Assignments¶
Two graded surfaces are individual: the Project Pulse assignments and the exams. Everything else is your team's work, covered on the Senior Design Project page.
Individual Project Pulse assignments¶
Five assignments, 20% of your grade, done alone on Project Pulse: the instructor's own Vue.js and Spring Boot application, and the codebase you watch in every Monday and Wednesday lecture.
Each one has the shape of real work on a codebase you did not write: open a well-formed issue, make the change, submit a pull request. Each is framed around the AI workflow.
Graded on judgment, not on whether it ran
An agent can make almost any of these produce working output. That is not what earns the grade. The grade is for the quality of the issue you wrote, the changes you decided not to make, the findings you correctly dismissed, the test you added because you understood what could break, and your ability to explain in the pull request why the change is right. "It ran" is the floor, not the bar.
| # | Assignment | Due | Task |
|---|---|---|---|
| 1 | Hello, Project Pulse | Fri Sep 4 | Set up and run Project Pulse, read one package with the agent and catch a claim the code does not support, open a well-formed issue, and submit a small pull request. |
| 2 | Spec a feature | Fri Oct 2 | Specify a feature Project Pulse is missing: the reminder emails every student whether or not they already submitted. Write the use case, turn it into an agent build-context, and evaluate what the agent assumed that you never told it. |
| 3 | Add a use case | Fri Nov 6 | Take one use case end to end with the agent, wire it into the traceability matrix, and add AI-generated tests that you have vetted. If the use case adds shared code, have the agent commit it first, in a commit of its own, and read it before the agent builds the rest (required, not graded this term). Then review the agent's pull request yourself before you submit it, and write that review into the pull request. |
| 4 | Triage the analyzers | Fri Nov 13 | Project Pulse's SpotBugs and PMD baselines hold 367 findings nobody has read. Make four project-wide rule decisions, then triage your slice: fix the real findings (one shown by a test), suppress the rest with reasons, shrink the baselines, and keep CI green. The triage is the assignment. |
| 5 | Hunt the bug | Fri Nov 20 | Localize a planted bug using logs, traces, and the agent. Fix it and add the regression test that would have caught it. |
Each is due about three weeks after the last, on the Friday of the week its topic is taught or the Friday after it. That Friday's studio then applies the same move to your team's project, so you do it once alone on Project Pulse and once with your team on your own code, in that order. Plan accordingly: Monday and Wednesday lecture is what you have behind you when the individual version is due. None lands during a break. Two share a Friday with a checkpoint, and both are deliberate: Checkpoint 0 is a lightweight kickoff gate and assignment 1 is small, and assignment 2 is submitted to TCU Online before class while Checkpoint 1 is presented in that day's studio, so they overlap in your week but not in the room.
You submit a link, and the pull request is the work. Each assignment is done in your own fork of a class copy of Project Pulse, not in the public repository. Each assignment has its own repository to fork; the link is on that assignment's page. Submit the pull request's URL to TCU Online by the start of class on the due date. The issue you opened, the commits, and the pull request description are the artifact being graded; nothing else is uploaded. Late work takes a 15% penalty per day and is not accepted more than two days late.
Exams¶
| Exam | When | Scope |
|---|---|---|
| Midterm | Mon Oct 19, in class | Weeks 1-7. What to study, week by week, is in the Midterm Study Guide. |
| Final | Mon Dec 14, 8:00 - 10:30 AM | Cumulative, weighted toward weeks 9-14: testing, static analysis, CI/CD, observability and debugging, maintainability and technical debt, ethics. The self-study open-source module is not examined. |
Exams assess whether you can reason about software engineering, which is the one thing the agent cannot do on your behalf in the room. Expect questions that give you an artifact and ask you to judge it: is this requirement testable, is this the right architectural decision and what did it cost, does this trace hold up, what would you have reviewed here.
What "individual" means when you have an agent¶
Read Working with AI. The short version: using the agent on individual assignments is expected. Submitting work you cannot explain is not.