Show how you work with AI.

Practise engineering decisions with AI.

Afterwards you get a private growth report on how you framed the problem, questioned the assistant's suggestions and explained your choices.

You read the task materials and talk your approach through in text, with an interviewer guiding you.

Try a scenario See a sample report

Work at your own pace. You can pause and come back later.

How the reports work

Illustrative example · invented run

01 / Read, question, explain

Review an AI suggestion

Interviewer

A structured checkpoint

Would you accept this change?

Explain what you checked and what you would check next.

Your answer…

Submit checkpoint

Scenario artifact

Read-only task materials

Does the shorter condition behave the same way?

Illustrative Python snippet

# Before
def count_before(items):
    if items is not None:
        return len(items)
    return None

# AI suggestion
def count_after(items):
    if items:
        return len(items)
    return None

Workbench

AI assistant

Assistant

The shorter condition preserves the existing behaviour.

You

The empty list used to return 0. This version returns None. I would check None, [] and a non-empty list before accepting it.

Message the workbench…

02 / See the evidence

Growth report

Private

AI Output Review

Adequate

Evidence

The empty list used to return 0. This version returns None. I would check None, [] and a non-empty list before accepting it.

What to develop

Turn the cases you named into checks with expected results.

This example shows reasoning and proposed checks. It does not show an executed test.

Using AI is fine. The report looks at your part of the conversation.

How it works

1
Work through the scenario with AI. You discuss the supplied materials in text. An interviewer takes you through checkpoints, and some scenarios also have a stakeholder you can question.
2
SHYWAI records the run. It keeps your prompts, the assistant's replies and your answers, and the analysis works from that record.
3
Read your private growth report. It has observations and qualitative assessments for each dimension, plus suggestions for what to practise in your next run. There is no overall numerical score.

What the report looks at

Problem Framing: How you clarify the goal, constraints and missing information.
AI Collaboration: How you give the assistant context, set boundaries and steer its work.
AI Output Review: How you question suggestions and check them against the supplied materials.
Debugging & Verification: How you form hypotheses and describe checks that could confirm or reject them.
Communication & Ownership: How you explain choices, trade-offs and what you remain responsible for.
Risk & Production Readiness: What risks and real-system constraints you notice in this scenario.

Sample growth report

Example only: an invented candidate and run, cut down to one dimension. A real report covers all six dimensions, each with a level, the evidence and what to develop. When the reviews name priorities, it also has a Start here section, and it has a summary from each review.

Example, not a real run

AI Output Review

Adequate

Continuing the illustrative conversation above.

Evidence

The empty list used to return 0. This version returns None. I would check None, [] and a non-empty list before accepting it.

What to develop

Turn the cases you named into checks with expected results. The record shows your proposed checks; it does not confirm that you ran them.

What you can share

From a completed run, you choose what to publish: any of the six dimensions, with its level and, where available, a quote from your own work. You decide which dimensions appear and whether their levels are shown. The rest of the report stays private, and you can revoke the page at any time. The page has a preview card for LinkedIn. For your CV, you can download that card as an image and copy a one-line summary of what you published, with the link.

Example, not a real page

Illustrative AI review · 2026-10-14

Problem Framing

Strong

The empty list used to return 0. This version returns None. I would check None, [] and a non-empty list before accepting it.

AI Output Review

Adequate

Communication & Ownership

Strong

How to read the evidence

The report is built from the recorded run. The AI reviews interpret it and can disagree, and you can read their reasons. Quotes under each dimension, in your report and on a public page, are checked against your recorded messages. They show only words you wrote yourself, never text you pasted from the assistant or the interviewer. Quotes in Process notes are checked against everything you typed, pasted text included.

A checked quote lets you trace where an observation came from. It does not make the conclusion correct. Use the report to look back at a specific decision and talk it over with someone else.

How the reports work and what they can tell you

What we don't do

It is not an AI detector. Using AI well is what the report looks at.
No proctoring: no camera, microphone or screen capture.
We don't rank people or recommend them to employers.
Nothing is published without your consent. Third-party AI models process your answers; see how we handle your data.