Mocks

One scored mock every week, transcript logged, scored against a written rubric with levels. I tell you plainly which level you hit — not the one you nearly hit.

Table of Contents

• Why Weekly and Why Scored
• The Rotation
• The Back-to-Back Protocol
• The Hire-Bar Scale
• How a Mock Runs
• The Transcript Format
• Scoring Discipline
• The Full-Loop Simulation
• References

Why Weekly and Why Scored

Two reasons, and the second is the one people miss.

One: nothing in this program is marked complete on the basis of reading. A track is complete when a drill passes, an artifact works, or a mock is scored. Mocks are the only instrument that measures the whole skill — including the parts that only appear under observation, like narration and recovering from being wrong out loud.

Two: a weekly score produces a trend line, and a trend line is the only defence against the two failure modes of a six-month program: quietly plateauing, and quietly panicking. Both are invisible without measurement, and both are obvious with it.

The Rotation

Six-week cycle, then repeat. Adjusted after each monthly diagnostic to weight your weakest track.

From week 9 onward, at least one mock per month is a back-to-back. From week 17, at least one per month is a full-loop simulation.

The Back-to-Back Protocol

Reported: the technical screen is two 60-minute rounds on the same day (../research/source-report.md row 6).

Fatigue is part of the signal, which means a mock run in isolation systematically overstates your performance. Almost everyone is measurably worse in round two — narration degrades first, then clarifying questions, then arithmetic. If you have never measured that degradation, you cannot manage it.

The protocol:

1. Coding round, 60 minutes, hard stop.
2. Five minutes. Not thirty. Five.
3. Design round, 60 minutes, hard stop.
4. Debrief both together, and score the delta between rounds explicitly.

Track the delta over time in the log. It should shrink. If it does not, the fix is stamina work — more back-to-backs — not more content.

The Hire-Bar Scale

Every mock is scored on this, not on L0–L3. Learn what the words mean.

Calibrate to Staff. AI-lab levelling is compressed and the "Senior" title reportedly carries Staff scope (../research/findings.md). A hire (senior) is not a pass for your target level.

How a Mock Runs

1. Cold. You do not know the problem in advance. No warm-up.
2. Recorded. Audio at minimum. Narration is scored, and you cannot score narration you cannot hear.
3. Hard stop. The timer ends mid-sentence if that is where you are.
4. I interrupt. Real interviewers do — with clarifying questions, with pushback, with "what if" at the worst moment. A mock without interruptions trains the wrong thing.
5. I do not help. If you are stuck, you are stuck. What I give is what a real interviewer gives: a nudge if you ask for one, and the cost of asking is recorded.
6. Debrief immediately. Fifteen minutes: the score, the two things that most moved it, and the one drill that fixes the biggest gap.
7. Everything you got wrong enters ../review/ at the 1-day interval.

The Transcript Format

One file per mock: mocks/NN-YYYY-MM-DD-<type>.md.

# Mock NN — <type> — YYYY-MM-DD

## Setup
- Problem:
- Duration / hard stop:
- Back-to-back? (if yes: which round, and minutes since the previous)

## Transcript
(paste or summarize; keep the exact wording of anything you got wrong)

## Score
- Verdict: no hire / hire (senior) / strong hire (senior) / hire (staff) / strong hire (staff)
- Per-dimension scores against the track rubric:

## The two things that most moved the score
1.
2.

## What I got wrong
| Item | The actual gap (not the symptom) | Fix | Into review/? |
|---|---|---|---|

## Narration self-score: _ / 5

## Delta vs the previous mock of this type

"The actual gap, not the symptom" is the important column. "I forgot to handle the empty case" is a symptom. "I don't write the edge-case test before implementing, so I only find edge cases when something breaks" is the gap — and only the second one is fixable by a drill.

Scoring Discipline

The rules that keep this instrument honest:

• Score down when unsure. Between two levels, take the lower one. It costs nothing to be told you are hire (senior) in week 6 and discover in week 14 that you were closer to strong hire. The reverse costs the offer.
• Score the performance, not the knowledge. "I knew that, I just did not say it" scores as not said. The interviewer scores what they heard.
• Hints are recorded. A solution that needed two hints is not the same as one that needed none, and the transcript must show it.
• No retroactive credit. Realizing the answer during the debrief does not change the score.
• I will say it plainly. If a mock is a no-hire, I will say no hire and tell you why. Softening it is the one thing guaranteed to make the real loop worse.

The Full-Loop Simulation

From week 17, once a month. This is the endurance test, and it is genuinely hard.

Five components, not four. Sources disagree on the onsite round count (4 / 4–6 / 6 components) and the agentic round is reported as a selective fifth (../research/source-report.md rows 16, 38). Preparing for more than you expect means an extra round is never a surprise; preparing for fewer means it is.

Score each round separately and score the trajectory. A strong round 1 and a no-hire round 4 is a stamina problem, and stamina is trainable — but only if you have measured it.

References

• ../diagnostics/RUBRIC.md — the level bands each track scores against
• ../research/source-report.md — rows 6, 16, 38 (loop shape)
• ../review/README.md — where every miss goes
• ../STATE.md — the running record of scores