Eighteen weeks, measured

Week 1: measure before you train

Experienced engineers tend to overrate where they start. A cheap, honest baseline tells you where the hours should go.

• Solve one unseen easy in 30 minutes or less, narrating out loud, in a blank editor without running it (try a warm-up mock round).
• Attempt two unseen mediums with a 30-minute cap each. Note your time, whether you stated the complexity without being asked, and whether you tested an edge case.
• Record yourself once. Note any freezing, silent stretches, or declaring done without testing.
• Write your target list: 8 to 15 companies across a few bands, with the level you are aiming for at each.

The weeks

1. Weeks 1 to 3Fundamentals ramp: read Python fluently, count costs, and understand lists and hash maps from the inside.
   00 The interview and this lab01 Python, one step at a time02 Counting steps: Big O03 Arrays and hash maps

   Every lesson's exercises; the module problems as they come. No timing yet.

   • Week 1 diagnostic done (above).
   • Module 1 checkpoint passed: loops and state feel automatic.
   • Module 2 checkpoint passed: you can state the cost of any loop.
   • Module 3 checkpoint passed: two sum, group anagrams and prefix sums from a blank editor.
2. Weeks 4 to 8Blocked pattern work, one pattern at a time: pointers and windows, stacks, sorting and binary search, linked lists, recursion and trees.
   04 Two pointers and sliding windows05 Stacks and queues06 Sorting and binary search07 Linked lists08 Recursion and trees

   6 to 8 problems a week, 3 to 5 in a row per pattern. Re-solves from the Practice queue every session. Struggle cap: 20 to 35 minutes, then read the approach.

   • Every miss logged in the error journal with its type.
   • Practice queue cleared at least four days a week.
   • Checkpoints passed for modules 4 to 8.
   • From about week 6: timed attempts against each problem's target time.
3. Weeks 9 to 12The remaining patterns, then mixing: heaps and tries, graphs, backtracking, dynamic programming, greedy.
   09 Heaps and tries10 Graphs11 Backtracking12 Dynamic programming13 Greedy and a little math

   New patterns blocked, then mixed sets twice a week. First mock rounds (warm-up and Google style) once a week.

   • Checkpoints passed for modules 9 to 13.
   • A mixed set each week: name the pattern before opening each problem.
   • At least three mock rounds logged.
4. Weeks 13 to 16Interview mode: recognizing patterns, the live-round script, testing without a run button, multi-part builds and hard problems.
   14 Interview mode

   Mock rounds become the spine: one or two a week, plus error-journal-driven drilling on the dominant miss type. Popular hards once mediums come comfortably.

   • Module 14 checkpoint passed.
   • Every mock run with the full script: clarify, plan, implement, trace, summarize.
   • One follow-up of each of the five types answered out loud in a mock.
   • At least one mock with another person, ideally someone who has interviewed.
5. Weeks 17 to 18Calibration and the readiness gate. Lower-stakes loops first; dream companies only after the gate.

   Two or more human mocks. Final 48 hours before any loop: no new problems, sleep protected.

   • Readiness gate passed (below).
   • Recruiter questions asked for every scheduled loop (below).

The shape of a week

Count sessions completed, not problems solved. A sustainable week repeated eighteen times beats a heroic week abandoned twice.

What the rounds look like (as of late 2026)

From the research dossier, with its evidence labels. Formats change; confirm each one with your recruiter.

Booking loops in the right order

• Run lower-stakes loops first and your top choices last; early loops double as practice and can produce competing offers.
• Never use a company with a long cooldown (Meta is reported at 12 months by default, even after a screen) as a practice loop.
• Postponing an unready loop at a large company is almost always free. Ask for 3 to 4 weeks with a specific date.
• Start slow funnels early on paper (Google is reported at 4 to 8 weeks, 12 or more with team matching); fast loops (Amazon, officially about 3 to 6 weeks) work well as early practice and leverage.

Ask every recruiter, before agreeing to dates

• What level is this loop calibrated for, and at what level would an offer be written?
• What is the round composition: which rounds are coding, design, behavioral? Do any rounds run code?
• What is the AI and outside-tool policy, round by round? (Silence is not permission.)
• What is the cooldown if I do not pass, at the screen and after the full loop? (Get it in writing.)
• What timeline should I expect, and does team matching happen before or after the offer?

The readiness gate

All of these before a loop at a company you care most about. If they are not true yet, postpone.

• Passed most of your last five mock rounds with unseen problems (the trend, not one result).
• Unseen mediums solved, tested and narrated in 25 to 30 minutes, complexity stated without being asked.
• You trace your own code and catch your own bug before saying done, every time.
• You handle a follow-up (complexity, scale, constraint change) with time left.
• AI and tool policy confirmed in writing for every scheduled round.

Your last five mock rounds

No mock rounds logged yet. Run one from Practice, then score it; the last five show here.

The last 48 hours

• No new problems or patterns in the last 48 hours: light review of your retired problems and journal only.
• Logistics checked: links, the editor you will use, a quiet room, one monitor, notifications off.
• Seven to eight hours of sleep both nights.
• One sentence ready for a hard first question: narrate, break it down, state the brute force, then improve it.