summarize. It is not a claim about every employer round.
Choose 30, 45, or 60 minutes to match the skill you are rehearsing, pause when needed, or work untimed while learning. Declare the timer and permitted tools so the attempt record remains interpretable.
Reset after a wrong direction#
Stop adding patches when the invariant no longer matches the approach. State what still works, name the failed assumption, return to the smallest correct baseline, and choose whether the remaining time supports a justified improvement.
Tell the interviewer or practice partner about the change. A controlled reset is more informative than silently replacing the entire solution.
Protect a correct simpler solution#
When time is short, finish the complete correct baseline, test it, and state its cost before attempting an optimization you cannot justify or complete. If you discuss an improvement, distinguish the proposed direction from code you actually verified.
Debrief a failed attempt#
Record the declared duration, mode, exposure, permitted tools, first obstacle, assistance, first failing case, state mismatch, and what remained correct. Separate planning delay, reasoning error, implementation defect, and testing omission because each produces a different next session.
A failed mock can be useful evidence. Do not convert it into a global readiness score or hide it by changing the status to solved after reading the answer.
Write errors that determine the next session#
Useful: 'At minute 18 I still had no invariant because I assumed a sliding window worked with signed values.' Useful: 'The BFS model was correct, but I marked visited after dequeue and duplicated frontier work.' Not useful: 'Too slow' or 'need more Hard problems.'
Interview-day checklist#
Verify the environment before expanding the catalog or changing your routine.
- Confirm the time, task format, language, editor, test runner, and communication channel.
- Confirm permitted libraries, documentation, internet access, AI tools, and debugger use.
- Prepare the required device, power, network, audio, camera, and backup contact path.
- Keep one familiar warm-up and stop adding new topics at the last minute.
- Know how to ask a clarifying question, report a technical interruption, and summarize partial progress.
Request accommodations through the employer process#
Ask the recruiter or designated accommodations contact about needed time, breaks, assistive technology, communication format, or other access changes. Do not infer that a practice setting controls the employer process, and do not disclose more personal information than the process requires.
End with one next-session recommendation#
Base the next session on the first observed error. A missing invariant calls for explanation plus a small trace. A sound approach with slow syntax calls for one focused implementation. A missed boundary calls for a minimal regression exercise. Do not answer every weak attempt with a larger random set.
Optional session-only rehearsal clock
Choose a practice budget or work untimed. The clock is pausable, stays only in this tab, and does not create performance evidence.
Worked example#
Recover at minute 20
During a 45-minute rehearsal, a learner discovers that their signed-sum sliding-window approach can skip valid ranges.
- Pause the implementation and state the failed monotonicity assumption.
- Return to the correct quadratic range-enumeration baseline and verify it on the counterexample.
- If the prefix-state invariant is understood, implement it; otherwise finish and test the baseline rather than bluffing the optimization.
- Debrief the failure as approach selection and schedule one untimed signed-prefix derivation next.
Outcome: The rehearsal preserves truthful evidence: the baseline was correct, the first approach was invalid, and the next session targets the exact reasoning gap.
Apply the idea#
In a 45-minute rehearsal, you abandon an invalid optimized approach at minute 20 but have a correct quadratic baseline. Decide what to do for the remaining time and choose one next session from the observed evidence.
Reveal the worked answer
State the failed assumption and preserve the correct baseline. If the improved invariant can be justified quickly, confirm and implement it; otherwise complete, test, and analyze the baseline and describe the improvement as unverified follow-up work. Record the failure as approach selection rather than general slowness. The next session should be an untimed derivation and trace of the missing invariant, followed later by a separate timed transfer attempt.
Choose the next session#
Choose the duration and permitted tools explicitly, pause when needed, and finish with one next session tied to an observed error.
Open timed or untimed practiceSource and review record#
This original Engineering Foundry guide was checked by Codex against the recorded research scope on 2026-09-05. It has not received independent human editorial review.
Evidence IDs: SRC-DSA-TECH-HANDBOOK-PLAN SRC-COMP-ATLASSIAN-OFFICIAL SRC-COMP-MICROSOFT-OFFICIAL