#The self-mock kit

The reading is finished. This page is the one you use.

Every other page here is input. This one produces output: a number, a recording, and a specific thing to fix. Without that loop, more reading changes nothing.

#1 · Why score at all

Because "that felt okay" is not information. After an unscored practice attempt you know only whether you enjoyed it. After a scored one you know that you never stated a trade-off's cost, or that you were still scoping at minute 22 — and those are fixable.

A number you gave yourself is worth more than a feeling, even when the number is rough. It is comparable across attempts, which a feeling is not.

Interactive simulation — needs JavaScript.

#2 · Setup — 3 minutes

[ ] Pick a prompt you have NOT read the write-up for
        -> question-bank.html, or pick blind from the eight shapes
[ ] Timer: 45 minutes. Visible. Do NOT pause it.
[ ] Stand at a whiteboard, or a blank digital canvas. NOT an editor.
[ ] Start a screen + audio recording. Phone camera propped up works.
[ ] Close every tab. No handbook, no search, no notes.
[ ] Speak out loud the entire time, to an empty room.

The recording is non-negotiable and it is the part people skip. You cannot hear your own filler, hedging, or silence while producing it. You can hear all three on playback, immediately.

No lookups. An attempt with a tab open measures your reading speed. The interview does not have that tab.

#3 · The scorecard

Score each 0–3 immediately after, before watching the recording. Maximum 24.

#	Dimension	0	1	2	3
1	Scoping	Started designing immediately	Asked one or two questions	Asked about scale and core actions	Wrote in/out on the board and confirmed it
2	Estimation	Skipped it	Did maths that changed nothing	One number changed a decision	Numbers drove ≥2 decisions, stated aloud
3	Structure	Interviewer would have had to steer	Drifted, recovered	Followed the phases loosely	Ran the clock; announced each transition
4	Justification	Named technologies, no reasons	Some choices justified	Most choices tied to a requirement	Every major choice traced to a stated requirement
5	Depth	Stayed at box level	One component at level 2	One at level 3	Level 3 on two, and offered the interviewer a choice
6	Trade-offs	None stated	Named without resolving	Resolved with a reason	Volunteered the cost of your own choice, unprompted
7	Failure	Never reached it	Mentioned redundancy	Walked the diagram killing boxes	Also named the assumption the design leans on hardest
8	Communication	Board unreadable; long silences	Followable with effort	Clear, mostly narrated	Narrated throughout; board readable at minute 45

Add it up.

Score	Where you are	Do next
0–8	Not yet a round	Drill the framework alone. Re-run the same prompt tomorrow
9–14	Recognisable, thin	You are reciting. Attack depth (5) and justification (4)
15–19	Would pass some loops	Push trade-offs (6) and failure (7) — the two most-skipped
20–24	Above the bar	New shapes, not repeat prompts. Book a real mock

Row 6 at a 3 is the single strongest predictor. Volunteering what your own choice costs — before anyone asks — is the behaviour that separates hire from strong hire, and almost nobody does it unprompted.

#4 · Watch the recording — 20 minutes

This is the highest-value part of the whole exercise and it is uncomfortable. Watch at 1.5×, with a pen.

[ ] Count the silences longer than 10 seconds.        ______
[ ] Count "um", "like", "basically", "sort of".        ______
[ ] Timestamp when you first drew a box.               ______   (target: ~13 min)
[ ] Timestamp when you started the deep dive.          ______   (target: ~25 min)
[ ] Timestamp when you first mentioned failure.        ______   (target: ~40 min)
[ ] Did you ever say a number and then use it?         Y / N
[ ] Did you ever say "which costs me..."?              Y / N
[ ] Is the board readable in the final frame?          Y / N
[ ] Count choices with no "because".                   ______

The three timestamps are the most diagnostic thing on this page. They are objective, they take seconds to collect, and they tell you exactly which phase is eating your clock. Nearly every failed round is visible in them.

#5 · Symptom → fix

What you observe	Actual problem	Fix
First box after minute 20	Over-scoping	Hard-stop scoping at 5 minutes, even mid-question
Never reached failure	Deep dive ran long	Watch the clock at 40; stop adding, start killing boxes
Many silences	Thinking without narrating	Say the uncertainty out loud — it is scored
Choices with no "because"	Reciting an architecture	For each box ask "what requirement demanded this?"
Only level 2 depth	Breadth reflex	Pick one component; force three levels before moving
No trade-off costs	The commonest gap	Add "which costs me…" to every decision, mechanically, until it is habit
Unreadable board	No layout plan	Scope box top-left, flow left-to-right, leave the lower third empty
Ran out of things at minute 30	Scoped too small	Add a requirement yourself: "what if this were multi-region?"

#6 · A four-week schedule that produces reps

Twelve attempts. Not twelve read pages.

Week	Attempts	Focus
1	3	Framework and clock. Repeat the same prompt on days 1 and 3 — the improvement is the point
2	3	New shapes. Target depth: three levels on one component
3	3	Target trade-offs and failure. Book one real mock this week
4	3	Full dress rehearsal. No new material

Rules that make it work:

1. Never do the same prompt twice in a week -- you are testing recall,
   not memory of yesterday.
2. Score EVERY attempt. An unscored attempt is practice you cannot compare.
3. Read the write-up only AFTER attempting, and mark only where you
   DIFFERED. Places you matched teach nothing.
4. One fix per attempt. Chasing all eight rows at once fixes none.

#7 · The twelve prompts

Do not read the write-up first. Grouped so consecutive attempts exercise different shapes.

#	Prompt	Shape	Write-up
1	URL shortener	Read-heavy key lookup	design
2	Rate limiter	Algorithms, distributed counting	block
3	News feed	Fan-out, hot keys	design
4	Chat	Stateful connections	design
5	Ticket booking	Strong consistency	design
6	Web crawler	Queues, politeness, dedup	design
7	Video platform	Pipelines, bandwidth economics	design
8	Ride-sharing	Geospatial	design
9	E-commerce	Breadth + inventory correctness	design
10	Key-value store	Distributed systems, undisguised	design
11	Logging & monitoring	Write-heavy, cardinality	design
12	Collaborative editor	Conflict resolution	design

If you only manage six: 1, 3, 5, 8, 9, 10. They cover six of the eight shapes.

#8 · The tracker

One row per attempt. A spreadsheet is fine; the columns are the point.

date | prompt | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | total | first box | deep dive | the ONE fix
-----|--------|---|---|---|---|---|---|---|---|-------|-----------|-----------|-------------
     |        |   |   |   |   |   |   |   |   |       |           |           |

Sort by column, not by total. A flat total hides everything. Seeing that row 6 has been 0 or 1 for six consecutive attempts tells you precisely what to work on — and that pattern is invisible in an average.

#9 · When to stop practising

You are ready when, across three consecutive attempts on unseen prompts:

[ ] total >= 18 every time
[ ] first box drawn before minute 15, every time
[ ] you reached failure analysis, every time
[ ] row 6 (trade-off costs) scored >= 2, every time
[ ] fewer than three silences over 10 seconds

Consistency matters more than a peak. One good attempt is luck; three in a row on prompts you had not seen is a skill.

And the thing this page cannot give you: a real person interrupting. Self-mocks train structure, depth and narration. They cannot train being pushed back on, being asked something you did not prepare, or being wrong in front of someone. Book one real mock. Everything here makes that mock more useful; none of it replaces it.