Coding · 55 min · Michael Poutre · Tue 13:00 PT LIVE — Sun 11:30 PT

49 hours.
The real timeline.

Sunday 11:30am to Tuesday 1pm PT. After sleep, ~24 hours of focused prep. Every hour matters. The hardest problems still get the freshest blocks; warm-ups get squeezed into the margins.

The honest read. This is a serious compression from the original plan. You won't get five reps per hard problem — you'll get two or three for the hardest, one or two for the rest. The lever you can pull: quality over quantity. Slower, deliberate solves with narration practice beat fast solves done silently.
First solve Peak-hours solve Redo from scratch Timed mock (50 min) Rest / fuel / admin
01
Sunday · starting now First solve every problem.
9hfocused
Hour
Mode
Problem
11:30–13:00
In progress
#7Batch Image Processor · finish
You're already in this. 1h30 to wrap: happy path → ProcessPoolExecutor → per-file try/except → retry with backoff → progress reporting. Don't get fancy — get it correct with clean failure handling. Cap at 13:00 sharp even if not perfect.
13:00–13:30
Fuel
Lunch · brief
Eat. Step away from the screen for 20 minutes. Then back in.
13:30–15:30
Peak solve
#3Stack Samples → Trace Events
Hardest problem on the list. Sketch input/output on paper first. Walk two sample traces by hand. Insight: compare consecutive samples; common prefix = still running, divergence below = ends + begins. Get a working solution.
15:30–17:30
Peak solve
#4Distributed Mode & Median
Talk out loud through tradeoffs: exact vs approximate, single-pass vs multi-pass. Count-min for mode, t-digest for median. Code one clean version end-to-end. Be ready to discuss the others — that's worth more than half-implementing all of them.
17:30–18:00
Reset
Walk · no screens
30 min outside. Let the hard stuff consolidate.
18:00–19:30
First solve
#1Web Crawler · BFS + concurrent
Single-threaded BFS within domain. Then ThreadPoolExecutor + lock on seen-set. URL normalization (fragments, trailing slash, host case). Discuss asyncio alternative aloud.
19:30–20:00
Fuel
Dinner
Real food. Off keyboard.
20:00–21:00
First solve
#2Deduplicate Files
Two-tier: group by size with stat(), then hash within size buckets. Chunked hashing pattern. pathlib.Path.rglob. Mention symlinks + hardlinks + permission errors.
21:00–21:45
First solve
#5LRU Cache (Python)
Review given impl, name what's wrong. Write OrderedDict version. If time, sketch DLL version on paper for tomorrow's mock.
21:45–22:30
First solve
#6Tokenize (Python)
Last block, lowest-stakes problem. Read given function twice. Extend. Write test cases. Bed by 23:00 sharp — Monday is the long day.
sleep · 23:00 → 06:30 (7.5h — protect this)
02
Monday · gauntlet day Timed mocks. All seven. Twice for the hardest.
13hfocused
Hour
Mode
Problem
06:30–07:30
Timed mock
#6Tokenize · 50 min
Warm-up problem first. Narrate aloud. 10 min review.
07:30–08:30
Timed mock
#5LRU Cache · 50 min
Both implementations. Defend the choice. Test edge cases (cap=0, cap=1).
08:30–09:30
Timed mock
#2Deduplicate Files · 50 min
Re-derive — don't recycle yesterday's code in your head.
09:30–10:30
Timed mock
#1Web Crawler · 50 min
Aim for clean code over clever code. Concurrent version. Comments showing the model.
10:30–11:30
Timed mock
#7Batch Image · 50 min
You started here Sunday. By now this should feel fluent. Focus on narrating the threads-vs-processes decision.
11:30–12:30
Fuel
Lunch · long break
90 min. Walk. Hardest two problems coming up — protect the brain.
12:30–14:00
Timed + redo
#3Stack Samples · 50min mock + 40min redo
Hardest problem gets double. 50-min timed run first; if stuck, walk through Sunday's approach; then redo cleanly. This is the make-or-break problem.
14:00–15:30
Timed + redo
#4Distributed Mode & Median · 50min mock + 40min refine
Same drill. Focus second pass on the tradeoff-articulation side. Spoken fluency > code volume here.
15:30–16:30
Redo
Pick your weakest from this morning · redo
Whichever 50-min mock bled the most time. Slow, clean second pass. End the work day on a win.
16:30–17:30
Cheat sheet
One-page reference
Concurrency primitives, hash strategies, BFS/DFS templates, OrderedDict snippet, Trace Event JSON shape. If it doesn't fit on one page, you don't know it yet.
17:30–18:30
Rehearsal
Opening 90 seconds · 3× out loud
Restate problem → 2-3 clarifying Qs → propose simplest end-to-end → flag the optimization → ask before coding. Calvin's framework. Make it muscle memory.
18:30–22:30
Stop
No code. Dinner. Walk. Wind down.
Cramming returns are negative now. Sleep is the single highest-EV thing you can do. Lights out 22:30.
sleep · 22:30 → 08:00 (9.5h — on purpose)
03
Tuesday · interview day Arrive warm. Not exhausted.
1hwarm-up
Hour
Mode
Problem
08:00–09:30
Fuel
Wake · breakfast · hydrate
Protein + carbs. Coffee at your usual dose — not the day to experiment.
09:30–10:30
Warm-up
#5LRU Cache · re-implement from scratch
Easy, fluent problem. Goal: blood in the fingers, not new knowledge.
10:30–11:30
Mental rep
Cheat sheet skim · visualize opening
Read the one-pager once. Close it. Picture yourself in the first 90 seconds with Michael — calm body, clear voice.
11:30–12:30
Fuel
Light lunch · walk · decompress
Nothing heavy. Outside if possible.
12:45
Setup
CodeSignal + Meet · 15 min early
Audio. Screen. Camera. Water within reach. Phone silent. Cheat sheet visible but unread.
13:00 PT
Interview
Go.
You've done the work. Now just be the candidate who thinks out loud.

Total reps per problem.

#3
Stack Samples → Trace
3 reps · ~3.5h
#4
Distributed Mode & Median
3 reps · ~3.5h
#7
Batch Image Processor
2 reps · ~2.5h
#1
Web Crawler
2 reps · ~2.5h
#2
Deduplicate Files
2 reps · ~2h
#5
LRU Cache (Python)
3 reps · ~2.5h
#6
Tokenize (Python)
2 reps · ~1.5h

Rules of the room.

  1. Restate the problem in your own words before asking anything.
  2. Ask 2–3 clarifying questions. Edge cases count.
  3. Propose the simplest end-to-end approach. Get a nod before you optimize.
  4. Narrate tradeoffs as you code. Silence reads as stuck.
  5. If stuck for 5 minutes, ask for a hint. That is the correct move, not a weakness.
  6. Test with one small case and one tricky case before declaring done.
  7. If you finish early, propose the next optimization. Don't fill silence with "I think that's it."