How I changed teaching after AI managed to do all my homework assignments
A computer science professor details how he completely redesigned an upper-level machine learning course's assessments—shifting to oral check-ins, larger codebases, and LLM-assisted grading—after coding agents made traditional homework trivial.
read it Offers a rare, practical look at how educators are fundamentally restructuring coursework and evaluation to survive the era of coding agents.
summary
A computer science professor shares how generative AI and coding agents rendered traditional homework assignments trivial in his machine learning course, forcing a complete redesign of student assessments. To combat unthinking AI offloading, he replaced written reflections and small coding tasks with mandatory 15-minute oral check-ins with TAs, larger real-world codebases, and video demos. He also leverages LLMs-as-a-judge to autograde code and provide automated feedback on student discussions, drastically cutting manual grading time while shifting evaluation toward classroom-based interactions and exams. The piece highlights a broader crisis in education where AI is breaking evidence-based teaching practices built on low-stakes homework.
key points
- Traditional homework and written reflections are completely obsolete because modern coding agents solve them without human intervention.
- Assessments have shifted to mandatory 15-minute in-person oral check-ins with TAs, video demos, and rigorous defenses of code.
- Assignment scope has been dramatically scaled up from small toy projects to massive production codebases like Zulip to exceed single-step agent capabilities.
- Automated grading using LLMs-as-a-judge and automated feedback loops cut TA grading time by 50 to 80 percent while retaining human oversight for point deductions.
who should read it
University instructors, educators, and technical professionals interested in the impact of generative AI on education and evaluation.
discussion mixed
consensus
Commenters agree that traditional homework and grading are completely broken by generative AI, forcing a painful reckoning across higher education.
pushback
Some argue that the root issue is students lacking intrinsic motivation to learn versus merely chasing grades, while others question the scalability of oral exams.
I was at EduLearn (a major edtech conference) this year and talked to many professors exactly about this. A lot of them are pulling work back into the classroom, but keep running into the same problems:
- some skills only develop through the actual writing (similar to how you can't learn to code just by watching YouTube videos)
- in-class assignments are not always scalable (and students end up using AI anyway)
- students aren't motivated enough to participate actively
I think looking at the writing process is a more realistic approach, so I built Turingo (https://www.turingo.net). It …
Shares a practical edtech perspective and tool for analyzing writing processes rather than chasing fallible AI detectors.
Generated by an LLM from the article and thread. Check the sources before relying on it.