Preparation for a loop is a scheduling problem with a measuring problem inside it. Lab and school studies found three habits that help: blank-page tests, growing gaps, mixed problem types. They never tested interviews, so each result below becomes a rule with a number you can check. Hour figures and thresholds are the author's estimate (not measured), and there is no guarantee of an offer from any plan.
Not advice. General information, and every example is invented.
General information for a book. Study results are summarised from published abstracts read on 31 October 2026, from lab and school settings, not interviews. Check what matters against the original papers and your own results.
Six practice rules
The rules below each name an action, a count and a way to see whether you did it. They apply results from Roediger and Karpicke (2006, testing beats rereading), Cepeda and colleagues (2006, gaps), and Brunmair and Richter (2019, mixing helps moderately, less in maths); Dunlosky and colleagues' 2013 review of ten techniques rates practice testing and spacing highest and rereading low.
- Cold attempt first. Every new problem or design starts blank, with a 10-minute timer, saying the approach aloud before you write. Check: you can name the minute at which you stated an approach. A solution read before an attempt does not count as a session.
- Revisit at growing gaps. Re-solve every problem you missed or needed a hint for on day 2, day 7 and day 21 (day 2, 7 and 14 in the 4-week plan), never twice in one sitting. Check: the third visit is written cold and clean. A pattern is held only when it is.
- One weakness, one number. Each session names one step where your longest pause fell (clarifying, estimating, naming the pattern, testing) and you time that step again next week. Check: the pause in minutes, from the recording, is lower by the rerun. If it is not, change the method, not the hours.
- Mix after you can code cold. From week 3 one session a week gives three problems from different patterns with the name hidden. Check: you name the pattern within 3 minutes on 2 of 3.
- Record one full round a week. Whole round, recorded, scored with that lesson's rubric, 1.5 hours. Check: a dated row in the log below. Re-watch only the longest-pause step, at most 10 minutes.
- Count clean, not volume. No source supports a problem-count target. Count rows that are clean: recorded, no hint, no restart, inside the limit.
INVENTED example ยท rule 3 in numbers 1 min
S. Rao records a system design round in week 1: 9 minutes pass before the first requirement is written down, and his rubric score is 3 of 8 on clarifying. That step is his one weakness. In week 2 he repeats only the first 12 minutes of three different prompts, timing the clarify step each time: 9, 6, 4. In the week-5 rerun it is 4 minutes and the clarify score is 6. Had it stayed at 9, rule 3 says change the method (he would write a fixed list of six questions and read it aloud first), not add a fourth hour.
Next stop: The weekly budget, 2 min
The weekly budget
Pick one weekly total you can hold for every week; a plan you cannot keep is worse than a smaller one you do. The full-round line is fixed at 1.5 hours at every budget because it is the only line that rehearses the real thing. Each column sums to its budget.
Table: Line, 6 h, 8 h, 10 h (6 rows)
| Line | 6 h | 8 h | 10 h |
|---|---|---|---|
| One full round, recorded and scored | 1.5 | 1.5 | 1.5 |
| DSA by pattern, timed, out loud | 2.0 | 2.5 | 3.0 |
| Design: system and machine coding in alternate weeks | 1.5 | 2.0 | 2.5 |
| Stories and pitch | 0.5 | 1.0 | 1.5 |
| Outreach, recruiter calls, admin | 0.5 | 1.0 | 1.5 |
| Total | 6.0 | 8.0 | 10.0 |
Week one is a test, not a reading list. Run each round in your loop once, recorded, and score each with its lesson's rubric: 3.0 hours is about 150 minutes of rounds and 30 of scoring. Pitch 1 minute, coding 45, system design 45, behavioural 15 and hiring manager 30 make 136 (use your invitation's times). An AI-allowed round (60) moves to week 2's full-round line, and a machine-coding build (lesson 7's timed drill, 90 with its notes) to the first week with 1.5 or more Design hours. Write two things per round: the score and the step where the longest pause fell. Then the two lowest rounds get the "weakest two" hours (2.0 in week 3 and 1.0 in week 4 of the 4-week plan; in the 10-week plan, tilt weeks 2 to 4 toward them), and you rerun the same rounds with the same rubrics. If you cannot say a score and a step for each round, you ran a practice session, not a test.
Loops with design or machine coding. The Design line goes to the design round in your loop. A gate round (lesson 7's timed drill, or a 45-minute design round) takes about 1.75 hours with scoring (author's estimate, not measured); a longer build takes the extra from this line. New to machine coding: lesson 14's minimum is about 10 hours; four weeks give 7.5 (Design plus weakest two) even when machine coding takes both weakest-two slots, so move 2.5 hours from the DSA line in weeks 2 and 3, or take the 10-week plan. With two or more design rounds, take the 10-week plan and give Design 3 hours a week from week 2, moved from your highest-scoring line, alternating the rounds, with a first full timed round in week 4.
When hours vanish, drop in this order (the author's judgement): rereading solutions; videos watched before an attempt; any problem-count target; summaries of design topics you are already green on; and last, outreach, the only line that produces a call. Keep the full round until the last week.
Next stop: The 4-week plan, 2 min
The 4-week plan
For a first loop two to four weeks away, 8 hours a week, 32 in all. Fill the "weakest two" column from your diagnostic, not from taste.
Table: Wk, Main job, Weakest two, DSA, Design, Stories, Full round, Outreach, Diag. / rerun, Total (5 rows)
| Wk | Main job | Weakest two | DSA | Design | Stories | Full round | Outreach | Diag. / rerun | Total |
|---|---|---|---|---|---|---|---|---|---|
| 1 | Diagnostic; read the letter | - | 2.0 | - | 1.5 | - | 1.5 | 3.0 | 8.0 |
| 2 | Open doors: resume and asks | - | 2.0 | 1.5 | 1.0 | 1.5 | 2.0 | - | 8.0 |
| 3 | Weakest two; first gate rounds | 2.0 | 1.5 | 1.5 | 0.5 | 1.5 | 1.0 | - | 8.0 |
| 4 | The rerun; last gate rounds | 1.0 | 1.5 | 1.5 | - | 2.0 | - | 2.0 | 8.0 |
| Sum | 3.0 | 7.0 | 4.5 | 3.0 | 5.0 | 4.5 | 5.0 | 32.0 |
Check: 3.0 + 7.0 + 4.5 + 3.0 + 5.0 + 4.5 + 5.0 = 32.0. If the week-4 rerun did not move your lowest score, the checkpoint table says what to change, and it is not "work harder at the same thing". Stories in 3.0 hours: four lead stories, not eight, two recorded takes each, then gate D's three mocks.
Where the gate falls in four weeks (one design round; the author's estimate, not measured). An AI-allowed round's two gate rounds take 2.5 of the weakest-two hours: the first in week 3, the second in week 4 with 0.25 from that week's full-round line.
Table: Wk, Gate work, From the line (3 rows)
| Wk | Gate work | From the line |
|---|---|---|
| 2 | Four lead stories, two takes each; a round that did not fit week one (else your lowest round again) | Stories; full round or Design |
| 3 | Design round 1; Mock 1; Mock 2; the C1 drill | Design; Stories; full round; Outreach |
| 4 | Design round 2; Mock 3 with one partner for 45 minutes; hiring-manager calls 1 and 2, different partners and days; the rerun | Design; full round; rerun |
Gate B's timed attempts take up to 7.5 DSA hours: the eight worked problems twice at 15 minutes (4.0) and the drill set's targets (208 minutes, about 3.5). The 4-week line of 7.0 fits only a reader near B; otherwise make coding a weakest-two round or take the 10-week plan (18 hours, about 10.5 left to learn). From zero: lesson 14. Gate C works the same way: two timed design rounds at 1.75 hours with scoring take 3.5 of the 4-week plan's 4.5 Design hours, leaving about 1.0 to learn lesson 8, which fits only a reader near C.
Next stop: The 10-week plan, 1 min
The 10-week plan
For a reader with two months or more: spacing and mixing have room and two loop windows fit. 8 hours a week, 80 in all; loop days sit on the calendar, not in the budget.
Table: Wk, Main job, Diag., Full round, DSA, Design, Stories, Outreach, Debrief, Loop prep, Offer, Total (11 rows)
| Wk | Main job | Diag. | Full round | DSA | Design | Stories | Outreach | Debrief | Loop prep | Offer | Total |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Diagnostic | 3.0 | - | 2.0 | - | 1.5 | 1.5 | - | - | - | 8.0 |
| 2 | Doors: resume, referral asks | - | 1.5 | 2.0 | 1.0 | 1.5 | 2.0 | - | - | - | 8.0 |
| 3 | Full budget; first calls | - | 1.5 | 2.0 | 2.0 | 1.5 | 1.0 | - | - | - | 8.0 |
| 4 | Full budget | - | 1.5 | 2.0 | 2.0 | 1.5 | 1.0 | - | - | - | 8.0 |
| 5 | Rerun the diagnostic | 3.0 | - | 2.0 | 1.0 | 1.0 | 1.0 | - | - | - | 8.0 |
| 6 | Screens; bunch loops | - | 1.5 | 1.5 | 1.5 | 1.0 | 2.5 | - | - | - | 8.0 |
| 7 | Loop window one: prep | - | 1.5 | 1.5 | 1.0 | 1.0 | - | - | 3.0 | - | 8.0 |
| 8 | Loop window one; debrief | - | 1.5 | 2.0 | 1.5 | 1.0 | - | 1.0 | 1.0 | - | 8.0 |
| 9 | Loop window two: prep | - | 2.0 | 2.0 | 1.5 | 1.0 | - | 1.0 | 0.5 | - | 8.0 |
| 10 | Offers and negotiation | - | - | 1.0 | - | 1.0 | - | 1.0 | - | 5.0 | 8.0 |
| Sum | 6.0 | 11.0 | 18.0 | 11.5 | 12.0 | 9.0 | 3.0 | 4.5 | 5.0 | 80.0 |
Check: 6.0 + 11.0 + 18.0 + 11.5 + 12.0 + 9.0 + 3.0 + 4.5 + 5.0 = 80.0. Weeks 6 and 7 turn from learning to rehearsing the loops you really have. Weeks 8 and 9 hold two loop windows with debriefs. Week 10: offers, negotiation (lesson 13) and the resignation plan (lesson 2).
Checkpoints
Change the plan on a result, not a feeling. Each row is the author's estimate of the first thing to try.
Table: At the end of, If you see, Do (6 rows)
| At the end of | If you see | Do |
|---|---|---|
| Week 3 | No reply from twenty approaches, no recruiter call | Fix the resume and widen the list (lesson 3). More DSA will not open a door that has not opened |
| Week 5 | Recruiter calls, but no technical screens | Re-record the pitch; ask recruiters what they hear |
| Week 5 | The rerun did not move your lowest score | Change the method: more full rounds out loud, less reading. Adding hours to a method that is not working is how people stay stuck |
| Week 7 | Screens lost on the same kind of question | Move hours to that topic for two weeks; retry with a practice-tier company |
| Week 9 | Loops, no offers | Compare debrief notes; repeat the weak round as a recorded full round before booking more |
| Week 10 or end of the 4-week plan | Still no offer | Extend by four weeks, not twelve; reset the list; keep the job |
Next stop: The readiness gate: when to book loops, 5 min
The readiness gate: when to book loops
This page merges the plan check, the drills of lessons 4, 6, 7, 8, 9 and 11 (lesson 4's drill is cheat sheet C1), and the behavioural mocks in part D below (rubric in lesson 10). Book real loops when all parts in your loop hold. All thresholds are the author's estimate (not measured); no reader data exists to say what a ready candidate's log looks like.
Clean means recorded, no hint, compiled and passed on the first run (for code; in lesson 6's sprints, the one run after the hand trace), the trace said aloud, no restart, and inside the limit: 15 minutes for a worked problem written from a blank file; in lesson 6's drill set, each drill's own target (drill 8: 18 minutes; drills 11 and 12: 30 and 35); 25 for any other unseen problem; your invitation's time for other rounds. A machine-coding build is clean with no hint, no restart, typing stopped by minute 82 and 7 of lesson 7's 8 checklist rows.
Table: Part, Pass when, If not, Done (7 rows)
| Part | Pass when | If not | Done |
|---|---|---|---|
| A. The plan | Each round in your loop has a diagnostic recording, a score and a longest-pause step; the rerun moved your lowest score; a recorded full round was in the calendar every week it fitted; you can say what you will change at each checkpoint | Rerun the diagnostic and apply the week-5 row | |
| B. Coding fluency (lesson 6) | On two separate days, at least 7 of the 8 worked problems of lesson 6 compile and pass the example printed with each on the first run, inside 15 minutes, in a blank file with autocomplete off. Then the lesson 6 drill set at 8 or more of 12 clean, at least 6 of them from drills 4-7 and 9-12, which lesson 6 does not solve. Staff and Principal pass on 5 of the 8 and 4 of those unseen drills, lesson 6's senior transcript said aloud and both of its code-reading screens ranked aloud, the second inside 10 minutes, ending with its Principal move (rule, owner, check) | Re-read the selector row and solution of each failed problem; retake no sooner than 2 days later | |
| C. Design (lessons 7 and 8) | Two clean full rounds in a row for each design round in your loop, on different days, the second in the last week before the loop. Pass mark: high-level design, 7 of lesson 8's 8 rubric rows with Numbers and Failure among them (Staff and Principal: Staff signals too); machine coding, lesson 7's timed drill at 7 of its 8 checklist rows, "Core flow runs" and "Correctness" among them; Staff: 8 of 8, and lesson 7's review drill passed | Book the same round again within the week, drilled on the pause step | |
| D. Behavioural mocks (lesson 10) | Mock 1 alone, with the AI card or a friend reading: five questions at 2:00 each, one probe after each (led without authority, lost a disagreement, a failure that cost, a project you chose not to build, why this company; Staff and Principal add a bet that failed; Principal tells S3 and S5-S8 at company scope, lesson 12's story bank). Mock 2 with a partner, 75 minutes: a 2:00 story, eight probes ("What exactly did you do?" to "What if they had said no?"), the story retold at 2:00, 1:00 and 0:30, a 10-minute hypothetical. Mock 3, three people at 45 minutes each (a project deep-dive at senior level; in the 4-week plan, one partner for 45 minutes), graded twice, on the telling and after the probes. Each mock, recorded, passes when every answer is solid or better on all seven rows of lesson 10's rubric, scope at your target level included (Mock 3: on both gradings) | Repeat Mock 2 with a different partner; probes you stalled on become the next week's weakness | |
| E. The screen (lesson 4; drill on cheat sheet C1) | The C1 drill recorded once, all nine rubric lines ticked | Redo it with the page face down | |
| F. AI-allowed round, if in your loop (lesson 9) | Two clean recorded 60-minute rounds with an assistant, on different days, each 11 or more of 14 on lesson 9's rubric with no zero; and lesson 9's practice loop at 6 of 6 on each of three runs, its hours taken from the DSA line. The two 60-minute rounds use the weekly full-round line, alternating with a no-assistant round (lesson 9, "The full round, alone"); in the 4-week plan, the weakest-two hours | Run the practice loop on the class you missed; rebook the round within the week | |
| G. Hiring-manager round (lesson 11) | Two clean recorded 30-minute calls using the lesson 11 drill, with different partners on different days, each at your bar: SDE-1 and SDE-2, 6 of the 7 rubric boxes; SDE-3, 6 of 7 plus a people answer from the evidence column; Staff, all 7 plus one decision at scope across teams; Principal, all 7 plus one decision adopted by several organisations (lesson 12's Reach row) and the bet that failed (lesson 12) | Repeat the drill within the week; the boxes you missed become next week's weakness |
Mock log. One record per round, fails included; a row that is not clean books the same round again within the week.
Table: Date, Round, Minutes of limit, Clean?, Score; longest pause (step) (3 rows)
| Date | Round | Minutes of limit | Clean? | Score; longest pause (step) |
|---|---|---|---|---|
The verdict. All parts in your loop ticked: ask recruiters to bunch your loops into one two-week window, and keep the weekly full round going until the first one. One part missing: book only a practice-tier loop and work that part. Two or more missing: do not book; extend the plan by four weeks and rerun the gate. A fail here is information about the plan, not about you.
Pitfalls
- Quoting a retention percentage. The studies report effect sizes and ratings, not recall percentages; the 10,000-hour claim has no support here.
- Practising only what you are good at. The diagnostic exists to stop it.
Practice with an AI: the prompt kit in the appendix has 1 card for this lesson.
Sources
5 sources
- Roediger and Karpicke, "Test-enhanced learning", Psychological Science 17(3), 2006, PubMed 16507066, abstract read 31 Oct 2026.
- Dunlosky and others, "Improving students' learning with effective learning techniques", Psychological Science in the Public Interest 14(1), 2013, PubMed 26173288, abstract read 31 Oct 2026.
- Cepeda, Pashler, Vul, Wixted and Rohrer, "Distributed practice in verbal recall tasks", Psychological Bulletin 132(3), 2006, PubMed 16719566; Brunmair and Richter, "Similarity matters", Psychological Bulletin 145(11), 2019, PubMed 31556629; abstracts read 31 Oct 2026.
- Macnamara, Hambrick and Oswald, "Deliberate practice and performance", Psychological Science 25(8), 2014, PubMed 24986855, abstract read 31 Oct 2026.
- Not sourced, so the author's judgement: all hour figures, the drop-first order, the checkpoints, the gate thresholds and the plan shapes.