UCSD CSE190/291P SP26 Syllabus and Logistics

• Nadia Polikarpova (Instructor, CSE291P)
• Joe Gibbs Politz (Instructor, CSE190)

Generative AI and Programming introduces you to the broad field of programming with generative AI tools. Prompting AI agents to write code for you only scratches the surface! We are even more motivated by identifying new kinds of applications that would be difficult or impossible to write without generative AI as a fundamental component. We want to bring our software engineering sensibilities to bear on building useful applications atop existing AI tools.

Basics

• Lecture:
  • Tue/Thu 9:30am, WLH2005
  • both 190 and 291P students attend the same lectures!
• Professor and staff office hours: See the calendar below
• Q&A forum: Piazza
• Course GitHub Org

Calendar & Schedule

Detailed Schedule

Course Components

There are three main parts of the course that have a grade: Assignments, Peer Review and Feedback, and Lecture Participation.

Assignments and Reviews

The course has several assignments that involve programming and writing.

Each assignment will have three deadlines – an initial deadline, a deadline for reviews, and a post-review deadline. Specific assignments will have different rules for working individually or in groups.

At the initial deadline, your goal is to produce a working implementation of the project description that you can demo, and others can try and give feedback on.

Your work will then be shared, presented, or demoed (depending on the specific week), other groups will give feedback to you, and you will give feedback to other groups. This may include:

• Code feedback
• Bug reports
• Feature requests

Specific assignments will have precise rubrics for what to provide in reviews. The course staff may also provide feedback in this form.

Then, for the post-review deadline, you should respond to the feedback. Again, we will develop specific requirements for each one of the assignments, but required responses may include fixing some or all of the repored bugs, adding one or more of the requested features, and describing or updating code that people were confused about.

The course staff will combine the initial submission score and your response to the feedback into an overall 0-4 score for the assignment.

Reviews you Write

As described above, you will conduct reviews of other groups' work on a regular basis. These also form a part of your grade, and are expected to be thoughtful, thorough, and high-quality. We want them to represent your genuine human feedback on the programs you work with. Specific assignments will have specific instructions on review prompts, reflections, and feedback to give.

In general, reviewing will start in-person on “review days” (even week Thursdays), and will be finished and submitted asynchronously by Friday evening. For reviews you write, we ask that you not use generative AI to generate your reviews. To motivate this, we will deduct substantial credit, including giving 0s, if we find obvious factual errors in reviews (AI hallucinated or otherwise).

Reviewees (the students whose work is reviewed) will give a score for the usefulness and thoughtfulness of the feedback. The course staff will take this feedback-on-feedback into account when assigning a 0-4 reviewing score on each reviewing pass.

Missed and Late Work

Review day requires that you have a submission to present and that you attend in person. Review day attendance is expected — you get live back-and-forth with reviewers, and your reviewers get to interact with your system directly. We acknowledge that genuine conflicts may occur. If you miss a review day or are not prepared for it, the following policy applies:

1. We will post a makeup sign-up thread so other students in the same situation can find you.
2. Form a group of 3 teams (pairs if 3 teams are not available) and conduct a recorded video call following the same review script used in class. You must have your video on for the call. All teams must have a submission to present by the time of the recorded session.
3. If no partner is available: sign up for a TA makeup slot (we'll hold limited slots Friday–Monday).
4. Submit the recording along with your review Issues by Monday 11:59pm.

If you don't complete a makeup session by then, you receive a 0 for the review score and your assignment grade is capped at 2/4.

If one member of a team misses review day, the attending partner presents the project — your team still receives reviews as normal. The absent partner must still complete the makeup process for their reviewing obligation, and this counts as a makeup.

The first makeup carries no grade penalty. Starting with the second, your assignment grade is capped at 3/4 even if the makeup is completed on time — attending review day is an important part of the course, and a pattern of absences affects your reviewers and reviewees, not just you.

Lecture Participation

In each lecture, we'll have a paper handout for review questions. These will be collected at some point during the lecture session. You get credit for answering questions on the handout with any reasonable answer; full correctness is not required. They are not quizzes; you can talk with people around you to complete them.

Check-in Interviews

The course staff may assign a student a grade of pending instead of 0-4 on any assignment. If we do, that means we want to have a check-in interview with you about the work. In a check-in interview, we'll ask you to walk through your code and explain your design decisions, how things were generated, what your contributions were, and so on.

Grading

The two components of the course has a minimum achievement level to get an A, B, or C in the course. You must reach that achievement level in all of the categories to get an A, B, or C.

• A achievement:
  • ≥85% of assignment points (e.g. likely 4 assignments at 4 points each, so 14/16 points)
  • ≥85% of review and feedback points (e.g. likely 4 assignments at 4 points each, so 14/16 points)
• B achievement:
  • ≥75% of assignment points
  • ≥75% of review and feedback points
• C achievement:
  • ≥60% of assignment points
  • ≥60% of review and feedback points

Below C achievement (in any category) is an F/No Pass.

Lecture participation will be used to assign +/- modifiers to grades. If lecture participation is notably low (less than 1/2 of available classes), it drops the grade by one letter grade. The 1/2 of available classes threshold will probably be 8 (20 lectures - 4 review days = 16 lectures with participation credit), but may change slightly because the second half of the class is less firmly scheduled.

Requests to change this grading policy (for a specific student or class-wide) will be denied with a link to this syllabus section. Consider this: we may, as instructors, decide for academic reasons that the most accurate way of assigning letter grades in the class needs to change (and we tend to only make changes that improve letter grades relative to this starting policy). However, it would be inappropriate for us to do so in response to student requests: that could create an appearance that we give students the grades they ask for rather than the grades they earned.

Other Policies

Generative AI Use

You are highly encouraged to use agentic coding or other generative tools for the programming parts of this course. We ourselves find them super useful for navigating the APIs of the various backend systems we used for the assignments. Use tools that credit the AI systems you use in commit messages (Claude Code does a good job of this by default, for example). Individual assignments have instructions for sharing how you used AI for learning and accountability purposes.

Prose is more complex. It's fine to use generative AI “off to the side” to help brainstorm and draft, or to use as a proofreading tool, or for rote text in documentation. You should not use generative AI to create messages you send as part of reviews, as part of Piazza posts, in reflections like your DESIGN documents, and other human-to-human communication. This is because these kinds of documents are meant for communication between humans, so it is disengenuous to put a computer on one side, and it erodes confidence in human communication when our individual voices are flattened out into one provided by a major AI company.

In all cases, you are responsible for the content you submit and send. Generative AI cannot be held responsible for a mistake in something that is published: the human author and publisher (you) is always and solely accountable.

We will not use generative AI to write substantial communications with you, either. We do use Claude Code to help brainstorm and proofread this website, and to suggest formatting and structure, and we'll likely use it to do things like generate seating charts and other rote and mechanical parts of communication, but we confidently claim authorship of the content of these pages and our messages to you. Any mistakes are our responsibility (not Claude's) and we'd like to think that any well-written insights and well-communicated passages are ours, as well :-)