A coding agent proposes a change; you decide whether the evidence supports it and whether you can explain it. Every example shows generation and verification. A test you can rerun is evidence; an agent saying “done” is a claim to check.
Twelve teaching weeks, then assessment
The course follows the public Fall 2026 calendar of Stanford CS146S, with two verification weeks added. Each teaching week has two 90-minute sessions: a lecture with live generation and verification, then a practitioner guest or a studio. When no guest is available, you work on your project or contribution in a studio with teaching support. Week 13 has the project expo followed by individual defenses. The schedule gives topics and deadlines.
Three lines of work
- Six assignments: A1–A6, due in weeks 1, 2, 3, 5, 6 and 10. Each earns up to 2.5 points (15 total). Keep the artifact, verification results and a short explanation, ending with one decision: accept, narrow, recheck or stop, and the evidence supporting it.
- Four open-source contributions: 30 points across two repositories. A small fix and a substantial change in the first repository earn up to 5 and 10 points in weeks 4 and 8; repeat in the second repository for 5 and 10 points in weeks 9 and 12. A TA performs acceptance review when the repository has no partner contact or the maintainer has not replied within seven days. Your mark does not depend on a maintainer replying.
- A term project: choose a topic and build runnable software with an agent, alone or with one other student. Proposal, technical specification, alpha and final milestones earn 2, 4, 6 and 8 points in weeks 2, 4, 7 and 12 (20 total). In a two-person project, milestone marks are shared; each student defends their own understanding separately.
Your term project has its own topic and repository. Attach an AI-use note once to each work item and reuse its evidence ID rather than copying logs into multiple submissions.
Explain your work without AI
CP1 in week 6 and CP2 in week 11 are thirty-minute, in-class paper checkpoints, each worth 7.5 points. The week 13 defense is worth 15 points: without a device, respond to a change or fault introduced in your own code during a fifteen-minute sitting. These three items provide 30 points of individual evidence without AI. Participation earns the remaining 5 points through guest questions, studio work and the expo.
Passing requires at least 60/100 overall and at least 15/30 across CP1, CP2 and the project defense. Each of those three items permits one resit after feedback when the first mark is below half; the replacement mark is capped at half the item's maximum. Invalid technical or approved accommodation sittings are replaced by an equivalent sitting without consuming that resit. Incomplete assessments stay missing. Read the syllabus for the full rules.
Practise making a decision
Save a prediction before discussion, inspect a candidate or trace, run a discriminating check, revise after feedback and try a changed input. After checking, choose whether to accept, narrow, recheck or stop. Another generation call helps only when it answers a question you can name.
Follow the safety practices in Week 0 setup before using your supplied coding agent. Ask for help when access or a concept prevents progress.