2026-09-18 23:31 UTC
Cheating on programming assignments has always been a thing. Attempting to detect cheating has always been a thing. LLMs have done very strange things to their relationship, though.
On one hand, producing code that passes the sniff test at the unmonitored submission level is now trivial; if a 2nd year programmer can write it, the cheapest, worst models can also write it.
1/
Replies (1)
-
@elb@social.sdf.org 2026-09-18 23:34
On the other hand, the distance between vibe coding up a small programming assignment and writing a trivial function under observation seems to have expanded greatly. Without any hard evidence, based only on anecdotes, it seems like the old methods of cheating (Stack Overflow, exchanging notes and small amounts of code with friends, etc.) led to SOME level of learning, while sloperating leads to ... none at all. 2/