Engineering Computing with AI Assistants
Objective: Set up the AI toolchain you will use all semester and practice the working pattern of this course - you specify the problem, the AI helps, and you verify everything - once with a spreadsheet and once with Python. Estimated time: 1-2 hours. Do Part 1 with class 1 (Excel) and Part 2 with class 8 (Python introduction).
Computational tools are how engineers turn equations into decisions, and AI assistants now write much of the first-draft code and formulas. That raises the bar rather than lowering it: when code is cheap, the scarce skills are posing the problem precisely, judging whether an answer is right, and taking responsibility for the result. Every assignment in this course includes Generative AI prompts that quiz you, plant errors for you to catch, and audit your work - this assignment sets up the tools and the habits.
Step 0: Set Up Your Tools
- Choose the Generative AI assistant you will use this semester (e.g., ChatGPT, Claude, Gemini, or Copilot).
- Install the course TA skill in your assistant (instructions in the archive for Claude, ChatGPT, and Gemini). It turns your AI into a course-aware TA that knows the schedule, the assignments, the TCLab projects, and the course AI policy: it tutors, quizzes, and debugs with you - it does not do graded work for you.
- Have a spreadsheet program ready (Excel or Google Sheets). For Part 2 you will also need Python: install it locally or use Google Colab in the browser.
The Pattern: Specify, Generate, Verify, Defend
Every AI-assisted task in this course follows four steps. A good prompt is a small engineering specification:
- Specify: context (who you are, what system), task (exactly what you want), constraints (units, level, format), and verification (how the answer will be checked).
- Generate: let the AI produce the formula, code, or explanation from your spec - and ask it to list the assumptions your spec left open.
- Verify: test against something you can trust: a hand calculation, a known solution, a limiting case, a boundary input.
- Defend: in your report, justify every choice in your own words. "The AI chose it" is never a justification.
Part 1 (Class 1): The Spreadsheet Round
1a. Learn. Use this prompt and answer the questions yourself before asking for explanations:
1b. Specify and generate. Write a one-paragraph spec for a small worksheet - a tank sizing table (volume of 5 cylindrical tanks from diameter and height, flag any tank over 10 m3), a unit-conversion table, or a course-grade calculator. Include the four spec parts. Then:
1c. Verify. Build the sheet, run all 3 test cases, and hand-calculate one output. If any test fails, decide: was the spec ambiguous, or the generated formula wrong? Fix the spec and repeat once.
Part 2 (Class 8): The Python Round
2a. Learn. A reminder for classes 8-10: you write all code yourself while the fundamentals are being built - the AI quizzes and plants bugs, it does not generate your code. Use this prompt:
2b. Write it yourself. Type (do not paste) a short script of your own: convert a list of 5 temperatures from degF to degC and print the average with a label and units. Run it and fix any errors by reading the traceback before asking anyone - human or AI - for help.
2c. The adversary round. Now use the AI the way these fundamentals classes will:
2d. Test the AI's judgment. AI assistants are confident even when wrong. Ask: "What are the limitations of your quiz answers above? Give one case where a rule you taught me fails." Then verify one claim from your conversations against a course page (for example, integer division on Python Introduction). Note whether the AI was right, incomplete, or wrong.
What to Turn In
Submit a report (PDF, about 2 pages; Part 1 with HW1, Part 2 with the class 8 assignment, or as directed by your instructor). You may use Generative AI to help write and format the report, but you must guide it to correct content and you are responsible for every claim. Answer these questions:
- Include your Part 1 spec, the generated layout/formulas, the 3 test-case results, and your hand calculation. What assumption did the AI make that your spec left open, and did it matter?
- If a test case failed: was it a spec problem or a formula problem, and how did the second iteration go?
- Include your Part 2 script (typed yourself) and its output. From the planted-bug round: what was the bug, what wrong output did it produce, and did you catch it before the reveal?
- From Step 2d: what claim did you verify, what did you find, and what does that tell you about when to trust AI output in engineering work?
- Which AI assistant did you choose for the semester, and name one thing that changed in its output quality when you improved your spec.
