Create Educational AI Tutors with GPT: How We Built Ada, Newton, and Grace
Three custom GPT tutors, a calculus bot named Ada, a physics bot named Newton, and an algorithms bot named Grace, show how a single prompt file can become a real classroom teaching tool.
Funding anything new inside a university can take a year, even for something as small as a classroom tool. Custom GPTs sidestep that entirely: if one person has a GPT Plus account, they can build a chatbot tutor that serves hundreds of students in a class, at essentially no institutional cost or approval process. Nik Bear Brown, Professor Bear, walks through three of these tutors that are already running in real Northeastern University courses.
Three tutors, three subjects
Out of the hundreds of custom GPTs built for various courses, three get the spotlight here. Ada is an introductory calculus bot used by Abby Williams in her class, providing step-by-step math guidance. Newton is an introductory physics bot used at Northeastern for intro physics classes taught by Shri and Tom Kelly, with dynamic visualizations built into its responses. Grace is an algorithms tutor built for an algorithms course, and its impact was visible in a very concrete way: after students started using Grace, office hours attendance dropped off almost entirely. A couple of people came the first week, and after that, students just used the chatbot to get their questions answered, a pattern visible both anecdotally and in the bot's usage statistics.
How simple the build actually is
The technical bar for building one of these is low. It's fundamentally a matter of writing prompts, and the creators made the exact prompts for all three bots available for anyone to look at and reuse. Ada and the others took roughly half an hour to an hour to build initially. Grace, for instance, was written by Nik Bear Brown together with a collaborator, taking about an hour to get the bot working. Since then, instructors like Abby Williams have taken over refining and maintaining the prompts for their own bots, tuning tone and content to match how they actually want to teach.
Real feedback, real usage
These bots aren't just built and forgotten. Surveys run on all three tutors show they average around 4.8 out of 5 with students. One story sticks out: a student who was afraid to ask questions in class because she felt she'd look unprepared in front of Abby directly turned instead to the chatbot, where that hesitation wasn't a barrier. That's a specific kind of value a bot can offer that office hours sometimes can't, a judgment-free space to ask a question that feels too basic to ask a person.
Scaffolding, not answers
A specific design choice runs through all three tutors: they're built to scaffold learning rather than hand over answers. What every teacher actually wants from a tool like this is for it to explain why an answer is right when it's right, and why it's wrong when it's wrong, not just deliver a correct result. This gets demonstrated directly with Ada: a student can upload a photograph of a handwritten calculus problem, since students often work out equations on paper first, especially when the notation is hard to type or format correctly, and Ada will check the work, pointing out exactly what was done correctly and where a specific, often small, error crept in.
Tone matters as much as content
Beyond correctness, tone was treated as a deliberate design element, not an afterthought. Abby specifically wanted Ada to be encouraging even when a student gets something wrong, which shows up clearly in how the bot responds: acknowledging what a student did right, being specific about what's incorrect, and being clear that a mistake doesn't erase the parts that were done well. That tone reportedly carries Abby's own personality into the GPT itself, encouraging and friendly even in the middle of correcting an error.
Key takeaways
- Custom GPTs let a single instructor with a GPT Plus account build a course-specific tutor for hundreds of students, without institutional funding delays.
- Ada (calculus), Newton (physics), and Grace (algorithms) are all in active use in real Northeastern University courses.
- Grace visibly reduced office hours attendance, since students got their questions answered through the bot instead.
- The bots are designed to scaffold understanding, explaining why an answer is right or wrong, rather than just outputting a correct answer.
- Student survey ratings for all three tutors average around 4.8 out of 5, and instructors have taken over prompt-tuning their own bots' tone.
Try it yourself
The full prompts and code behind Ada, Newton, and Grace are shared publicly for anyone who wants to build a similar tutor for their own course, work that grew out of Nik Bear Brown's collaboration with Northeastern University faculty and his nonprofit, Humanitarians AI.
Full transcript(auto-generated, with timestamps)
[0:01]Professor Bear here. Show and tell. Uh we're going to discuss some open AI custom GPT chat bots that we've created for some courses. We've created hundreds of custom GPTs. They're super easy to create and they're really effective. What's beautiful about the custom GPTs is if you've ever worked in an educational institution, getting the money to fund anything is virtually impossible. It can take you a year to just get a little bit of money to fund something. But with the uh custom GPTs, all you have to do is create them. If one person has a uh a GPT plus account, they can create a custom GPT for hundreds of students in
[0:49]The class. So it takes the custom GPS are super s simple to make. Basically, you give it prompts and for all of these GPTs here, uh we put all of the actual exact prompts that we use. So these are the prompts using ADA. So, this is Ada, an introductory calculus bot that's used by Abby Williams in her class. All of these bots I'm going to talk about, I'm just going to talk about three. We have hundreds of them. I'm going to talk about um N, which is a physics bot, intro introductory physics bot that's used at Nor Eastern for the introductory physics classes with Shri and Tom Kelly.
[1:29]Um, Gradius, which is for my uh Corsair algorithms course. after using grace in my Corroser algorithms course I just don't have people come to office hours anymore everything gets answered through the bot I mean the first week maybe a couple people come but after that people just using the the chatbot to answer everything and we can see that from the usage statistics and from you know the fact that people stop coming to office hours and ADA is for introductory calculus so here are the prompts super easy it takes about half an hour to make these We have some data uploaded these Abby has taken over writing this bot. So
[2:10]Basically we wrote this uh Dev Shaw and I wrote this initially and then uh it took maybe an hour to write the bot. Students love it. Uh we've put some uh surveys on these. They average around 48 out of five for all three bots. And what's nice about it is Abby told me that one of the students came up to her who had math here. She was afraid to ask Abby questions because she would feel stupid, but she uses the bot. Okay, so let's try Atta with a handwritten problem. I'm told by Abby that often in calculus people write equations on pieces of paper and then upload, you
[2:51]Know, what they take from their camera. Uh because math is sometimes hard to write in a proper way. maybe don't know latte or other things. So, let's go ahead and try that. So, we're going to upload just a little handwritten um so this may or may not be right and we're going to ask ADA to check it. What every teacher wants is they want scaffolding. They don't want the answer to be spit out. They they want if it's right, why is it right? If it's wrong, why it's wrong. And what's also beautiful about this is uh tone. Nabi's very friendly and so she she took over prompting it. She wants it
[3:35]To be encouraging even when you get things wrong. What you did right, what you did wrong. So this this has this, you know, this is not correct, but you got this part right. So good job on getting this part right, but you made a sort of a trivial little error here and it would be correct only if you made this change. This is very much happening. is very much the tone of yeah you got it partly not you are wrong but you're sort of right but you know you you made a mistake here and so not only is it very useful these tools but it
[4:06]Even has the sort of a tone of Abby in the GPT where it's encouraging and friendly even when you make a mistake so that's it for this and we'll go on to the next show and So like, share, subscribe, and I'll see you in the next show and tell.
More videos
2:08Bridging the Pixel Gap in Browser Automation.
2:23How One Narrow Safety Rule Can Make an AI Less Safe Everywhere Else.
2:04Why splitting a chunk from its document makes it retrieve for the wrong question
4:20Three You Can Take Back. One You Can't.
2:21Why a 50-turn agent pays for the same screenshot 35 times unless it caches the pixels
1:53