Project Overview: Assignments, AI Levels, and Model Problems

SENG 365 — Software Engineering

Author
Affiliation

Neil Ernst

University of Victoria

Published

September 11, 2026

Six Assignments, One Team, One Problem

  • Teams of two, one repository
  • Not an open-ended product build. A structured design exercise: specify, design, implement a slice, then revise both under a change request.
  • 30% of the course grade total: 6 × 3.75% for A1-A6, plus 7.5% for debrief and peer-review participation.
  • Focus is design reasoning, not product breadth.

Role of AI

Phase Assignments What it tests
Manual foundation A1, A2, A3 Can you specify, design, and build a slice yourselves?
AI-assisted revision A4, A5, A6 Can you use AI to pressure-test and extend that work without losing judgement?

The Six Assignments

Assignments are described in detail here.

# Focus AI level Output
A1 Requirements specification A Scoped requirements packet
A2 System design A Design packet: alternatives + decisions
A3 Tested prototype slice D Working vertical slice, tests, short explanation

Phase 2: AI help

# Focus AI level Output
A4 AI-assisted requirements revision C Critique + revised requirements for a change request
A5 AI-assisted design revision D Design revision: alternatives, risks, prompts
A6 Tested and deployed prototype E Deployed version, test evidence, final reflection

AI Levels

What an AI Level Controls

AI levels describe how much responsibility AI may take in producing the work. They never transfer responsibility for the result away from the team. You still have to explain, defend, and repair anything you submit.

See the course outline for details.

Why A1/A2 Are Level A

  • You need your own fluency with requirements and design vocabulary before you can evaluate what an AI proposes.
  • A4 later asks you to use AI to critique your own A1.
  • Level A note: AI-generated requirements or design content is not permitted in the A1/A2 submission itself. You are asked to be responsible (mainly to yourselves) in using your own brain.

ai-log.md: Mandatory From A3 On

For Levels C, D, and E, your submission must include docs/ai-log.md documenting, for each significant AI interaction:

Item What to record
Prompt or task What you asked AI to do
AI output What it produced, summarized or excerpted
Evaluation What you accepted, rejected, or revised
Reason Why the team made that call
Verification How you checked the result

Submission Mechanics

Repository Structure

Fixed for both model problems — do not restructure it:

docs/
  a1-requirements.md
  a2-design.md
  a3-retrospective.md
  a4-requirements-revision.md
  a5-design-revision.md
  a6-final-reflection.md
  ai-log.md
src/
test/
README.md

Details

  • Written work in Markdown, in the team GitLab repository.
  • Each written assignment: 800-1200 words, unless a template says otherwise.
  • Keep file names the same.

Grading and Debriefs

  • 6 assignments × 3.75% + 7.5% debrief/peer-review = 30% of the course.
  • Part of the mark is generated through in-person debreifs.
  • Rubrics and marking results will be handled in Brightspace.

Debrief

  • Every assignment will include a short debrief or targeted TA questions:
    • Why did you make this decision?
    • What change does your design handle well? What would still be expensive?
    • Where did AI help, mislead, or hide complexity?
    • Which part of the system can each partner explain independently?

Common Marker Guidance

  • Unassigned features or product expansion don’t raise your mark.
  • Every claim should trace to the model problem, requirements, design, tests, code, deployment evidence, or AI log.
  • Completing the listed artifact is usually Proficient, not automatically Extending
    • extending work shows judgement in why you chose this, what you rejected, and the consequences.

Choosing a Model Problem

Two Options, Comparable Weight

Both are drawn from Candidate Model Problems in Software Design (Shaw, Kang, Petre, 2025), scoped down for this course. Pick one this week.

The TAs will take your team number and initialize a Gitlab repo for the two of you by Wednesday.

Option 1: BikePark

Help cyclists find secure bike parking near an unfamiliar destination.

BikePark starter repository

Option 2: Traffic Signal Simulator

Let a civil-engineering class “play” with signal timing and see the effect on traffic flow.

Traffic Signal starter repository

Picking

  • Both are sized and scoped to reward the same kind of design reasoning
  • Pick on interest and comfort with the domain, not perceived difficulty.
  • Read the full PROBLEM.md and README.md in the starter repo before committing; the “Scope guardrails” section in each tells you what not to build.
  • The dedicated requirements lecture goes deeper on both problems and on using GitLab to elicit and track A1 requirements.

First Week of Lab

  • Go to your lab.
  • Form your team (exactly two, from same lab).
  • Pick a model problem.
  • Read README.md and PROBLEM.md for your chosen problem.
  • Tell your team name and ids to the TA to create teh Gitlab project.
  • Start brainstorming the system boundary, stakeholders, etc. and populating GitLab issues for candidate requirements.