Lesson 2 of 4 · Free to read
Exercises
Exercises — Foundations: Modern SDLC + the AI-Augmented Engineer
Watch & learn
Course video
Course video
Keep a prompt log while you work. Speed without a trail is not a skill — it is luck.
How to use this file
- Each exercise is time-boxed and has a measurable success criterion.
- AI prompt hints are copy-pasteable into Cursor.
- Do the foundational ones first, then intermediates; advanced are stretch.
Exercise 1 — Map an SDLC for a product you know
Goal: produce a mermaid flowchart LR of the 7 SDLC phases for a real product, with feedback arrows from Learn back to Plan.
Time: 30 minutes
Level: foundational
Modules touched: [01-foundations]
Setup
- Pick a product: your school capstone, a GitHub library you've contributed to, or a feature from your job.
- Create a file:
learner-artifacts/<your-name>/sdlc-map.md.
Steps
- Sketch the 7 phases as boxes connected left to right.
- Add a back-arrow from Learn to Plan.
- Add a one-line note per phase: what artifact this phase produced for your product. Mark "(skipped)" for phases that didn't happen.
- Save and commit.
Success criteria
- The file has one valid
mermaid flowchart LRblock. - All 7 phases are present.
- At least one back-arrow exists.
AI prompt hints
- "Ask Cursor: 'Turn the following 7 phases into a mermaid flowchart LR with a feedback loop from Learn to Plan: [...].'"
- "Ask Cursor: 'For each phase, suggest one artifact a small team would produce.'"
Solution sketch
A learner picks "Spotify Wrapped" or "their FYP" — both have all 7 phases visible. The mermaid block has 7 nodes and 7 forward edges + 1 back edge. The Learn phase often has "(skipped)" — that's the discussion point.
Stretch
- Annotate each phase with the team role(s) primarily responsible (PM, Eng, SRE, Designer).
Exercise 2 — Annotate AI-augmentation opportunities
Goal: extend Exercise 1's diagram with 3+ AI-augmentation annotations, including the actual prompt you would use. Time: 30 minutes Level: foundational Modules touched: [01-foundations]
Setup
- Open
learner-artifacts/<your-name>/sdlc-map.mdfrom Exercise 1.
Steps
- For 3 phases, add a bullet list below the mermaid block: "AI prompt I would try in this phase: …".
- For each prompt, briefly note: what the input would be, what output you'd expect, what could go wrong.
- Pick one prompt and actually run it in Cursor. Paste the response (or a 5-line summary) into the file.
Success criteria
- 3+ documented prompts.
- 1 actually executed prompt with the response (or summary).
- Each prompt has an input / expected output / failure-mode note.
AI prompt hints
- For Plan: "Group these 50 customer feedback notes into 5 themes with a one-line description each."
- For Test: "Generate 5 negative test cases for this function focused on boundary conditions."
- For Operate: "Given these recent commits and this stack trace, list the 3 most likely root causes ranked by likelihood."
Solution sketch
The "actually run" step is where most learning happens. Expect AI output that is plausible but wrong in places — flag the misses in the file. The grading is on the grading, not the asking.
Stretch
- Run two different models (e.g., Claude vs GPT) on the same prompt. Note the differences.
Exercise 3 — Identify what doesn't change
Goal: produce a written list of 5 things AI does not change about the engineer's job, with one sentence of justification each. Time: 20 minutes Level: intermediate Modules touched: [01-foundations]
Setup
- Open a new file:
learner-artifacts/<your-name>/what-doesnt-change.md.
Steps
- Brainstorm 8–10 things AI does not change.
- Pick the 5 you feel most strongly about.
- For each, write one sentence justifying it from your own experience or what you've read.
Success criteria
- 5 items, each with a justification.
- At least one item references a real anti-pattern you've seen.
AI prompt hints
- "Ask Cursor: 'Critique this list of 5 things AI does not change about engineering. Which is weakest? Which is missing?'"
Solution sketch
Common good answers: accountability, taste, understanding the customer, hard bugs being in the spec, working with humans. Common shaky answers: "creativity" (too vague) and "writing code" (false — AI does change that).
Stretch
- Pair up with another learner. Trade lists. Pick the strongest item in their list and the weakest in yours. Discuss why.
Exercise 4 — Reverse the question: what should NOT be AI-augmented?
Goal: name 3 places in the SDLC where AI augmentation is a bad idea today, with reasoning. Time: 25 minutes Level: intermediate Modules touched: [01-foundations]
Setup
- Append to
what-doesnt-change.md.
Steps
- Pick 3 specific tasks (not "Design", that's too broad — pick "writing the failure-mode analysis for a payment processor").
- For each, name why it's risky: data exposure, lack of accountability, missing context, etc.
- For each, name a partial AI use that would be safe (e.g., "AI drafts FMEA categories but humans rate severities").
Success criteria
- 3 specific tasks named.
- A clear "why risky" + "partial safe use" for each.
AI prompt hints
- "Ask Cursor: 'Help me reason about whether a financial regulator would accept AI-drafted compliance documentation for a payments product. What are the standard objections?'"
Solution sketch
Good answers usually centre on: customer-facing apologies, security postmortems, postmortem decision-making, accessibility audits, anything regulated. The "partial safe use" forces nuance — banning AI entirely isn't a real answer.
Stretch
- Find a real public policy from a company you respect about AI use in code. Compare to your list.
Exercise 5 — Build a "day in the life" timeline (advanced)
Goal: produce a 1-day timeline of an AI-augmented engineer's work — every interaction with an AI tool marked, with a note on whether it added or removed friction. Time: 45 minutes Level: advanced Modules touched: [01-foundations, 02-ai-assisted-development, 17-ai-augmented-workflows]
Setup
- Open
learner-artifacts/<your-name>/day-in-the-life.md.
Steps
- Imagine (or recall) a real workday: 9:00 to 17:30, with breaks.
- Note every meaningful task (≥10 min) and every tool you'd use.
- Mark every AI interaction with
[AI]and one of:+friction/-friction/=neutral. - At the end, sum the AI interactions and write a 3-sentence reflection: did AI net add or remove friction in this day, and why?
Success criteria
- 12+ tasks logged.
- 5+
[AI]interactions tagged. - A 3-sentence net-friction reflection.
AI prompt hints
- "Ask Cursor: 'Given this list of tasks, suggest 3 places where I missed an opportunity to use AI.' Then critically evaluate the suggestions."
Solution sketch
Most learners overestimate AI's net positive on day 1, then come back with a more nuanced view later. The reflection is the assessment, not the timeline. Watch for "AI made me faster but I trusted a wrong suggestion in the PM-meeting prep" — that's the kind of insight we want.
Stretch
- Repeat for a different role (PM, designer, SRE). Compare which roles see net add vs net remove.