English · Español

Lab 00 — 1-Hour Mock Interview Checklist (Asciinema-Recordable)¶

🇪🇸 Drill de auto-entrevista de 60 minutos: 5 preguntas × 12 minutos cada una, grabable en asciinema para revisión posterior. Mezcla whiteboard, coding, paper read, design y behavioral.

Setup (5 min before starting)¶

Terminal in a clean directory: /tmp/mock-interview-$(date +%Y-%m-%d).
Pen + paper at hand for whiteboard questions.
asciinema rec session.cast started. (pip install asciinema if needed.)
Timer visible (terminal or phone).
No browser tabs open except the one for the prep file. No googling during the session.
One bottle of water.

The 60-minute structure¶

Block	Time	Question type	Source
1	12 min	Whiteboard ML question	Random from `theory/01` (Q1-Q25)
2	12 min	Coding drill	Random from `theory/04` (drills 1-12)
3	12 min	Systems design prompt	Random from `theory/02` (prompts 1-5)
4	12 min	Paper read + discussion	Random from `theory/03` (4 papers)
5	12 min	Behavioral anecdote	Random from `theory/05` (anecdotes 1-10)

How to pick the question¶

The night before, run:

# Pick one of each type uniformly at random.
python3 -c "
import random, datetime
seed = int(datetime.date.today().strftime('%Y%m%d'))
random.seed(seed)
print('Whiteboard:  Q', random.randint(1, 25))
print('Coding:      Drill', random.randint(1, 12))
print('Systems:     Prompt', random.randint(1, 5))
print('Paper:       Paper', random.randint(1, 4))
print('Behavioral:  Anecdote', random.randint(1, 10))
"

Don't peek further into the question content until the timer starts.

Block 1 — Whiteboard ML (12 minutes)¶

Min 0-1. Read the question aloud. Restate in your own words. Ask one clarifying question to no one in particular (interview muscle).
Min 1-9. Speak the 3-paragraph model answer continuously. Do not consult notes. Write on paper if it helps.
Min 9-12. Self-fire the L1 → L2 → L3 follow-up tree from theory/01. Speak each answer for ~60 seconds.

Self-grade. Did you cover all three paragraphs? Did you go all 3 levels deep? Score 0-3.

Block 2 — Coding (12 minutes)¶

Min 0-1. Open a blank .py file. Write the function signature.
Min 1-10. Implement the drill. No internet. No copy-paste from the reference stub.
Min 10-12. Run a quick correctness check (synthetic input, eyeball output, or a print of the shape).

Self-grade. Did the code run without errors? Did the output match expectations? Did you hit the common bug listed in theory/04? Score 0-3.

Block 3 — Systems design (12 minutes)¶

Min 0-2. State 3-5 clarifying assumptions aloud. (The interviewer is your phone's voice recorder.)
Min 2-5. Capacity math on paper. QPS, GPUs, memory budget, dollar cost.
Min 5-9. Sketch the architecture diagram on paper. Label every box.
Min 9-12. Name at least 4 failure modes and the mitigation for each.

Self-grade. Did you put a number on every estimate? Did you cite a relevant phase? Score 0-3.

Block 4 — Paper read (12 minutes)¶

The paper-read block is compressed — in a real interview it's 20-30 min. This is a calibration drill.

Min 0-1. Title, abstract, conclusion. Write the 1-sentence claim.
Min 1-4. Figures and tables. Mark the headline figure.
Min 4-9. Method section. Re-derive the central equation on paper.
Min 9-12. Speak: claim, headline figure, equation, one weakness.

Self-grade. Did you nail the claim in ≤25 words? Could you reproduce the central equation without the paper open? Score 0-3.

Block 5 — Behavioral (12 minutes)¶

Min 0-1. Read the anecdote prompt. Take 30 seconds to plan the STAR structure on paper.
Min 1-3. Tell the story aloud. Hit Situation, Task, Action, Result, Reflection. Target 90 seconds.
Min 3-12. Self-fire follow-ups:
"What was the dumbest thing you tried first?"
"What would the opposing view say?"
"What did your manager think?"
"How would you do it differently with infinite resources?"
"What is the next thing you would build on top of this?"

Self-grade. Did you hit 90 seconds on the main story (±15s)? Did you use first-person "I" throughout? Score 0-3.

Post-mortem (10 min after the session)¶

Stop asciinema (Ctrl-D).
Replay at 2x speed. Note where you hesitated, where you used filler ("um", "uh", "like"), where you went off on a tangent.
Score the session: each block 0-3, max 15. Track the score over time.
Update gap list. What did you not know? Pick one item to drill before tomorrow.
Commit the cast (optional): learners/borja/mock-interviews/YYYY-MM-DD.cast with a 1-paragraph reflection in learners/borja/mock-interviews/YYYY-MM-DD.md.

Cadence¶

Weeks 1-2: 1 session every 2 days. Target ≥ 8/15.
Weeks 3-4: 1 session every day. Target ≥ 11/15.
Final week before a real interview: 1 session in the morning + 1 in the evening; alternate blocks.

Optional advanced variants¶

The "noise" variant. Have a friend interrupt you mid-answer with a follow-up. Simulates the real interview rhythm.
The "wrong assumption" variant. Misread the question on purpose and see how soon you correct. Trains noticing your own mistakes.
The "second chair" variant. Pair with another candidate; you grade them, they grade you. Calibrates your own scoring.
The "across companies" variant. Run a session targeted at one company per week (per theory/06). Tune the question mix to that company's signals.

Sample scoring tracker (markdown)¶

Date	Question	Whiteboard	Coding	Systems	Paper	Behavioral	Total	Notes
2026-05-23	Q1, D1, P1, Pap1, A1	2	3	2	2	2	11	Forgot RoPE detail

→ Final lab: 01-paper-pitch-cards.md