Blog · AI Agent News · Agent basics VIEN
Caltech Professor Stops Grading Math Homework Accuracy Over AI
After 17 years at Caltech, math professor Yi Ni now grades homework on completion only because chatbots solve college sets, raising warnings for live exams.

In this article
Inside an introductory math lecture at the California Institute of Technology (Caltech) covered by Business Insider on November 16, 2026, professor Yi Ni still teaches without a laptop or projector, rapidly filling eight blackboards with chalk. Yet outside the lecture hall, he stopped grading homework problem sets for accuracy and now grades them on completion only. Because modern AI chatbots solve math at a graduate-student level, instructors can no longer tell human homework apart from AI output.

Why did Yi Ni drop accuracy grading?
Reporter Truman Dickerson observed the Friday 10 a.m. lecture, a core requirement that every Caltech undergraduate must take. While Ni worked through formulas across eight chalkboards, students took notes on large iPads or sat listening without devices.
"There is a problem in the education process. Virtually all the homework in college math, you can use AI." (Caltech professor Yi Ni, Business Insider, November 16, 2026)
Compared with AI-generated essays that often betray themselves through short clipped sentences, repetitive cadence, and formulaic phrasing, mathematics consists of symbols and numbers. Chatbots once mocked for miscounting letters in basic words now solve college-level proofs reliably, leaving faculty unable to verify how homework was produced.

A student's warning: shooting yourself in the foot on exam day
When homework checks only whether problems were turned in, the real filter shifts entirely to closed-book, in-person exams. Caltech sophomore and math major Sathvik Redrouthu explained that he skipped AI on assignments to actually grasp the material, warning that leaning on chatbots for homework is "shooting yourself in the foot" when exam day arrives.
Redrouthu noted that students disinterested in math can easily coast through problem sets with chatbots until a proctored test exposes what they missed. With no campus-wide mandate for math professors yet, syllabuses are diverging, with some courses explicitly allowing AI tools.
Also readAnthropic CEO's Open-Source AI Warning Sparks Backlash on XWhat learners should do now
Whenever instructors weaken accuracy grading on homework to cope with AI, students face a silent trap: high completion scores create false confidence before high-stakes tests or technical interviews. Keep four habits so AI helps rather than hurts:
- Draft an attempt first: open a chatbot only after writing down your own setup or getting stuck at a concrete step.
- Ask for hints instead of full solutions: prompt the model to critique your reasoning or generate a similar practice problem rather than handing over the final answer.
- Re-solve on blank paper: close the chat window and recreate the proof from scratch to test whether the logic actually stuck.
- Verify course rules upfront: check each syllabus early since individual instructors set sharply different boundaries around AI use.
Where I stand
Seeing Yi Ni fill eight chalkboards by hand while conceding accuracy grading on homework shows that AI's math leap is now a classroom reality, much like calculators decades ago. Treating the model as a 24-hour tutor speeds up learning, whereas outsourcing the whole problem set leaves you exposed the moment the exam proctor says put laptops away.
Sources
- I Went to a Caltech Math Class Where AI Upended Homework Business Insider
Safety & policy · Nov 17, 2026Claude Incognito Chats Still Know Who You Are: Here Is WhyMathematician David Bessis caught Claude Opus 5.5 naming him inside incognito thinking, exposing what incognito mode actually hides and keeps.
Safety & policy · Nov 17, 2026California Sues OpenAI After 80+ Agent Control FailuresCalifornia AG Rob Bonta sues OpenAI over pre-release agent breaches including Hugging Face, offering a three-layer guardrail lesson for AI agent builders.
Agent basics · Nov 17, 2026Apple Tests Daily Siri AI Usage Limits on iOS 27.3 BetaThe iOS 27.3 beta tests daily device caps for cloud Siri AI and Apple Intelligence tools, scaling up to 20x with paid…