How I changed teaching after AI managed to do all my homework assignments
thelastsoftwareengineer.substack.com
thelastsoftwareengineer.substack.com
I rely on written exams that are open book, but forbid any use of computers and smartphones in class.
The university insists that homework can't be optional, but it lost its meaning. I've tried to explain to my students that it's in their interest to think about home assignments on their own, but the majority will obviously use gen AI, and it's such a waste of time to check and grade LLM output.
I just had an experience where all students were given "sample problems" to try at home and prepare for the written test. Many of them just dumped the document into an LLM, asked it to produce some "exam guide", and showed up with that thing printed out, asking me during the test to explain what LLM output meant.
It seems like the university also has many people pushing for AI use for everything, but I teach basic stuff where the goal is to make students think on their own and digest some fundamental ideas, LLMs can produce perfect solutions, but relying on them is pointless.
Perhaps some people do not belong at university? It is just a waste of everyone time, to insist 50% of population studies until 25 years old.
2. We're also actively giving them less paths in life to begin with. How many high school grads @ 18 will find any job whatsoever without additional school/training, and how else will they afford housing with zero income? The military is the only consistent employer at that age, and the only one who'd provide shelter with the deal.
3. On top of all that, it's in university's best interests to keep 1) and 2) the way they are. Their incentives start at recruitment, and end at enrollment. Past keeping them in college for 4-6 years, they do not care about performance, post-grad job placement, nor societal wellbeing. And part of that is due to a lack of pressure in the recruitment stage to make universities focus on such metrics. Turns out a fancy new gym is a cheaper investment.
Hopefully it doesn’t mean extending high school two years into college, still prepping to make a bet not yet made.
As long as students ask the teacher what they're going to need understanding of maths for they are not prepared adequately.
I don't know about you, but I sure didn't have my life set at 18.I could have easily been a mechanical or electrical engineer if my alma mater rejected me. And that was more focused than most students.
It pretty much means what it says on the tin. Before hoisting an 18yo with the sudden responsibility of an adult, show them what lies beyond their GED. That's not just limited to college; there's many more roads out there and seeing what options motivate you can let you try to specialize while you're still in your prep phase.
That's precisely why some student go into a major undeclared or even take a gap year after graduating. No one gave them that exploratory transition time, so some make it for themselves.
People are choosing to go to university. It might have something to do with having better job options and statistically higher pay…
And do a final exam on oen and paper and nothing else, and an exam that relies more on thinking than rote memorization of formulas (or print the relevant parts of the course on the exam document itself).
This makes the class really awful. I had professors with schemes like this, maybe calling on someone's name randomly 1-2 times in class to answer a question they'd just asked, just to ensure everyone is scared into paying attention.
At least in this course students seem to be taking their strict 'no AI' policy quite seriously. Then again this is also the kind of course you do when you want to actually learn the material. It provides a qualification for further higher education in Swedish, so if we don't legitimately learn the stuff there's just no point as we won't manage in future courses.
If I knew most of my students were using LLMs I would encourage the rest of them to use them too. I would then change the marking criteria such that if they get a single point wrong that's 20% off the entire grade. 5 mistakes gets you a big fat zero.
LLMs are great but they make mistakes. At this point your students would have had to spend so long checking, re-checking and triple checking the LLM's output that they will have accidentally learn what they need to. They will likely need to cross-reference multiple LLMs and at least have read their output which is likely an upgrade on today.
Or some variation of the above, I'd be interested to hear your thoughts.
Reviewing is just not the same as working through a problem yourself. You often can take shortcuts when checking answers for correctness, which generation (with mind or LLM) of the solution can't take. This isn't limited to Math but equally true for programming and many other skills.
The people that do their own work would be at a huge advantage as the LLM(s) can work as reviewers instead of doing the work so there is more chance of catching a mistake before you lose 20%.
On top of that there will be some issues where the LLM is just plain wrong. Being able to work out when that is the case is a hugely useful skill in 2026. It is also something the people who blindly feed their work into an LLM and print its answer without reading it will be unable to discover. Much to their own detriment.
Now this i think applies in lesser degree to the comment I was replying to. With maths, and some comp sci problems, you often can check an answer much faster than actually solving it, eg by using simple substitution for simple math problems. That is a useful skill, but a different one from solving the problem yourself. It employs different types of skills.
That's because we write much smaller quantities of code than we read. Not that LoC is a useful metric, but I'd be surprised if I write more than a few hundred lines a code a day in a legacy codebase. Most of the time is spent planning and deliberating over what to write, and perhaps revising code a few times. Meanwhile, understanding what a snippet of code does can easily spiral into having you read thousands of lines of code across several modules.
But that's for work. writing is a stronger learning tool than reading, and we needed to write tens of thousands of lines of code before we got into the door and started reading more than we wrote. It's still preferable for students to do the same and write as much for their homework as possible.
Explaining that paradigm shift from learning to working isn't trivial, and is often poorly done. But it's needed if we ever want people to frame these tools as productivity boosters, and not shortcuts through tedium. Learning by its nature involves some tedium.
Students will complain to the admins and waste more of your time.
1. Make the exam the only thing that determines the grade (or technically, proves that the student learned enough);
2. Give out test assignments students can do, but don't have to;
3. Offer students for the professor to grade their work if they want to get some feedback on how they're doing.
Would that lead to less time wasted by tutors grading AI generated answers? From where I stand, students should be allowed to prepare for exams any way they like, with or without their professor's help. If they managed to learn, they pass.
Of course, this a lot easier for some subjects than others. Subjects like Film Studies relied on exams much less, and assignments much more.
I'd rather a MUCH larger amount of professors and teachers come to class to find their classes deserted (by their own hand), and indeed, their own justification for existing gone. The world would be a lot better off. Gatekeepers are anti-human.
They were disliked by many other students, though, and some people really think university grades should be about things like diligence, compliance, time management, etc., rather than pure subject mastery.
Small nitpick: you teach math, why don't you use an automated grader?
I agree homework should be optional and your college is wrong. Even without AI, for some students it's easy to cheat and maybe for some students it's a waste of time. Maybe you can assign it a very low score, like 5% of the final grade, and/or award the grade based on attempting rather than getting the right answer.
LLM verifies output of another LLM. The circle is complete, and nobody has any doubts that this is just a theatre.
With manual grading, that one student who genuinely tries still has a chance to meaningfully engage.
If the proof isn't trivially translatable to lean, is it actually a good proof?
That student presumably performed poorly on the test. Your approach is working as intended here, isn't it?
Also, is the AI-generated exam guide all that different to a human-written guide they could have downloaded a decade ago?
I think this development should make all educational programs reconsider the flip-model: homework at school and video lessons at home.
Homework sucks. Learning should be at your own pace (play/pause/rewind).
In general university lectures are horrible (of course there are exceptions). The lecturures oftentimes are required to teach as a condition of research funds, so for many it's not a priority anyways. On top of that most couldn't care less if their teaching style is didactically sound at all.
For grade weights, I weight the homework the lowest of any grading category, with in-class quizzes next, and exams scores the highest.
In any case, even if the uni insist on homework, the teacher can simply mark everyone who hands in their homework 100%. It is impossible for the administration to police it, or if they do, they'd need to hire someone to mark the homework (which conveniently solves the teacher's issue).
Students who copy all solutions from LLMs maybe learn something, but it's like 5% compared to what someone can learn by doing the assignment without external help, or even by asking their peers. It's a bit disheartening, because I invest quite some time and thought in designing the assignments.
It was very different before LLMs.
But it seems like the temptation to get perfect solutions from an LLM is too big for many students nowadays, they lack the maturity to understand that using LLMs for basic assignments is counterproductive.
I find bright people who are self-motivated tend to do problems on their own, but this is not the modal student. This is like the top 10-15% of any class. The modal student is motivated by getting a degree so they can enter the working world, and sees coursework as an obstacle to that. They'll take any and every shortcut to get the highest grades with the least amount of effort. However, particularly with first and second year undergrads (but I've also seen the same behavior in grad students), they do not understand the purpose of pedagogy. That doing problems is where you learn the material.
So if you make homework optional, the 10-15% hold steady and do it but the bulk of the class will stop doing problems and tank their grades on the exams. They will not internalize the lesson after that first bad exam, because it is morally and practially cheaper for them to believe they had a bad day or that their professor (me) is just a terrible meanie who constructed an exam with problems they've "never seen before" and "had no connection to the lecture" (actual evaluation comments I've had when I've tried to the homework-optional method).
Homework still provide valuable practice for students who are truly committed to learning. An automated system could provide an initial assessment and feedback. However, students can request human feedback paired with an in-person meeting. Such requests would require a mandatory explanation of what feedback they want or what the issue is. Professors and TAs will therefore only read and answer such requests themselves and will not waste time on AI output. At the same time, the system will still support students who are genuinely interested in learning.
Students who use AI to do their homework are not interested in such feedback anyway, and they will also avoid in-person meetings because it will naturally reveal that they are unprepared and did not do the homework themselves, which is humiliating.
To make this system work, we should present the in-person meetings as collaborative 'working sessions' or group office hours. Encouraging students to sign up in pairs or small groups also reduces individual anxiety.
The main channels for grading remain exams and oral presentations.
These can be written by AI. Or do you mean something more like the defending of a thesis?
Myself as a student i take my class notes and the subject matter the teacher provides and have Ai create multiple choice quizzes. It's a much quicker way to learn tho not as quickly as AI glasses I mentioned.
I wonder what the financial cost of AI will be to teaching? Does teaching get more expensive? More things need to be supervised, more things need to be presentations or in person?
The problem is not LLMs, but your university. In my undergrad, almost all the math courses had no required HW. They'd assign it, but you wouldn't turn it in. You'd come to office hours for help on the HW, or ask in class (they'd often dedicate the first 10-15 minutes of each lecture to Q&A).
It was awesome for people like me. No time wasted on neatly presented HWs. I'd do it quickly, verify the answers, and study for the exams.
The flip side was there were many exams (you don't want your grade catered if you do poorly in one exam). A course like calculus could have 4-5 "midterms", and then the final. Often, they'd drop your lowest midterm score so you're allowed one bad day.
I can’t imagine as American how that would go down with most American students.
How about cutting out the open book so that the students who don't study and to their homework actually fail? When I was in school many years ago "open book" was only used for special ed students and others who couldn't carry a regular academic workload. Homework was assigned, but our grades were determined solely by testing our mastery of the material. If you could master the material without doing the homework or showing up for class on non-testing days, more power to you. College has become more like kindergarten with all of the hand-holding.
In the classes I took with open note exams where I did well, I barely ended up consulting my notes anyway. The students who didn't study could flip through the textbooks all they wanted, but they lost so much time doing it that they couldn't complete all of the problems on the exam. For the students who studied well, the textbook was way easier to jump around in because they knew it well and knew what they were looking for.
Am exam that is designed to be closed book/closed note where a specific student is allowed some amount of notes as a disability accommodation is just different from one that is designed to be open note/open book from the start. The latter, testing the same material, is generally longer or harder to make better use of the format.
This was too long ago for me to remember the details of but, just making an example here, we might have had a course on homogeneization of elliptic problems using a model PDE with a div(A grad u) term throughout, and then the exam was about a PDE with a curl(A curl u) and some other added terms.
Notes carry the method and theoretical tools but you then still need to apply them to a new problem.
So, if anything, our open book exams were harder and a better test of understanding and mastery. You couldn’t rote your way into a good grade.
We had a formula "cheat sheet" on every exam paper, but if you don't know when or why you're supposed to use a formula, it might as well not have been in the cheatsheet at all
Force people to go all in with “impossible” tasks and see how they do with the power of LLMs. Then spend some significant time teaching people _with_ that solution in place what was actually built and what would have been a better architecture? A … software archeology session if you will.
That’s what most of what people do would be anyway - building PoC fast and then figuring out how to scale them?
Perhaps (sarcastically) this is the new approach, because this is now the point where they’re forced to struggle and therefore actually learn. An exam every other day, let them ask the questions.
I know this is absolutely impossible because colleges and universities have headed in completely the opposite direction, but the solution: Literally just expel them. Just like they did when people cheated when I was in college. Don't make it a small punishment so that they turn it into a risk/reward assessment. Colleges are overpopulated anyway; just trim the fat of students who shouldn't be there. Do a follow up oral exam interview about their project and if they can't explain what someone like me spent 30+ hours building and can speak about for 4 hours if asked, then expel them. They can make your coffee at the nearby Starbucks, who cares.
If you don't do this then understand that you are telling your students not using AI that they will need to begin cheating like the others to be able to keep up with the changes. OP mentioned they massively increased the scope of their assignments, which means OP is forcing their students to use AI now.
Or rather, I do have ideas, but they do require substantial restructuring. E.g.:
1) Strict, in-person (no exceptions), electronics-free exams with written and oral portion.
2) Transition to guided independent learning: students are given written materials and recommended exercises to study. They are shown how to use AI assistants productively. Instead of lectures, there are scheduled discussions, where students can ask questions and listen to additional explanations from the lecturer. They can also seek feedback for completed exercises. However, none of this contributes to grades in any way.
With such an incentive realignment, I think it's possible that higher education could be salvaged. But... apart from general institutional inertia, there is also the political angle standing in the way. Everybody needs a degree in the "developed economy", and the above changes are basically the opposite of the nobody left behind policy that enables half the population to attain one...
Can't you just not grade it at all? If the student has done their homework, just consider it passed. And then don't count it at all for the final grade.
They don't even have to be a significant part of the rubric.
Just ensure the students know what they don't know.
But now you could have the students take a written exam every other week and use a LLM to grade it?
I'm pretty sure Opus 5.5 has good enough vision capabilities to auto grade with the right prompt.
I imagine this would work very well in the current period since there’s no reason to use LLMs for this homework. Of course even back then many students didn’t do it and got low grades, but you have to be responsible at some point.
I think the problem is the students also take ~4-5 subjects per semester, and each of them may be just as demanding. High school is even more spread out with 6-8 subjects per semester. There's only a certain number of hours during the day for study and homework. For many, math homework ends up being the most time-intensive. Also these exams determine their GPA which in turn plays in part in determining their standing for the next few decades of their lives.
Reading this has made me understand that the recruitment industry will soon have to adopt the default position that anyone who graduated after mid 2025 is likely misrepresenting the depth of their understanding; people who graduated before that are the equivalent of low-background steel, until our processes catch up.
Maybe there’s still room for me.
Don't do anything productive for the society but still get to live a decent life on benefits, job security etc combined with addictive social media have ruined most kids will to study.
And the schools are even worse. When I was in school, teachers encouraged learning things; whether for school or outside, with teacher or self help. Everything was fair game and supported. Debate was normal.
Nowadays the teachers in schools are like a cult. Kids are tuned to reject anything from outside; anything that contradicts teacher is a no go zone. Anything that's "too advanced" compared to what teacher is doing is forbidden. Kids are scared of learning by themselves because the teachers punish it. What do you think these kids will do when they grow up?
That strikes me as backward. The higher the expected monetary value of good grades, the more students will prioritise grades over deep learning.
This seems pretty inaccurate to me at just about every turn.
The fire under their asses is for grades and a degree because they've erroneously been told they need one or else they'll be flipping burgers.
It's such a cliche thing to say but cheating at school cheats yourself of the education you could be getting but, since that's not what most people are there for, they cheat. They are there because they've been told that they need a degree to make a decent living and the degree doesn't technically require learning.
The primary difference now is that the both costs of cheating and ability to verify cheating on the instructor side have decreased dramatically, in tandem. There's also been a kind of mass effect, a cultural change in the attitude of students, where the sheer prevalence of cheating among their peers has made them less likely to feel fear, shame, or guilt over cheating. Students in the past who were caught were likely to confess immediately. Students now are increasingly belligerent and more likely to deny cheating, not just because they believe the professor can't 100% prove it, but also because they believe being caught out for cheating when most of their peers are getting away with it is unfair.
If you’re forced to attain a pass rate, and can’t watch so many fail, then the certificate becomes worthless in the workplace. Or just offer paid retakes and watch AI make you money.
If you insist on persisting with coursework then a short but heavily-weighted oral part seems to be the only answer, something that will diminish teaching time as a downside, as the article and comments about the flipped classroom touched on.
Students could just use the AI to teach/tutor the material to them instead, but the institutions are clutching the pearls too tightly.
I don't think "learning the material by other means" is the exam being "gamed".
Please include a description of what you're linking to. People aren't likely to follow a link to an unknown YouTube video.
Why is that not desirable? As long as the exam does not allow ai usage during it, i don't see the problem.
I don't think it's shocking to realize that most 20 year olds (especially those growing up and seeing that success is not tied to intelligence nor hard work at all)are taking advantage of a convenient tool to cheat when given the chance.
As long as they learn.
Then there was the written exam at the end of each course, which constituted 100% of the grade. The only reason that got abolished was because it made too many students fail, and the board hates that. So that became a race to the bottom, and apparently, we can still go lower.
It's reaching levels where most uni diplomas don't matter any longer. The only thing that will then matter is your network. That's the end of social mobility.
The article gives examples of LLMs being wrong. An AI grading of an AI submission would presumably give a good mark, to a wrong answer.
I think I understand what you mean though, and I kind of agree. I just don't like the use of the word chaff, like something you would discard permanently.
I'm pretty sure I was "chaff" at some point (maybe multiple) in my life growing up.
"This is math, you can find all solutions to the course problems in old material since I've been teaching here for 20 years, but know this you're cheating yourself and at the end of the semester you're taking a three hour test on pen and paper so good luck"
I see zero reason why AI should impact teaching. You can grade exercises as information to the students as to how well they're doing, but simply have an exam at the end of the course. Needlessly to say like 60-70% flunked through Analysis I and Linear Algebra. But they're adults and they're there voluntarily, so why is this an issue
My hope is that this is a wake up call to teachers / professors to find a better way to evaluate students. To determine who has actually come away with a real understanding, vs those who were running around copying the assignments off of their peers anyway. These same people are now just skipping the peers and going straight to the all knowing oracle. It's the same as it ever was, degree mills.
The example questions given in the article do not assess wrote learning.
> It's been distilled to the point that the lowest of the lowest common denominators can pass.
Have pass rates been increasing? Unless I missed it, the article doesn't say so.
> forced to learn things they will never use
A computer science degree is not a coding academy, it's a basic grounding in an area of study. A computer science graduate should have some understanding of theoretical computer science and of computer architecture, say, even if they're unlikely to apply these topics directly in their careers.
> My hope is that this is a wake up call to teachers / professors to find a better way to evaluate students. To determine who has actually come away with a real understanding, vs those who were running around copying the assignments off of their peers anyway.
Academics are already aware of the importance of fair assessment, but it isn't easy, especially with LLMs in the mix.
> These same people are now just skipping the peers and going straight to the all knowing oracle. It's the same as it ever was, degree mills.
Students that cheat, and degree mills, are two different things.
Now that we have systems that can pass for students at this baseline level, educational systems are facing a reckoning and we can, hopefully, move past cookie-cutter education solutions and recognize the need for more hands-on, engaged education. Will it be an easy transition? No. Will all schools adapt successfully? Also, no.
But will we finally have a path away from standardized testing, middle-of-the-road education? I am hopefully optimistic. We'll rediscover the need for human growth and development in our schools. It's either that or we lose. I'll choose the optimistic route and join and support those who believe the same.
I'm not convinced children need "hands-on, engaged" education at the age of 10. At some point there's no shortcut to being socialized that patterns in numbers are interesting (which is one of the national goals of maths for 3rd graders in my country). 10 year olds won't find that stuff "engaging" on their own, they must be informed by authority figures that it's a thing worth paying attention to.
Too much of this "fuck the schools" mentality treats children as perfect little beings that must only be supported to grow into wonderful adults, when the reality is that socializing children is a brutal exercise in fitting them into norms.
https://pietersz.co.uk/2009/10/schools-maths-wron
> when the reality is that socializing children is a brutal exercise in fitting them into norms.
No, you really, really should not be brutal to little children.
You describe the homework as
> basically lots or practice in basic arithmetic
Which is absolutely what you need for these sequences to have any meaning. There is no "pure math" to have a game about, without the "pathetic" examples of Lucy's 5 apples and Tom's 2. If somebody hadn't (brutally) drilled the concept of numbers into her skull, the sequences of random symbols would not mean anything to her.
Engaging with your child, who has already been partially socialized, is not proof that this socialization wasn't necessary. For that, you would have to have locked your daughter in the basement for those 7 years, and then find that she was spontaneously able to do multiplication. That is not going to happen.
> No, you really, really should not be brutal to little children.
I hope you understand that _brutality_ as spiritual. It's the brutality of going against nature, of putting a monkey in a suit and locking it in an office for 8 hours a day to feed itself. You don't beat math into your children, but they sure as hell don't learn it from eating candy and watching Minecraft let's plays.
She was quite capable of expressing what she was not interested in. You are just refusing to take experiences that do not match your preconceptions.
There is a huge different between engagement and "being informed by an authority figure" which implies they have no choice.
> She loves you, and is reflecting your engagement back at you in hopes of getting more validation. Children do that a lot.
Read the blog post. She demanded more of things she enjoyed rather than what I wanted her to do (finishing her home work). That is not seeking validation.
> If somebody hadn't (brutally) drilled the concept of numbers into her skull, the sequences of random symbols would not mean anything to her.
No, nothing brutal needed to drill the concept. You can teach very gently and make it fun.
> Engaging with your child, who has already been partially socialized, is not proof that this socialization wasn't necessary. For that, you would have to have locked your daughter in the basement for those 7 years, and then find that she was spontaneously able to do multiplication
Strawman. I did not say socialisation was not necessary, I said it did not have to be brutal.
> I hope you understand that _brutality_ as spiritual. It's the brutality of going against nature, of putting a monkey in a suit and locking it in an office for 8 hours a day to feed itself.
Who gets locked in an office?
> but they sure as hell don't learn it from eating candy and watching Minecraft let's plays.
That would be neglect. no one things complete neglect is bad.
> I hope you understand that _brutality_ as spiritual. It's the brutality of going against nature
nature is a lot more brutal than modern life.
I am both validating your experience, and rejecting your analysis.
> Who gets locked in an office?
If you can't engage with that statement in more thoughtful way than that, we have no common ground for a discussion.
My analysis is based on facts and observation.
Yours is based on what exactly? It is contradicted by the observed facts.
You seem to think kids always do what their parents show they want them to do, which is detached from reality.
I learned math from flash cards my mother was using for my brother. He got M&Ms for getting them right, so of course I wanted to learn and join in. Many years later, I graduated with a degree in math.
So you absolutely can learn math from eating candy. I did.
I worry that the availability of this path will be extremely unevenly distributed, amplifying existing educational gaps along class, race, and language-proficiency lines.
That might be better than a forced average denominator, but the social effects of further amplifying those gaps are pretty severe and don’t take that long to happen. Combine that with the fact that education isn’t like new miracle drugs, where improvements inevitably tend to become more and more widely available to different groups, and I’m not as optimistic about the future.
The real question for education is whether we can get 'tutor-quality' education from ai interactions. One on one education is leagues better than classroom education; it's the one-true-result in education research. There's a massive opportunity here to improve outcomes, and in a generally available way. You can't clone a good teacher, but you can give everyone an ai tutor.
Of course the onus is on me to do this at all, I don’t need to in order to pass the class. I set my own standard for what it means to ACTUALLY understand anything rather than just say “oh yeah that makes sense” and move on.
- 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 replays how a Google Doc was written, highlights unusual edits (large pastes) and tracks how that text changed afterwards. Surprisingly, several professors reported that students often underestimate how much LLM-generated text is left in their final drafts, so just seeing the replay changes the conversation.
It's not meant to be a verdict machine like the existing AI detectors (another big problem in education), more of a starting point for the kind of conversations the author's TAs are having. In fact, we're also beta testing a feature that suggests a few questions for the professor to ask based on the replay.
And yes, it's not perfect and autotypers do exist, but they're dumber than you'd expect and nowhere close to mimicking a real human (I hope to address that soon).
Interestingly, this applies not just for learning, but also to a great extent in production/work contexts too.
Isn't the answer the 'flipped classroom' - students are assigned reading/studying to do before a class and the class becomes more interactive, answering questions/discussing topics/solving problems etc depending on subject. Of course, this is more expensive than a hall with 400 people listening to a lecture with a couple of tests/essays.
I've heard of places where the minimum points for an assignment is 50% even if you never turn it in which is insanity.
My main point is: flipped classroom works great in some contexts but is no panacea and requires setting clear expectations.
Aside: as English is not my native language: "you fail them" can mean both "you give them a failing grade" and "you fail to provide the support they need", right?
Yes, but as written the latter would not be a reasonable parse. For that sense: “you have failed them”, sure; “you are failing them”, sure; even “in acting so, you fail them”; but “and then you fail them” as an entire sentence would be very weird. The emphasis I placed on “fail” also doesn’t feel particularly compatible with that sense of the word.
I also felt having multiple explanations (in a regular class where you could also read the book in advance) from regular lecturers helped me in gaining understanding because having something explained in multiple ways made sure there would be at least one way, or a combination, which made it 'click'.
I always liked the regular lectures where you would be required to do some pre class reading and a smaller set of exercises the most, as you would be engaged in the class itself and would be getting repetition with different explanations.
It’s a virtuous circle though, as many universities are there for the money, so it’s a beautiful exchange.
Flipped classrooms have incredibly weak causal evidence. However they are popular in spite of this - perhaps because of this if you're cynical - as the educational research field is pretty questionable in its standards and the profit motive is a strong thing lurking in the background. Wikipedia has a big warning on the article for a reason, but even then it has serious issues.
"Active learning" is a thing (insofar as it has meaning rather than be a useless buzzword catch-all) that does work. But the "flip" specifically does not, or rather, sometimes a mix of extra effort on behalf of the teacher and other loosely related effects occasionally can make a modest impact that makes up for its failings.
Most of the people pushing flipped classrooms are the same people who think that students learn best when they "discover" concepts on their own, maybe with a bit of guidance. This is the other, more serious, lie in education right now, akin to the whole idea that you didn't need to teach phonics to kids when they learn to read.
The end goal of education is to help you understand how things work, with problem-solving being a means of developing and applying that understanding. LLMs shortcut the former and turn the latter into a pointless arms race that we cannot win.
I loved the example of increasing the complexity of learning projects, but I still feel like one question is missing: are people really trying to learn the same thing we were trying to teach them in the first place?
* Alternate between theory and practice in the course of a single 2-hours slot. Theory is necessary, but the attention span of people is very limited. * Grade either a homework assignment in the form of a real project with specifications. You give the assignment half-semester, so that students who start early can ask questions. Otherwise, grade a in-session exam, with every resource available, including the course and the whole Internet. However, you don't assess the knowledge, but if the student is able to apply their knowledge to the different tasks at hand.
These approaches are now completely moot in the age of AI.
I was surprised the first time a colleague asked me to do an oral exam. I thought it was a burden on people with poor social interaction skills. It went surprisingly well: you could see in a couple of minutes if the student had integrated the concepts of the course, even if they were shy. Now I wonder if it's the way to go. There's one caveat, though: it doesn't scale.
There’s no way I’d have learnt everything I know now about software in today’s environment. So much of software development - debugging techniques, architectural decisions, structuring data etc is learned through trial and error. As the models get better is becoming easier and easier to just not look too closely at their output. I suspect it’s human nature.
In my very limited experience (I have German major degrees in Informatik and Philosophy, I was a TA for Logic in Philosophy) the more advanced the topic the thinner the attendance anyway.
I mean some people are good at Tetris and some aren’t. Tetris can tell you which ones are which.
They've shown that it is entirely possible to scale a system like that, as long as the society sincerely values the role of an educator.
Finally, if a student feels their result is unjustified, reevaluate the recorded session.
Usually you know after a few minutes if it's going to be a fail, and then otherwise you only need to figure out where on the passing scale the student ends up.
That's the point. As a professor you're allowed to outsource the grading of written exams to you assistants (ongoing PhD teaching stuff). But you must do an oral exam personally.
> usually quicker for the well prepared students
That's true. More time is needed only in case the answer is missing and you have to check whether you hit accidentally the only poorly understood material piece or there are more troubles.
2 hours is a lot! In my universities, classes were either 50 minutes, or 70 minutes. I could handle 50 minute ones just fine, but would often have trouble with challenging courses that had 70 minutes. Particularly for stuff like math, I hadn't properly digested the material in the first half, and now he's already proving theorems with that material in the 2nd half. I can follow the logic, but the ideal is when you can digest the prior material before being exposed to theorems relying on them.
Even when alternating theory/application like you say, the amount of attention I'd give to the application in the 70 minute class was less than the 50 minute one. I could follow, but I wouldn't think critically.
Similarly, I did well in interviews where I could use the internet, but I didn't do well in interviews where I couldn't use the internet.
Everyone has a testing format they're good at. I'm curious about how this kind of thing is evaluated, and whether there are any papers related to this.
I think structuring a course to be taken using a frontier model subscription is outrageous. If you refine the assignments each semester to be only solvable by a frontier model, your students will always be exposed to the actual pricing of said models. It may be 20 USD/month for now (which is already too much to ask in a syllabus in my opinion), but what if prices increase; where do you draw the line? What's next -- make the students pay for cloud credits in DevOps courses?
I agree with the need to increase the weight of oral examinations / interviews with TAs though.
This is how we do PRs now.
The education systems were already hopeless before LLMs and now they got caught with their pants down. Of course some problems were unavoidable, but this overreliance on homework was always a problem in my eyes.
Why would AI undermine quizzes? I took an intro to Japanese class where we would have a quiz at the beginning of each class which I quite enjoyed. I can’t see how AI would change the utility of that.
This gives me/us a lot of angles to understand what progress has been made throughout the semester as well as to consider and react to individual differences between students. In my book, it’s also resistant to faking it convincingly throughout all modalities. However, scaling that approach to big entry level courses would require more money for teaching than any research group in Europe I’ve ever seen has available.
— watches, calculators, discrete technological devices. I've seen kids encode text as patterns of dashes on their desks, sharing the cipher solution with their desk mates in other periods.
The problem isn't skirting the effort. The problem is systemic. Why does the system reward taking shortcuts? I think it's because most areas are focused on short horizon tasks, min-maxing effort:output. It's not a problem with those technical fields. It's a problem with the current crop of the C-Suite across the entire occupational plane. They're short sighted.
If someone manages to learn without doing the homework, or finds some way to learn better using an LLM, fine. The exam will show that.
I frequently see this in remote interviews for senior developers. These are conversational interviews, not coding tests, but some candidates seem incapable of having a conversation without AI.
I can't imagine how these candidates would be able to meaningfully contribute to a whiteboard design discussion with real people.
For a previous client, we resorted to a simple in-person pen-and-paper screening questionnaire with 5 very basic questions. Many candidates declined to attend, and most of those that attended failed dismally.
I generally think this is okay because the students are getting more feedback, but some students are starting to call this out as hypocrisy. At Dartmouth, there's a debate running because the Provost was using LLMs to improve his writing. He's not in a class. He's not breaking any official rules. But somehow people are feeling like he's cheating.
E.g. if students learn better through gameplay - can assignments become like game levels?
Rather than building for verifiable outputs - build assignments that gauge process and help elicit what they have learnt not what they have produced.
The problem is that the student is in charge of the tutor and often tells them to just do their homework for them and not check whether they have learned anything.
LLM providers should formalize a tutoring mode. We should have prompts available to create this mode. This should be the default way students use LLMs. LLMs may not be good enough to do this now, but I think there is already enough raw intelligence and its a matter of training models to be good tutors, to be experts and specific subjects, and developing software systems for tutoring.
In this model there are few lectures. This is more similar to Montessori but with on demand tutoring. The teacher provides objectives and materials for the students and tutors. The tutoring process can be done during class for teachers to ensure it is working well. There can be larger blocks of office hours for students to do their tutoring during.
As AI gets more intelligent, it will also be capable of doing all the grading. This includes oral exams. Teachers need to proctor the exams in person.
Anyone will be able to tutor themselves on any subject. Proctored examinations to certify knowledge do not need to be very expensive. The value of the school environment will not be to transfer knowledge from a teacher but to have peers working on the same thing and teachers and an environment invested in training the next generation.
It was also interesting to see that everyone is now forced to use LLMs to do the assignments. I wonder if students are using any kind of subsidised access to LLMs here.
This means that education needs to have a high teacher to student ratio. This is not feasible in any society currently on this planet, because we simply cannot afford to have that many teachers per student.
Once you have a wishing well - i.e. you own an eduction in baking or dentistry or law or math or whatever - you can ask it anything and it will grant you your wish. If others want something from your well, they have to come and ask you. What you give them has your name attached to it. So, your reputation rides on the quality of what your wishing well produces.
Prior to AI, universities chose which wishing well they would specialize in building and offered to share that knowledge with students.
In the age of AI, the creation of personal wishing wells has become trivial but universities still think they are in the business of teaching people how to create wishing wells.
Ultimately, the only truth left is for owners of the wishing well to start offering the product of their wishing well to the market and attach their name to it. If there is a market for their product, great. If not, go make another wishing well.
Education used to be pets. Now, it is livestock. Universities are not willing to accept this shift in mindset.
But you're not doing that, you've just given up and allowed any use!
> How to use AI coding agents effectively is not a learning goal
Exactly
> violate evidence-based best pedagogy practices, and I made them anyway.
Of course you have, when have educators cared much about best practices
> We have since built infrastructure to autograde code and written reports of homework with LLMs (institutionally approved LLMs)
Further reducing the point of your own existence
> Debriefing after such a failure is a good learning opportunity to talk about automation bias and different forms of human oversight (and how this is really hard in practice).
It would be if there were any learning, but since the incentives are for more AI use and less learning, there will be no application of the supposed learning, so you'll continue to have those 80% fails
Why do you say so?
This is not at all my experience, as a recent teacher it strikes me how much the teaching methods are supported by science.
I recently did a course on teaching bachelor-level education, and had to study a couple of books. They all seemed to quote their sources and I got the impression that they're rooted in science.
I've also read news stories and stuff that support the idea that a lot of "education education" is of questionable value. This story is pretty old by now but has stuck with me: https://www.latimes.com/archives/la-xpm-2010-may-10-la-oe-zi... . A relevant quote:
> Of the 1,300 institutions awarding graduate degrees for teaching, Harvard’s director of teacher education told a 2009 conference, only about 100 prepared students for the classroom. The rest, she added, “could be shut down tomorrow.”
Some of these experiences/articles are about teaching K-12 schools rather than teaching college, but there would have to be a pretty huge difference between the two for me to think that college-level pedagogical theory was meaningfully better than that for younger ages, and I've seen nothing to give me that impression.
Now the question is, are these curriculi being rebuilt? Here in The Netherlands, yes.
Lest anyone think this is a brilliant idea: an LLM would likely be much better at doing the ~~RLHF~~ RLMF, so you haven't actually gained anything. But I feel like there may still be the kernel of a good idea somewhere in here.
Perhaps your assignment is to iteratively train the AI, which is pre-prompted to seek out every reasonably possible failure mode when it generates solutions. So the learning mechanism is to train an adversarial AI on the subject matter as a way of learning it yourself. This does not solve the problem of cheating with an LLM, it's an alternative pedagogical approach.
This is the way.
I think society is moving back toward nature at rapid pace. More adversarial in general, and advantage taking behavior is the way to survive if you don't have a solid community. Society will be culled to a much smaller percentage of people than are present today, and it probably won't be pretty. Smaller communities instead of larger communities, with self-sufficient systems in place and a lot of warfare...
Don't see how AI overlords controlling everything will work out. If no one can buy the food from the large producers, then they will dissolve too. Supply chain will fail at some point in the next couple years maybe? Diesel is already outpacing wages for truckers.
The people with all the money are positioning themselves to reap the assets from others to give themselves as much leverage as possible.
At this point I don't really see how it could go any other way.
I don't know. I'm trying to figure out what's going to happen, but it doesn't really seem like anything good is going to happen because the trend is wanton disregard for morality in exchange for stealing wealth.
Makes me really sad. I'm also on fixed income so I worry quite a lot about my future.
AI can talk the talk, but it can't walk the walk.
A simple "explain why this is the best solution" will make it shit the bed 10 out 10 times.
Only because you're obsessed with being able to assign grades and fail cheaters.
No obsession required.
The ones who don't want to learn will cheat themselves of an education. This is their own problem and we shouldn't punish the actual students to catch these others.
The situation reminds me of what my teachers used to tell me. "Learning how to look up things in a book will be a skill you'll need all your live and you won't carry a calculator with you at all times when you are grown up." Both assumptions were proven wrong before I was even out of school.
If the only remaining value of education is "to become a well rounded person" then why waste your time on it? It used to be you went to university because that was the only place to get knowledge from. Which is also why it was dominated by the rich and powerful. These days university feels a lot like paper money. Everyone just pretends it has a real value and because everyone agrees, that's where the value comes from.
It's little effort for a large benefit. But anything you learn past middle school in formal education is a moderate to large effort for a small to neglectable benefit. And you'll forget it anyway if it is uninteresting and not useful in your daily life.
Anything interesting or useful in your daily life, you'll pick up by yourself even without formal education.
Knowledge is not a tool for answering the absolutely minimum questions you encounter in daily life, but a way of seeing the world. Reading allows you to use language in ways you could not imagine with spoken word alone, just as math or physics or computer science enable new ways of thinking inaccessible without intense study.
Our brains can't hold infinite amounts of data and a lot of information will become inaccessible if not used regularly. This is also why most of us do not speak twelve different languages, despite this being a supremely useful skill that is relatively easy to pick up compared to semiconductor physics.
Clearly teaching got a lot better, since we now have a new teaching tool at our use, the LLMs. It's evaluating that got harder, because students can use the same tool to cheat in traditional forms of evaluation.
If "teaching the students" means the same as "evaluating the students" to you, maybe you shouldn't teach in the first place.
It’s like copy-paste Wikipedia presentations 2 decades ago. People used to learn things doing research for a project. Then suddenly it became copying off Wikipedia and people not even pre-reading what they copied into PPT.
Even devoid of all the economic and social pressures that grades impose students would compromise their learning experience. Ever regretted looking up the solution to a puzzle in a video game or sneaked a peak at the crossword puzzle solutions?
One of the hardest part of teaching is keeping students from self-sabotage. Evaluation helps with that.
Beyond that evaluation is of course also useful just as feedback.
The only people really having a crisis here are teachers who want to feel relevant.
One major problem: how do you get a job for which a degree is required, if you teach yourself via books/internet/AI? How do you get into graduate schools for which a degree is required? How do you envision advanced degrees working without funding or labs or mentors?
All of education is undergoing dramatic seismic shifts because of AI right now. Attempting to insult teachers for talking about it doesn’t reflect on teachers, it reflects on you.
The kids are fine; the rent seekers aren't.
And to your larger point: the entire university/graduate/lab/scientific system was created to formalize what was originally citizen scientists trying to understand the world from their basements. Scientific journals are just a formalization of nerds writing letters to each other about what they figured out. Consider from first principles how much of what exists now actually still serves that original purpose.
You want it to be a jobs program? Sounds good. Take their computers away and make them use pencils and you can create even more jobs.
You want to do science? Figure out how to do science. You want the kids to learn? Figure out how to teach them.
Just don't protect a jobs program under the guise of something nobler.
Figuring out how to teach kids is exactly what this article is doing. How do you propose to teach kids, when they refuse to put in the work, and reach for AI for everything? Your suggestion to make them use pencils seems naïve/ignorant/obtuse; this article is written by a computer science teacher. You can’t force people to use only pencils for homework (as the article points out), and it would be ridiculous to use pencils for writing computer programs in 2026. That suggestion is a non-starter. FWIW, the impression I’m getting is not that you have any helpful suggestions for education, but that you don’t have experience teaching and don’t understand what’s happening in schools right now.
Who is protecting jobs programs under the guise of something nobler? This seems like a straw man unrelated to your previous comment. I’m getting the feeling you’re anti-education for some reason, maybe have some bone to pick with the university system, is that accurate? Asking kids to teach themselves or forcing them to use pencils are not going to be successful ways to improve education.
And you're taking the pencil comment weirdly literally, like it apparently went completely over your head. We're clearly talking past each other. I don't see any purpose to talking about this further.
What heavy assumptions are you referring to? If you mean assuming that kids probably won’t skip university and instead will be self-motivated enough to use AI non-stop for several years in order to teach themselves enough to get a job as a biologist or a ballet dancer or a civil engineer, and if you mean assuming that companies and organizations probably won’t suddenly start accepting self-AI-taught candidates as being the same as having a degree… if these are the assumptions you’re referring to, then you’re right - I am assuming those things and I believe them. No matter how AI changes things, it’s a pretty heavy assumption you’ve made to think kids will self-motivate an education just because information is available, especially given that was always possible and still chosen so rarely that it effectively doesn’t exist. And it’s a pretty heavy assumption to think that universities and organizations will change together in a way that can vette the merit of prospective candidates any time in foreseeable future. AI will change a lot of things, but you’re underestimating our social structures and what things like the accreditation system do for the workforce and the economy. You can argue it’s got problems, but assuming like you did that AI can replace it anytime soon is a very heavy assumption.
If the pencil comment and suggestion to take away computers wasn’t literal, then you’re right, I have no idea what you meant. Feel free to explain. If you can’t, then I’d agree, there’s no further purpose.
> Students can and will learn better with ai “professors” than they do now.
No, that’s not true. You have no evidence to suggest that.
> The problem is that higher education is not about learning ultimately.
It is, actually, and many institutions do a good job of this. Some don’t, but they’re not all just degree mills.