Peter Gault: Writing as a superpower

Peter Gault. Nice to see you. Great to see you as well.

When did we first meet? You were you were a Fast Forward alumni, right? Yep, I I was in in the 2015 cohort.

And were you 2018, 2017? 2020. 2020.

Oh, wow. It all blends together, but yeah, I remember seeing your pitch when you were first getting started and it’s amazing to see the progress you’ve made over 5 years. Yeah, I remember Quill was one of the nonprofits that they you know, kind of paraded around as like, "This is what you’re trying to accomplish.

This is what high quality and success looks like." And it’s funny because I don’t know if you feel this way, but if I look back on, you know, where we were back in 2020 and where I am today, I would have I felt like, "Wow, I I would have been you know, enthralled to know that we were able to grow to the extent that we did." And yet now I find myself you know, quite anxious about continuing to grow and I’m sure you find you’re in the sort of the same boat where you’re you never really get to sort of like rest on your laurels. There’s always more to do, more work to do, more funds to be raised, teams to be hired. If anything, it just gets more complex.

It feels like a board game you’re playing, but every round is more possible moves, more options, more things to consider. And it just keeps getting more and more complex. It’s exciting though that we’ve both started with nothing and then both these organizations that are now having a national impact and and reaching thousands and tens of thousands and hundreds of thousands of kids.

It’s a really incredible. Yeah, well, millions in your case. Yeah, Peter, why don’t you just give our audience tell them a little bit about quill.org and and what you do.

For sure. We are a nonprofit that helps students improve their writing skills. We really see writing as a superpower that as a young person, when you can write out your own ideas, when you can explain your thinking, build an argument, use evidence to support it, that’s an incredibly valuable and import- important skill for kids.

But, it’s so hard to become a good writer. You need so much practice and feedback. And that with AI, there’s this incredible opportunity to give kids immediate feedback and coaching on the writing, and give them those practice activities that help them build those skills.

For us, what’s really critical is doing that in the context of the really important courses that kids are engaging with today. And so, that’s things like what’s happening in English classrooms, and what’s happening in history classrooms, but also things like AI, and how can we teach kids AI through writing? And that’s just this incredibly important topic that is new, and everyone’s grappling with, and trying to figure out what do we all need to know 10 years from now?

Is just this really difficult question, but one where it’s absolutely critical that young people really we get this right for them, or at least open the door to what’s happening in the world. And Quill, we can pull up maybe some screenshots or some just some visuals in the video, but you know, what what you built is it’s it’s this student-facing product. Yeah, can you just describe sort of like what what happens when somebody like logs into Quill?

What what what sort of what does the product look like? Yep. So, when you go to Quill, first of all, everything is free for all students and teachers.

So, that’s absolutely critical to our mission, that you go to the Quill homepage, there’s a quick sign-in button. Teachers quickly import their students, they’ll use a service like Google or Clever to import their students, and then they have access to more than 1,000 activities that they can assign. And what we see is that we have this range of activities supporting students through grades 4 through 12, primarily in the context of ELA, but we are now creating new offerings for social studies and STEM classes.

And that in these activities, they’re about 10 to 15 minutes where kids are writing on different prompts, and then receiving feedback as they’re writing. And these are all very quick prompts. They’re one-sentence prompts where the kids are able to quickly build an idea and then get feedback to revise and strengthen it.

And the core of Quill’s in that feedback loop that students will write, they’ll be struggling with a concept and they’ll continually and patiently be given feedback. And with over like five rounds of feedback, the students are able to really build these really critical skills. So, and and and Quill it’s like bite-sized.

It’s like you’re not going in and replacing an English curriculum, you’re basically providing teachers with, you know, these opportunities sort of integrate technology in a way that gets students to sort of like in real time think critically, write in response to different prompts, build critical thinking skills, and also just build their writing skills and grammar. And it gets sort of the real-time sort of coaching and feedback. And this is even before generative AI.

You’re actually giving you had sort of like a system for feedback. Have you been using AI since? I mean I mean language models seem, you know, pretty well suited to a product like Quill.

Yes, we’ve been building our our own AI since 2018. Back when it was NLP and you would build these models, it was very basic back then. You could do things like grammar analysis.

People think about Grammarly, for example, which has been around for many years and they’ve had their own AI. We had had a similar process where we would use AI to analyze the students’ writing looking for certain patterns in the syntax. And then using that to trigger feedback and coaching.

And so with Quill, we never fix the writing for you. We always help you to build the skills so that you can learn it yourself. And so under the hood, our AI looks similar to Grammarly and some of those other platforms, but the student experience was the exact opposite where again Grammarly would just fix your writing for you, we wanted to help you build these really critical skills.

And so that was some of our initial AI work. We built our own algorithms, we trained this meta model that had tens of thousands of different training responses in it. And we built these datasets by hand.

And that was really time-intensive work. And really critical is getting these AI models to be highly accurate. And so around 2022, we had completed all of this work where we had these amazing models, spent six years building them.

They were working really well. And then the bombshell of large language models comes. And now we’re in this whole new era.

And the LLMs are amazing in many ways, but it’s a completely different technology from an approach of building your own fine-tuned model. And so over the last year, that’s meant for us essentially scrapping everything that we built over the last six years and rebuilding natively on generative AI, which allows us to do a lot of sophisticated things that we couldn’t do in the past. But it’s a comes with its own challenges and risks as well.

Yeah, I mean, what are some of those challenges? Yeah. So So one of the big problems that all of the players who are using generative AI to help students are grappling with now is that these underlying models are trained to be very helpful for students.

And so what that often looks like is that when a student is a struggling to produce the answer, the LLM says, "Oh, I see that you’re stuck. Have you thought about blank?" And then it will tell the student what to say because the LLM knows the answer and it wants to help the student. And telling the student the answer is the fastest way to help them.

And so what we have to do on our end is build guardrails so that the LLM isn’t giving away the answer and that the thinking is on the student. And for us, what that looks like is building a multi-step prompt. You have a sort of a chain of thought process where it’s first producing an answer and then checking, is this the right feedback for students?

Is the student doing the thinking? Or is this an example of feedback that is simply revealing the answer to the student? We didn’t have this problem with our old models.

That with our old models, we control the entire model ourselves, we control the output, and so we we didn’t face that risk, but with the LLM, it’s unpredictable, and you can build guardrails around it, but if you don’t have those guardrails in place, it can sometimes go off the book. And yet, you you switched wholesale to language models, and and that’s because you saw some some serious potential. I mean, like what what prompted that that pivot, you know, given that you had something that was already working quite well, like being used by millions of students, and these new models have, you know, obviously lots of limitations, and they hallucinate.

Why why did you make the decision to Yep. To go all in? Sort of like a couple of really thing big things for us that drove this decision.

The first is that these models are a lot smarter. And so, while they can be unpredictable, you get the benefit of a much more fine-grained analysis of the writing versus when we created our own models, it was sort of like putting the writing in categories. And the LLM just can understand things in a very nuanced way, depending on what model you’re using and how smart it is, but that unlocks for us the ability to go deeper.

A lot of our work is focused on writing, but it’s in service of critical thinking, that students are building an argument using evidence to support it, and in that process, we want to sort of be able to go deeper in thinking about like, what are other questions a student could be asking, how could they use more evidence to support this claim? What’s a really good thesis statement that summarizes this idea? These are all pedagogically research-validated strategies that we couldn’t previously build into Quill because they were too complicated, and the AI wasn’t sophisticated enough for these strategies.

And so, with generative AI, what’s so exciting is that some of these really advanced strategies, things that researchers have known for 50 years work really well, we can now build in a way that we couldn’t build 2 years ago. And so, for us, that’s really why we made the switch is to unlock sort of these deeper forms of learning that we previously couldn’t couldn’t address. Yeah, yeah.

I I I keep coming to quill.org as an exemplar for, you know, how AI can actually cuz there’s a lot of, you know, hyperbole around like AI is going to transform education and I think sometimes I, you know, you know, my advice to folks is that, you know, AI is a really powerful tool, but it’s really not the it’s not the solution. It’s sort of like part of it’s something that we can use on that journey and I think Quill is, you know, I don’t think of Quill as an AI tool. I think of Quill as a a reading and a writing tool.

Now, you’ve integrated AI in a way that really enhances the the sophistication and capabilities, but it’s what you didn’t do is scrap quill.org and start a whole new company. And, you know, I think a lot what a lot of the stuff that’s in the field right now, it’s like these wrappers. And I think some of the some of some people will say things like, well, you know, teachers will be able to just build these things for themselves and I and I think just like reflecting on like the amount of work that went into fine-tuning and even sort of like the prompt, you know, I don’t know if I’d call it prompt engineering, but you you’re basically thinking about like prompt design.

I mean, is this something that that like any teacher can just do for themselves or or do you feel like there is actually maybe a bit of barrier to entry to like really effectively using LLMs? Yeah, it really depends on the task. I think there are certain things that a teacher can do them themselves and can be really effective.

There are things like draft me emails to parents based on my students’ progress. And those emails could have taken you a couple hours to write and that could now be a 10-minute process. And so, I think there are some examples of ways in which AI can help to speed up communication workflows.

Some other examples, there there are a whole bunch of things that it can do well. We’re seeing some areas though where you’re stepping backwards a bit when it comes to education. So, in the world of education, the high the is high-quality instructional materials.

And there’s been this huge effort over the last 10 years to go from these not-so-great programs that were being provided by some of the big publishers to much better, smarter, more sophisticated curricula that could be used across an entire school from grades K through 12. EdReports led the charge here with how it rated curriculum programs. And the goal here was to go from incoherence, which is how education felt often, to a coherent program.

What was happening in parallel were some tech plays that didn’t plant pan out. So, you had sites like BetterLesson that tried to aggregate teacher lessons, and that they would host 100,000 different lesson plans. And for a teacher trying to create coherence out of 100,000 different lesson plans, impossible.

We then saw Teachers Pay Teachers, which took that idea and added a payment mechanism on top of that, but still didn’t have that coherence of K through 12 in a program that worked for the entire school. You still had individual teachers making individual decisions. And while sometimes that can work well, oftentimes it led to students not having a consistent learning experience.

And so, as we think about AI today, the next version of Teachers Pay Teachers is Magic School. And so, you have teachers generating lesson plans, and rather than say having a coherent learning experience, you’re sort of recreating some of that that incoherence. And so, I think that’s sort of the big question is how can AI fit into this bigger picture of coherent learning?

And that happens by starting with the curriculum. It starts with what is the teacher doing in the classroom, and then how can you layer the AI in in a very surgical and mindful way so that it’s supporting the classroom instruction, as opposed to trying to sort of move away from that altogether and propose something that lives completely outside of these existing programs. And so, I think that’s really where the smart players are today is they’re trying to work within the constraints of what really good programs look like and Quill’s certainly been successful by taking that approach.

Yeah, I sent you something over text. It’s a study that someone on my team shared. And this is from Microsoft Research.

Let me pull it up. Just so I can get the date right. This was published oh, this year.

So, very recent. I think it was like literally like a few weeks ago. And the the sort of headline is as they looked at the use of generative AI in knowledge work.

They found this correlation between knowledge workers that are increasingly reliant on generative AI actually have a deterioration of their critical thinking skills. And this is not, you know, it it it it’s not necessarily terribly surprising. This is something that people were I think postulating for some time that this would become a crutch.

It’s interesting that it’s starting to play out and actually be validated by research. And and and it it sort of brings me to this question of like, you know, there are some folks who are who are there there’s they see what ChatGPT and other language models are capable of. And their reaction is, well, we just need to teach kids how to use these tools because these tools are the future.

Like, we can’t you know, we can’t put Pandora back in the box. And so I’m curious what your response to that is. You know, is is it the case that we just need to sort of pivot and adjust educate the goal of education from teaching students how to write critically to teaching them how to let’s say become really effective prompt engineers.

Mhm. As someone who is employed, you know, prompt you know, I don’t know if you’d call the folks on your team prompt engineers, but they’ve been doing quite a lot of prompt engineering. Absolutely.

Yeah, I know we’ve created thousands of test prompts and that’s very much what they do. To me, I guess they they are there’s more overlap here that prompt engineering is writing that you are writing prompts and I do think the ability to write well does enable you to do things like query an LLM, and being able to specify what you’re looking for, and why you need that information. In so much of my own experience using LLMs, it’s not that first query that gets you the answer.

It’s that process of digging in and interrogating the answers. And so I do think that that is very similar to the process that English teachers are trying to create for their students. They’re reading a novel, they’re trying to unpack the novel, trying to interrogate it.

And so that critical thinking skill of trying to unpack information, I think that skill remains critical for students, and that writing is a way to build that that ability to interrogate information through building your own ideas. There is certainly a counterargument that as these models get smarter and smarter, there will be less need to even prompt engineer. And you’ve made this argument to me a few times, and I actually am starting to come around a little bit more to this than I was initially.

So I’ve been using the deep research function that’s in ChatGPT and a few of these other tools now. And what this does, which the the previous models didn’t do, is it asks you questions to understand what you’re looking for. And so rather than you needing to ask the questions yourselves to really unpack and get that information, the LLM can better like read your mind, so to speak, and figure out what it’s looking for.

And so I do still believe that writing and critical thinking remain absolutely critical skills, and that our ability to navigate this world and all of its complexity, if you don’t have these skills, I feel like you’re going to going to be left behind. That being said, in terms of AI as a skill, what that skill is, I think it’s very up in the air of what will AI look like a few years from now. It’s really hard to predict that.

The one thing I do know, though, and I do think this is a useful comparison, is I think about the area the area in which we all knew how to read maps. And you look at folks like taxi drivers and folks who had to drive for a living and they would build this knowledge of the world through driving where for a lot of us you know you would get that by reading maps and trying to have that spatial recognition. Today it’s a little bit of a lost art.

I love maps so I try to force myself to sometimes not use Google Maps just as like a fun exercise in thinking but if you’re not forced to figure that out day by day it does change how you think and so it’s certainly a big ethical question of what does thinking look like in a world where AI can can think itself in a really meaningful and deep way. Yeah that the maps example is fascinating because I’m I’m someone who who really struggles with sort of like geospatial reasoning and I find myself you know even with maps and we even with Google Maps sometimes getting lost. In including in New York City which is hilarious cuz New York City is one of the easiest cities to navigate and I actually find myself I’ve become much worse and increasingly dependent on Google Maps where you know it’s like New York is a great example like you shouldn’t need Google Maps you should be able to just like look at sort of the grid like roughly okay I’m on like 41st Street and you know it it I shouldn’t have to sort of open Google Maps and like do the thing where you sort of like are like turning around and trying to get a sense of like which direction you’re pointing and I don’t know that like my you know the deterioration of my ability to navigate the world is it’s definitely bad but in the the hierarchy of skills that maybe I’ve lost it doesn’t necessarily impact me day-to-day like I you know at the end of you know generally I have access to to my phone in many cases.

Critical thinking though like if we sort of abstract prompt engineering what you’ve described sort of like the process of writing which is really sort of like an underlying skill that leads to and maybe is like one of the necessary if not sufficient but certainly necessary conditions for critical thinking skills. It it feels much more serious if we’re in a world where there’s this reliance on AI to supplement critical thinking especially in a world where you know, one of the things that you described me even deep research like I’m actually I have a deep research query running right now on that study. I asked like, oh, can you find some more studies that maybe corroborate this or not?

You know, being able to effectively use it I have to actually, you know, critically evaluate the output and many cases it’s made stuff up or it’s not quite right. And so that’s not really a question. It’s more of just like sort of like an open concern that I think is sort of like underpins this the big question about AI in education is you know, do how do we balance the utility of these tools and the fact that like if you don’t know how to use the tools, you’re going to be left behind.

That’s that’s probably true. But if you use the tools and if maybe like you spend too much time using the tools, you actually have less you’re sort of less effective as a complement to the tools. Yeah, I’m not quite sure how how it’ll play out.

You know, I think we think of this this I like to think of it like an orchestra and that as the conductor of the orchestra, you are got a symphony of different queries that are doing different requests and that the human being is driving that and that I think the the orchestra conductor remains and that that skill of how you manipulate LLMs and find information, it that will remain a critical skill for sure. But I think there’ll be a difficult question of how do we build those skills and what does it writing even look like when writing is a co-creation process? And I do think that those are big questions where today when I write, I’m building a thesis.

I’m taking a point of view on the world and there’s a world where you can ask the LM like, "What should I think about this?" And you can imagine the LM giving you your opinions on things as opposed to having your own point of view. I’m somebody who has a lot of opinions. I love doing debate throughout high school and college.

I found that to be the experience that built my own critical thinking skills most effectively. And so I do think that for young people to really important that they have their own opinions and they have this ability to think critically. But I do think that to do that they need to spend a lot of time doing this and they need to be given those opportunities.

And if they don’t, there’s certainly a world you can imagine where that gets outsourced to the LMs in lieu of their own ideas. I think for me the the reason I feel so confident in my our advice to educators being, you know, you shouldn’t be focusing on teaching students to use AI tools, you should be focusing on critical thinking skills. It comes from this like, you know, even when we’ve done hiring and we thought about, okay, we certainly and we’ve actually struggled to hire people who are sort of like super users of AI.

I I just think like our generation is still like there’s some anxiety about like, is it even fair to you like, is it cheating to use AI to to to write something? And so there’s some hesitation. But if I think about like, you know, like we’re hiring someone to help as as we discussed to help with like fundraising and one of the things you really pushed on is like, you know, being having exceptional writing skills is critical to that.

I still feel like in my hierarchy of of skills that we’d be looking for for that role, I would way rather have someone that’s a really good writer. I feel like they I could teach somebody to use a language model if they have the writing skills. I don’t know if it’s the reverse.

Like if someone’s really good at chat chugging out a prompt, I feel like the minute that I need something that goes beyond what the LM is capable of, they’re going to sort sort hit that roadblock and that’s feels much harder to teach. Like I don’t know that you can teach someone to be a good writer on the job. It’s sort of something that you basically have either developed during your educational experience or not, and and some people maybe have a knack for it.

For some people, don’t. I mean, has have you had any something like how has your organization thought about sort of like talent, you know, given that you obviously have a lot of people that are using the AI almost on a daily basis? Like, do you do you have a sense of like what skills are really Yeah.

You know, set them up to be really effective prompt engineers or users of AI tools? So, I think both with prompt engineering and writing, we have a pretty specific definition of it, or at least I do, which is that writing is knowledge. And I say that to say, and this is definitely a hot take in the world of education, there’s been a big push called the knowledge matters campaign, which is the idea that your ability to write is predicated on your knowledge of the world.

And so, when you ask kids, for example, to write about baseball, and they’re big baseball super fans, you’ll get these long essays of baseball strategy and hits and great teams, and they’ll have their writing will look really great because it’s a topic that they have knowledge about. But then you ask them to write about a book that they’re not interested in, and the writing looks quite different. And in this world, writing isn’t just a skill of you can construct a sentence, writing is your knowledge of the world.

And so, I say that to say that when we think about like a fundraising writer role, for example, we find that the world of philanthropy is very fractured. There are many different causes and issues and strategies and theses of how the world can be improved, and as we’re working with different partners, our work needs to align to their work. And so, that really is requires knowledge of how do our partners think?

What do they think the future will look like? How does Quill align to that vision? And as we’re writing, it’s not just an artfully constructed sentence, it’s about how our mission align to their mission.

And so as we think about AI and and as we think about these critical thinking skills, students need to know a lot of stuff about the world. They need to know how it works. They need to know what these different ideas are to be able to be a member of those conversations and to be able to contribute their own ideas.

And so I do think as we think about how AI will develop, there’s sort of this underlying question where the more that you know, right, the more you can contribute. And for us, that’s really critical. And that when you have that knowledge, the writing will flow from there.

And I so I say that all to say that those skills look quite different. That the world of knowledge is often thought of something that like a history or social studies teacher does, but it doesn’t exist across all classes. And that’s also a shift that’s starting to happen right now.

There’s a big movement to to move towards knowledge as the main way in which students are assessed as opposed to things like reading comprehension as a skill, where being able to read any article on any topic is less valuable than knowing about particular topics. And that particular knowledge is more valuable than that general ability to read. And there are lots of proponents for both sides of this.

But I think we’re starting to see how knowledge itself is durable in a world where some of these skills become less important. Yeah, it’s like the the I don’t know if you’ve been following the all all the noise being made about vibe coding in the Y Combinator Mhm. Survey that came out.

And and for for our our listeners and viewers, this is a survey that Y Combinator did with their their latest cohort. And this is like the leading technical I mean like not These are some of the most technical founders right in the world. Like these folks know how to code.

And full I I think it was like a quarter of the cohort reported that 95% of their code is written by AI. Yeah. And so like right out of the gate, the headline was like, "Vibe coding is here, and if you’re not, you know, using AI tools, you’re at risk of being left behind." But then, when you sort of like, if you listen to the full podcast, and you hear sort of like the nuance, they also will say that, "Well, it’s also the case that, you know, vibe coding is great for getting you to an MVP.

It’s, you know, a great tool." But what they also have seen is that the most effective founders and companies had folks who had, I guess what they’re calling classical coding, which I find funny, but had classical coding skills, to be able to do things like debug, which AI isn’t very good at, at least right now. And And you actually have engineers on your team. I mean, are they are they vibe coding?

Like, have you like had a had conversations internally about this? I was just talking to our CTO, Aquil, about this. And he was sharing that at Quill, it’s around about like 10 to 20 hours of extra productivity per week.

And so, it’s not that we’re doing 95% of our coding through AI, but it’s certainly helping us with particular problems. It’s great at writing SQL queries. It’s really great for certain projects.

That number will just keep increasing though. I think as we build our own knowledge of how to use AI, and as these tools get more reliable, I think we’ll certainly see, you know, in a couple years, we’ve got six engineers at Quill, which is a small team, you know, relative to some of the big ed tech players. It’s not one engineer, like some small nonprofits, where like you only have one person.

But what it means is if they can have the output of say 12 or 18 engineers, that’s a huge win for us. And so, I do expect that number to increase, especially as these tools get better at debugging. And that still is a big obstacle right now, where they write code, sometimes it’s good code, sometimes it’s not so great code.

The ability for the AI to refactor itself though, for it to improve its own code writing skills, that will certainly happen. It’s already happening, but that will just keep getting better and better. And so, I think there is a big question where one there’s a lot of advice like what jobs should people be pursuing now and there’s this huge question mark around like is software engineering the path to like economic prosperity?

And I don’t know the answer to that question, but it seemed to be that was like hey, you’re trying to find sort of a path towards having a comfortable and and economically successful life, that was the path. And I don’t know if that will be the case in a couple of years and I don’t I don’t know what happens in that world. But I we’re certainly starting to see some of that.

I’m curious if you are hearing from others how that conversation is shifting now about things like what does CS education look like? Because again, you do have the ability to code, it does allow you to manipulate these systems and so that becomes incredibly valuable. But if the learning curve of building those skills is too high relative to that ability to just use an LLM, yeah, I don’t know what happens.

Yeah, I was just So, I’m going to I’m going to read a quote. This is from Tom Blomfield, he’s a partner at YC. He’s my my my husband’s former boss, founder of a company called Monzo, which is multiple billion-dollar valuation now, very wildly successful.

And he’s been experimenting with live coding and and here here’s his take. He says, "Software engineers are like highly paid farmers tending their crops by hand. We just invented the combine harvester.

The world is going to have a lot more food and a lot fewer farmers in very short order." Yeah. So, so my response to this is I think we have to take very seriously that the folks that are really at the bleeding edge of these technologies, who are really getting hands-on, that’s a pretty consistent take. I don’t know that I’ve talked to many engineers that are like totally complacent and saying that like AI is just, you know, is just a fad.

I think there’s actually folks who thought it was a fad that are coming around, but most of the folks in CS who are like, you know, actually building are like, yeah, coming to terms with the fact that this is this is going to make software engineers more efficient. And you know, you sort of just by extension, right? Like if you can get more productivity out of one engineer, you don’t necessarily need 20, maybe you only need, you know, 18 or 16 or whatever the number is.

And I yeah, I I I I I worry about I think I think because education has traditionally been very oriented towards these like super discrete career pathways. You know, my parents are immigrants and for them it’s like doctor, lawyer, maybe engineer. And that was basically it.

Everything that or bust. And like I went into political science and they were like aghast. They were like, well, you know, you can go to law school still.

And they they actually still ask me if like, "Are you Have you thought about going to law school?" and yeah, I I I think you described it really well. It’s like the reason that we hyper focus on those, you know, those pathways is you know, for someone who grew who who who grew up, you know, in a lower middle class household, you know, one of the most certain ways to achieve economic mobility was to go into one of these careers where you’re, you know, essentially going to be guaranteed, you know, like you know, six-figure income. So, yeah, I mean, if if we have at some point there’s probably going to be this shift where you know, if companies start laying off certain percentage of their engineers, and then now they’re flooding the the job market and then you have all these sort of graduates that are coming into the job market and and then you’re in high school and you’re trying to decide, "Okay, do I go into And I don’t think it’s just computer I think computer science is the canary in the coal mine just because it’s such I mean, you know, software is a language and so language models are especially adapt.

It’s like there’s there’s there’s a lot less friction to applying it into that specific type of knowledge work, but I don’t think, you know, accountants or financial analysts or, you know, lawyers. I I I think I think a lot of knowledge work is really, you know, you know, in the crosshairs and it’s just a question of like how long it takes for institutions to kind of figure out the implementation and if it get past the friction. And yeah, and bring me to the sort of like how are how are you going to add value?

Because being able to write lines of code is probably not sufficient anymore. You know, my my sense is that if you you said you have six engineers, let’s imagine you had 100. You know, and if you were thinking about, okay, well, I’m going to get rid of let’s say I can get rid of 10%.

You know, the question would then follow and maybe you can answer this for me. Who are the engineers that you keep? Who are the engineers that you get rid of?

Like do you have a sense of like what the complimentary skills beyond just like knowing the the software language that, you know, are going to be important? Well, my own intuition is that folks who have jobs now I don’t know what big corporations will do, but I think in general those folks will be able to find your jobs, but I do worry a lot about those new graduates. I think when we think about hiring, we’ve hired folks who are brand new software engineers and we always know that there is a bit of an investment up front that sort of coming out of a boot camp, you’re not quite ready to be a full contributor to an organization, but that over 6 to 12 months you do become a contributor and that as LLMs get stronger, that ability that that entry point I think is going to be what’s most at risk.

And so that’s certainly something I’m really worried about. I do It’s so funny and my parents also pushed me to become a lawyer as well and that feels the most at risk of any of these professions. When you look at the deep research tool which I love and I keep running out of my 10 credits per month and I need to upgrade now to the $200 plan because it’s so useful, but it’s a complete game-changer in its ability to take a week’s worth of research and do do in 10 to 15 minutes.

And so we see this future where the LLM can do these sophisticated research tasks in a way that would just take a lot of time to do ourselves. I spent so much time plumbing the depths of Google searches on page 50 or or whatever to try to find some information. And now that that can be done automatically very quickly, that really changes the needs of the workplace.

And so I don’t quite know what the answer is here, but I do think that those folks who have jobs will probably be in a better position. But for those folks who trying to enter the workforce, yeah, that that there’s a really big question mark of what do these entry points look like now? Yeah, and you have I don’t know if this is an announcement or a leak, but there was, you know, the news that OpenAI is going to be offering like a $20,000 AI agent.

I you know, I have to you have to take sort of all this through, I think, a lens, right? Which is that these companies are also trying to command big valuations there at, you know, beholding to their investors and trying to justify and cuz they need to also raise tons of capital to train like their the next-gen models. So I’m I’m like less, you know, I I’ve I’ve there are there are folks on who are saying like we’re 1 to 2 years away from AI being better than any human engineer.

There are also also folks I I think Amodei at Anthropic has said we’re maybe 1 to 2 years away from like artificial general intelligence. Whatever that means is a whole you can spend 90 minutes just talking about that. But I think even if we’re like much more conservative, even if we say, "Okay, you know, 5x that." It still doesn’t really put any of this outside of the realm of like, you know, if you’re in like middle school, like we’re still talking about like your first job out of college or maybe even while you’re in college, you know, us getting to this moment.

And and just to like prove the point, so the ChatGPT deep research just finished its work. And so as I said I I I I gave the initial initial research research PDF from Microsoft Research, it summarized it. And then I I I didn’t even like spend that much time prompting.

I just said, "Okay, could you conduct some research to identify other sources that address this question?" And as you said, it kind of like you know ChatGPT before would have just immediately started generating something. In this case it stopped and actually asked for like which direction do you want me to go? And I didn’t even, you know And I think this is like for me this is what concerns me is that you know, prompt engineering doesn’t have to be terribly thoughtful.

I mean you can sort of just like kind of like hack your way through it and you still will stumble into sometimes, you know, high-quality outputs. I haven’t had time to actually read through this, but you know, it pulled some legitimate sources. I It found the Microsoft Research report.

It pulled something from you know, Springer Open. It sort of outlined it, talked about sort of the benefits and opportunities. And then it did this clever thing.

It’s like very in-depth. It did this it did this table. And it’s like, "Okay, here’s a table that summarizes the key benefits and risks." And so it basically did this like you know, you know, very I mean I was going to actually ask like, "Oh, this is really dense.

Can you simplify it?" And it And then I got to the point where it already actually done that. And you know, I I I I guess it brings back to like what’s the point of education? It still feels like and I think this is a good news for teachers.

And maybe tell me what you think about this. It’s like education doesn’t Like like the good news is education actually doesn’t need to change that much. Like to your question about computer science, like we still need to teach students computer science because their ability to be really effective vibe coders, if that’s what we’re going to call it, will actually be predicated on whether they have the ability to evaluate and critically analyze the outputs and debug and also the computational thinking skills and, and and those students that have that knowledge are going to be by far the best able to add more value alongside AI.

And so, and then the other good news is that I don’t know that education really is going to have to solve the problem of teaching students to use these AI tools. I think the tools, it’s like we didn’t have we didn’t have education didn’t have to solve the problem of teaching students to use their phones. And I think in large part, you know, the same goes for like the internet and social media.

Like these these are technologies that kind of by design be become sort of like ubiquitous and and seamless. So, I I agree with you, but I have a couple of hot takes here and I have a few ways in which I think there are some pressing questions. So, so one of those I think is when you look at education, I completely agree with you that it remains vital and more vital than than ever.

That if an LLM can produce a 10-page report for you, your ability to read and understand and build your own knowledge from that report is critical. That that the LLM can’t just download the information into your brain. We’re still hopefully many years away from that scary reality.

But this is all to say that that sort of becomes we we’ll live in a world of more information, not less, right? That’s certainly something that we know is true that information is not going to become more scarce. And so, your ability to parse it will become more critical.

There are more immediate challenges though of like what does education look like today? You know, you’re talking about middle school students and it’s wild to think that a 10-year-old a decade from now, you know, this question of is it AGI a year away or two years away? It doesn’t really matter.

It’s going to come whenever it comes and there’s no one definition of it. It’s, you know, lots of definitions. But 10 years from now we know the world’s going to look quite different than it does today.

And we know it’s going to look quite the world today looks quite different than the world 10 years 10 years ago did. And so we’ve been worried about driverless cars, for example, and that is another huge impact on society. And that’s taken a lot longer than people have expected to become this ubiquitous thing.

But, you know, my friend was just in LA and in a driverless car and it was driving him around the city, right? So, it’s like these things are real now. And so I do think if we take a 10-year horizon, I do think there are a couple of really critical things that we do need to do now.

One of those is when you look at ELA instruction, one of the big changes over the last 10 years has been towards less fiction writing and more nonfiction writing. So, when you looked at English education, one of the big heavyweights in the space was a woman Lucy Calkins and she ruled the roost when it came to literacy and she really loved fiction writing, getting kids to write stories about their lives, and it was a really fun and engaging experience for kids, but it didn’t build the critical thinking skills in the way that a nonfiction text does, where you have to build an argument, you have to find sources and evidence to support it. And so when you looked at classroom instruction, about 90% of writing was fiction writing and about 10% was nonfiction.

And there’s been a big push to shift so that maybe 70% is nonfiction and 30% is fiction. I love fiction writing, it’s not that it should go away, but kids need to be given these opportunities today. And so if you really look at what’s happening in classrooms, how much nonfiction writing is happening, so we’re directly connected to how quickly and how effectively we’re building critical thinking skills.

So that’s one really critical question. A second though is like what should computer science education look like? And should we keep teaching JavaScript to students, for example, which has become the mainstay?

And I know I’m sort of stepping my foot into a very hot water here, but I’m I expect that in a couple of years, like we won’t teach JavaScript as the primary language that kids engage with as their CS classroom. And I don’t think CS classrooms will go away. I think they’ll become more vital, but that classroom will look quite different.

And I do think that having coding skills is important, but if there’s not an economic driver of that skill, if JavaScript isn’t a language that we use anymore because Figma your Figma designs turn into front-end code automatically, what is the role of JavaScript in that world becomes a a big question mark. So, I do think those are some of the questions that aren’t happening today, but will happen within the next 2 years. Yeah, I mean I think that’s like the the right aperture.

I mean it’s it’s not that, you know, do I think we need to continue teaching computer science? Absolutely, yes. I also think that the have and the have not like we everybody worries about this digital divide and like, oh, you know, the kids that don’t have access to AI are going to be left behind and I I actually worry that the digital divide will look more like the kids that are over-reliant on AI will be left behind and the kids that toiled through learning JavaScript, Yeah.

Even if they don’t know even if they’re not using JavaScript specifically, they have gone I mean like, you know, I’m not an engineer, but my sense is from all the engineers that I’ve talked to is, you know, you struggle through learning your first software language, your second software language and then at a certain point you kind of develop the instincts for being able to like learn new languages. But the you can’t skip that process. It’s like there is something sort of like this that the productive that you struggle that comes is like, you know, Malcolm Gladwell’s like 10,000 Is it Malcolm Gladwell the 10,000 hours?

You know, you have to put the time in and you know, I I think we need to be really clear that like AI not only can it not s- replace that time, it it risks making it much harder to motivate students. And I was talking about this with like the creative director of like one of these like really big design agencies and we were talking about sort of AI art and he made this point that like, you know, the the motivation the the the the incentive structure for like learning art usually goes something like you spend, you know, a year drawing and doodling and struggling to draw a human face and then you eventually get to a point where you can create something really cool that you’re proud of that’s unique and and your own and that drives you to like learn more techniques and to spend more time and if AI makes it so that and I think I and I’ve already and I’ve talked to students who are interested who are, you know, artists or burgeoning artists and I asked them for their take on on AI art and they’re generally not excited about it because they’re like, well, now I spent all this time learning how to draw and my friends are creating way better like stuff that looks cooler whether or not it’s art I think is a separate discussion and it’s like demotivating and maybe the same would apply like why would you spend all that time learning JavaScript if you can get, you know, literally a working video game with a single prompt which I’ve seen now with with Claude code and with, you know, Gemini 2.5 Pro it’s it’s kind of wild actually what you can get with a single prompt. Yeah.

The that it can build an entire application is is completely wild. And I think that’s what I’m concerned about is that the incentive isn’t there. You know, I think I think the farming example is perfect where people still grow their own crops, you know, people will have vegetable gardens or they will have a artisanal farm and it’s not that farming completely gone away outside of industrial farming but it certainly looks quite different from when 90% of society were farmers, right?

And I think that that’s the question of like if that incentive structure isn’t there, the productive struggle I think is an incredibly valuable learning experience. So I want to be crystal clear here that while I think the JavaScript will go away as my own hot take, I’m not saying that it should go away. It’s just that I think the incentive if the incentive isn’t there the sort of value out of doing this and and the time it takes to to get there versus spending that time on something else, right?

Education’s all about opportunity costs that you have very limited time in the classroom. You’ve got like 30 weeks per year of instruction of instructional time and that time flies by and so what do you spend that time on? It becomes that really pertinent question.

And is spending that time learning What was it? Hand coding? What what was the the new Vibe coding or classical coding or like artisan coding?

Yeah. Classical coding now, I mean, that’s my first time hearing it but it’s already too funny that that’s now in the rearview mirror. So yes, so all that to say that those are all things that I think will become questions about 2 years from now.

Again, I don’t think these are happening today. The world of education always a little lags behind a little bit of of the workforce and these things sometimes take time. But I think that it’s valuable to try to get in front of these questions and try to think about what does that What is the best use of that time?

And And I don’t think anyone has the answer to it but certainly what is vibe coding is a is a huge question to to figure out and unpack. Yeah, but I there’s I think it’s easy with AI to go sort of go down the the glass half empty road and and there’s a lot more to talk about. I mean, we haven’t even gotten to sort of like artificial general intelligence where what do you even do in a world where nobody has to work and you get to a place where there’s like maybe it’s very important and interesting philosophical questions but they’re also questions in which I see very little agency for myself and our organization and frankly for the education system to to fully address.

I think it’s you know, to me AGI is like a question about sort of social safety nets and our ability to you know, figure out like the the fiscal policy such that we have the resources to be able to provide people. So anyways, it’s like a it’s almost like a political you know, political organizing question. But the glass half full version of this is also you know, I think one of the big deterrents to students coding like like going into computer science pathways is you know, today or at least let’s say 2 years ago, it was really hard and required a lot of like annoying work and effort to get to a place where you could create even a rudimentary or interesting video game.

And I think with Vibe coding in the hands of the right teacher, you’re not necessarily replacing class with Vibe coding 101, but your first day, like not even the first week, your first day in introduction to computer science, you are creating a video game. And to me like I would not I didn’t even have a computer science class in my high school, but I can tell you I probably wouldn’t have taken it, but I could but if if on day one as I was able to create, you know, some sort of it probably would have been some sort of fantasy, you know, Lord of the Rings type of video game thing, but that might have hooked me, you know, that might have actually like drawn me in and so I I’m curious, you know, just to bring things back to Quill, you know, one of the things that that we’ve been really impressed with your team’s it was it was it was been your team’s ability to, you know, not just use you know, your technology platform to really effectively build sort of the critical thinking skills and provide feedback, but also as a way to like really efficiently provide teachers with like current and just interesting and engaging topics that students are just respond well to when you’re putting out baseball was like I think well well taken, right? It’s like students are more more likely to lean into the learning experience if it’s something that they actually are, you know, interested in or or or feel somewhat passionate about.

And and to that end, I know that this is something we partnered on or is like creating some specific activities around artificial intelligence, which is a bit meta, right? Cuz we’re almost like using AI in the back end to help teach students about AI conceptually, but just to sort of can you just paint that picture of like what what are those activities like? How have those been received?

Yeah, they’ve been some of our most popular activities. We’re seeing that kids are really fascinated about these topics, you know, AI is so interesting in so many different ways. And so we have things like how AI is advancing animal conservation.

And this is one of those areas where AI is amazing that it is helping to protect endangered species and doing things like being able to use AI to protect elephants or being able to use AI to with whales. These are these ideas that really get kids excited about the future. And as we’ve been building these new activities, we’ve been getting some emails from students, which almost never happens, where the students are sharing their opinions and saying, "Oh, you covered this, but what about that?" Or this feels too optimistic on this particular topic or what about this other question?

And so you’re really seeing that the students are talking about these issues, they’re unpacking them, they’re debating them. And Quill had never really gotten into that level before with kids. And so for us that’s a huge win and and 100% what we’re trying to do in these activities is to really help students to be curious and excited about the future and to be able to think critically about it.

And so focusing on AI knowledge is just an incredibly interesting way of opening up this door for kids. We’re seeing this happening the most in English classrooms. That when we’re building these activities, they can be used in a STEM classroom, they can be used in a CS classroom, but that English teachers are looking for these opportunities to get their kids debating ideas, to build their own opinions and their own ideas.

And that this content has just been an incredibly rich opportunity for kids. And so we’re rolling out a whole series of new activities over the course of the next year focused on all of these really fascinating topics. So how AI is impacting art and creativity, for example, and how AI is impacting things like the future of work, how it’s impacting algorithmic bias for example, and how researchers are addressing and changing AI to mitigate bias.

And these are all really critical topics that we think kids will really be be really excited to to dive into. Yeah, I I love the the call out for algorithmic bias and like AI ethics because I sometimes am frustrated when people will describe AI literacy and they’re like, "Well, the key is that like students just need to know about algorithmic bias or like they need to just know about like the risks and benefit of AI." And and I worry that they like that sort of there’s the approach of AI literacy and thinking of it as a content knowledge is actually not quite there because what really matters is not so much like the awareness that it exists, but providing students with the agency to actually like you know, start to dictate, you know, what what their knowledge of AI algorithmic bias, like how that is now informing their perspective on you know, if and how they should be using AI. And like I think what’s powerful is that we don’t have answers to all these questions and that’s maybe some of the most interesting questions to then pose to a student because they have frankly as much entree to the conversation about AI art to give another example as as anybody else.

Do you it What’s like you know, what what is like the what what is coming now like like two or three years from now? Like how is how is Quill different or how is it the same? I mean, is is is your vision for growth more more scale and reach or or do you have like also sort of like a product vision that is maybe expanded expansive beyond where, you know, what you’re currently providing?

So, we’re really thinking deeply about what are those big questions that kids are going to be excited about that teachers are going to be able to find to be quite valuable. And one example here is the researcher Joy from MIT. She’s been doing a lot of research on things like facial recognition and how these tools can sometimes not have enough training data to represent different ethnicities and and races and have misclassification as a result.

Rather than just saying, "Hey, this is a problem." she was able to build her own data sets and retrain the models so that they were able to be more responsive. And to us, I think that’s just a really powerful example of how AI is a really malleable As you feed more data into AI, you change its output. And that can be really be used for good.

It certainly can be used for bad purposes as well. But that ability for students to dictate what AI is and what its output looks like is a really powerful thing. And I think we’ll see that this next generation of students they’ll inherit this technology and they’ll control it and they’ll be able to choose how it’s used.

And that understanding that they can change how it works rather than it being just this black black box that’s beyond our control, I think is a really important lesson for us to teach now. And so these particular case studies of retraining a system and improving it making it more effective, these are all examples of how AI can change. We’re big believers in this idea because that’s what we do every day at Quill.

When we’re creating feedback for kids, we’re building our own custom data sets and that we’re not just taking the output of the LLM and serving it to the student. We’re building data sets of more than 100 responses to a particular question for example, where we’re mapping out what are all the different things that kids are saying and how would teachers engage with these students if the teacher was sitting down next to the kid and working one-on-one to give feedback. And by building these data sets, by showing those exemplars of how students are writing and how teachers engage, we’re able to inject that all into the LLM to give it our own opinions of what good learning looks like.

And so we believe ourselves that So absolutely critical for good education, But there’s also this meta concept that for kids, they need to know that AI is malleable and it can be changed and that doing that impacts their lives and can impact their lives in a positive way. And so we think those are some of those big ideas that we’re excited to tackle over the course of of the next year. There’ll be this very meta level where we want to cover Quill’s own AI and explain it as they’re using it.

And so we see this all as a little turtles on turtles on turtles here. But we see this as a really powerful opportunity where we can unpack the how we train our AI ourselves and how we try to mitigate bias within Quill and use that as a learning opportunity for kids. But in doing that work, it’s really to try to help them understand that AI again is not just a static thing and that they can change it and they’ll own it and that in owning it they can hopefully steer it in a good direction.

Yeah, and I I can see why Quill is really well placed to do that because your approach is all about having students like hone their ability to sort of articulate their opinion or criticism or support for something. And I think that’s something that you know, I think students actually have the example that you gave in terms of just like even like the the partnership that we have in students where you get getting really activated. It’s like students once you sort of provide some scaffolds, they become quite articulate in their in the development of their opinions about AI.

But I don’t think that happens just by accident. I don’t think just because they’re using it and they’re digital natives that they necessarily have the the tools to to become really informed and I and I and I think about like, you know, the the TikTok algorithm and the fact that you know, we don’t need to teach students that algorithms exist. They’re talk about the algorithm, you know, like they they sort of innately understand that there is this sort of like thing in in the shadows that dictates the content that they see.

What we have found is that and we had this activity a while back that was actually like it would challenge students to train their algorithm to feed a certain type of content and you know, I don’t think that kids necessarily realize like all of the ways and I think that I did the it was like owls. It was like how how can you get as much as much owl content on on Tik Tok as possible. And this is the thing we use in the classroom so we kind of phase it out as more just sort of like a you know, an at home activity.

But you you’d have students that were like, oh, I didn’t realize like there were so many different aspects to how I use these apps that were were generating and it makes them like I think much more resilient as a when they see something it’s like well, you kind of have control like you’re seeing that kind of because you’ve been creating the reward mechanism for the for the tool. And it also shifts this narrative cuz I think again like the AI conversation can get can get very depressing if you feel like you don’t have agency. And I think what’s important is even with jobs, even with the future of work, you know, when you talk to like this is like Daron Acemoglu I think who was really pushing this is like we and David Autor as well.

I think it’s actually Autor. It was like it was closing remarks to one of his his talks that he gave recently and he was like, look, you know, AI is not going to happen to us. Like this like like when we talk about like what does the future look like in terms of jobs in terms of its impact on society like we are going to make decisions about the degree to which we use it to automate skills, the degree to which we prioritize you know, building human capacity alongside it and given that the kids today are really going to be the primary recipients of the those decisions, I think it’s it’s it’s both powerful and also like quite necessary, right?

For them to have the to be part of that. But it’s like it’s not like I think what you’re what you’re describing is not so much as like giving students a seat at the table. It’s like you can’t just give them a seat at the table.

They have to have have the the rhetorical tools to be able to like participate in the conversation and really contribute. Absolutely. It’s being advocates, I think is critical.

And I think that that example of controlling your algorithm is a is a fascinating one because I do think there is sort of the sense of this technology and you could What is an ideal algorithm? What content do you want to see? What makes you happy and joyful?

Should there be more cute animals in your feed because that makes you happier? Those are all questions where ideally students are the act driving that, right? That they’re driving their own engagement.

And they often probably don’t think about that or consider that idea. And I think that’s where Quill and aiEDU really step in to try to give them this chance to reflect and think about these questions that they might not have thought about. You know, our programs work really well together because Quill provides an introduction to a topic that we’re covering an article.

We’re get letting kids write about it. We’re getting them to build arguments and use evidence. And in doing so, we’re really opening the door.

And then from there, aiEDU with your lesson plans and your activities really goes towards that building experience of how do you take this and run with this and build something new. And so, I do think we’re both really trying to give kids these opportunities to reflect on this thing that impacts our lives right with the amount of screen time happening. You know, kids are spending what like 8 hours on their phones or something like that.

You know, it’s impacting our our lives in such an insane and intense way. We all know this and we all know that it’s not the most healthy thing. But, to give kids a chance to reflect on that and to think about that, you know, I think it’s really great the work that we’re doing.

And there aren’t a lot of folks right now doing this work. I think it’s really important that as this technology evolves that kids are seeing this happening and so the work you know that we’re doing together we’re in the early innings of this work right? The AI is going to be around for the rest of our lives.

It’s not going away. It the snowball is only going to keep growing in mass and so doing this work is really going to become more and more vital. Yeah, I mean I mean just to cuz I was going to ask like what you’re obsessed with.

But maybe I’ll refine that question to you know, compared to like where you were last year and based on what you’ve seen, I mean, are you how is your thinking about sort of just the timeline that we’re on changed? I mean, do you feel like are things accelerating? Are things sort of just like steady state high velocity?

Slowing down maybe? We definitely think So So for us the big thing is is the rebuild on generative AI and to us it really feels like day one. That there are a ton of opportunities for us to go deeper in building thinking.

These are things like teaching students how to build a thesis statement which for me was one of the biggest things I remember struggling with as a student that I would be writing an essay and I’d have to build a thesis and no one ever taught me what a thesis was. I just didn’t have that class or none of those teachers covered that and I’d be like, is this a thesis? Is that a thesis?

Like what what should I be saying here? And this is actually a hard skill for students to like build a thesis that captures their entire point of view and getting practice with that. There’s research that shows that this is incredibly impactful.

Building a topic sentence even is incredibly impactful. But there’s not a lot of instruction explicitly that gives kids those opportunities. And so for us that’s sort of something that we’ve always wanted to do.

It’s been on our road map for 10 years and the technology was never there for us to allow a student to build their own thesis and then for us to be able to evaluate and provide feedback and coaching on it. And so that all feels very doable today in a way that was not doable again. Even in that sort of first iteration of generative AI where it wasn’t quite there.

Now you have that really fine-grained analysis. And so I think Was that like GPT-4? Like what what what was was the inflection point in terms of capability?

In our in our own journey, there’s been I think two really big inflection points. The shift from GPT-3.5 to 4, and then for us the introduction of Gemini Flash 2.0. We spent a lot of time on GPT-3.5 when it first got released.

You know, we’ve been using this technology within like a I don’t know, a month or two of when it first became available. I think within a week of the API access. And it was not reliable at all.

It was just so bad at hallucinating and repeating itself and all of these problems. We spent so long trying to make it reliable. And what we should have just done is waited to be honest.

Like at the time we thought this is the technology, we got to make it work. So we spent so long trying to build these guardrails. And it was a good learning experience for us.

But certainly we spent so much time playing with this thing which ended up, you know, throwing out all that work. Four though was a big step forward where four was able to sort of give us much more reliable analysis with a caveat that four was really slow and really expensive. We’re helping millions of kids per year.

We’re giving feedback on around 500 million sentences every school year. And so at that scale, four would have cost us something like six to 10 million dollars per year to run. That’s bigger than our entire budget of our organization.

And so you saw this powerful technology, but it wasn’t at this sort of scale. It also was too slow. You know, for us when kids are writing, we need to give them feedback in under a second, right?

We can’t sort of have that long analysis period. And we knew that models would get faster and cheaper. But for us the Gemini Flash model has represented a really fast model that gives really great output and is reliable while also being cost-effective.

And we see that Flash 2.0 there’s still room for growth. That versus the really powerful pro models, there is a gap there. And that gap will get closed over time, but it certainly is at that point where we feel confident that we can deploy it to production in a way that’s real and scalable.

And so for us, that’s been a really exciting threshold. And that has only been in the last 6 months that this technology is available. I think it came out like in August of last year.

And so while there’s been a lot of FOMO around generative AI since essentially as soon as it came out, the truth is is that there was a sort of period of getting from this technology exists to it’s reliable and it’s fast and it’s cost-effective. And I think we’re in that territory today. We only entered into that territory very recently.

And so for us, that’s critical towards actually being able to use this at scale versus our own models where when you build your own model, you have a lot of control over it. You can control the cost, the model design. But that was just a very slow process.

So now our our iteration loops are a lot faster as well. And something that allows us to do a lot more than we ever could do in the past. Yeah, I think I’m curious if you’ve struggled with this because I I’ve often had conversations with with foundations and you know, they’re trying to figure out what their AI strategy is.

And there’s this like sense that what they need to do is invest in an AI nonprofits and I don’t know if I’d consider Quill an AI nonprofit. I mean, you’re using AI, right? But I mean, I think actually like you’re an organization that is like you know, working to use technology to help students, you know, read and think critically and you know, articulate their opinions.

But you’re you’re really nimble and you’ve been able to really effectively deploy AI. And it’s And it’s interesting is in the venture spaces is actually something that a lot of VCs have been talking about. Like even like like Andrew Ng’s AI fund.

I mean, their whole thesis is like, you know, look for companies that are really well placed to leverage AI to like rapid iterate and get to sort of like an MVP. But yeah, I mean, I’m just curious like from your perspective, I mean, do How can we do a better job of sort of articulating cuz cuz I don’t think it’s just funders, I think it’s like school district leaders as well, like the buyers. I know you’re a nonprofit, but but to your to your sort of like customers, let’s say, I mean, is there Have you Have you run up against, you know, folks saying, well, ah, I’m really trying to figure out what like the generative AI tool is that we buy?

And is Quill really like you know, Quill isn’t competing per se with like Gemini. Like they’re they’re they’re very different tools. And at the same time, it feels almost more important for teachers to be using something that has sort of like the pedagogical structure that you’ve put in place than to just sort of like have this like multi-purpose tool that doesn’t necessarily have the like the deep thinking behind it.

Yeah, I think for us, we are very much a literacy nonprofit. That our goal is to build strong readers, strong writers, and strong critical thinkers, and that’s what we’re all about. That’s the end game here.

How do we help millions of kids build these skills? And AI is just a tool that we use in that process. And so, I think we are an AI nonprofit because we have that expertise in AI.

As a nonprofit, a third of our team are software engineers and product managers, and that our team is doing all of our own in-house AI development. So, we have that skill, but it’s a little bit like saying we are a nonprofit that uses software. Or like an internet nonprofit.

An internet nonprofit, right? Or a database nonprofit, right? That all software uses databases, and that’s just part of it.

And so, I do think that right now there is this class of organizations where AI is part of their model of delivery, but in 10 years every organization will use AI in some capacity. And that distinction of like, are you an AI nonprofit or not, will go away. It’ll be a question of like, who’s building novel use cases on it.

You know, cuz there are a lot of these thin wrapper tools where they’re just taking the output of ChatGPT and trying to sort of wrap some service around it. And sometimes that can be valuable, but sometimes that’s not valuable. And so that I think that’s a very different question of that will become very easy for anybody to do.

And you won’t be an AI nonprofit because under the hood a model is helping you in some capacity. And so I do think those are some of those distinctions that apply now but won’t in the future. I think the more interesting question though is that AI is getting a mixed reception in schools.

We are working on our messaging right now and we don’t talk a lot about AI on our website. We have our AI program for kids, but it’s not splashed across the homepage. And we figured maybe we should do more on this, right?

The AI is so central to our work, we’re developing it, we’re using it. Like let’s make that sort of part of the sort of special sauce of Quill. But teachers reacted pretty negatively to it.

When they see the word AI, they’re worried that this is just going to be a cheating tool, that this is going to replace them, that in our own surveying of teachers, the sentiment was fairly negative that as an being an AI company is not why they love Quill. And that calling ourselves an AI company felt like the wrong step forward. And I think we all sort of saw like crypto company, you know, I don’t want to have the like crypto company vibes, right?

And so that’s a little bit of the feedback that we’re hearing from teachers today. And I think it’s nuanced because there are some really good folks using AI and then I see a lot of like not so great products as well. So the sort of space has a mixed reputation right now because of those like really ethical AI players doing usually very narrow use cases and doing it in a highly customized way.

And then folks who are just sort of promising the world with AI in a way that doesn’t actually deliver what students and teachers need and comes with a lot of potential problems as well. Yeah, I mean I was actually just talking to, one of the biggest school districts in the country, and they actually have someone that’s like leading their their generative AI strategy. And they just banned, I won’t say the name of the company, you probably guess, one of the big wrapper, yeah, one of those popular sort of for-profit wrappers, because what they found is that it was you know, there’s like some instances of teachers using it really well, but there was like lots of instances where it just like it wasn’t being used effectively.

And you know, we actually have to go to lengths to like we I open up almost all my meetings now with school districts and I’m just like, "We are not we are not the AI implementation project." like we’re actually in most cases our advice to schools is, you know, when they come to us and say, "Oh, what AI tool should we be, you know, providing to students and teachers?" We’re like, "None. Like you’re like you should actually, you know, pump the brakes, focus on the question is more like, how do we provide a sandbox for teachers to actually start to experiment with stuff?" I think Quill’s interesting because it’s to me like that’s actually easier to take to a school district because it’s already aligned with priorities that they have, right? Like schools and the NAEP scores really underscore this like, you know, many if not most schools have like significant ground to cover in terms of literacy.

And so solving for that problem I think resonates with like a much broader audience. I am interested though in this sort of like meta component that you’re talking about where like as students are using Quill they’re also kind of like learning about how Quill works. I’m curious if there’s like a teacher-facing component to that as well because I’m I’m fascinated with like, how do we build it like I almost worry more about teachers and students.

I think the students are going to figure it out like far more quickly cuz they’re just saturated with it and they’re sort of, you know, they’re very tech-forward. Yeah, do you I mean does does Quill like does Quill have a teacher-facing component? I mean do you see any any opportunity there to just sort of use your your platform as a way to help teachers kind of see sort of like what it looks like to implement AI, you know, really effectively on the back end?

Yeah, so so everything I’m talking about now is a project that we’re working on and something we’ll hopefully be shipping with AIEDUO sometime over the upcoming school year. So a big caveat that this is not yet live, but we do want to build activities specifically around how we build our training data sets and how we use those data sets to evaluate writing. And then this is a really important topic because evaluation of writing is being used for things like Quill or Quill is a very low stakes practice platform where kids get to practice and receive feedback.

It’s also being used for testing purposes. Every year folks like the College Board hire tens of thousands of educators to grade all the AP exams for example and there’s a ton of work that goes into those evaluations. And that as AI builds these skills, evaluation of writing is going to become part of education.

And so getting that right and making sure it’s reliable and accurate, those are all really critical problems. And that you solve those problems again through good data. And so for us this is a really critical topic.

We want to introduce to students, but also there’s an opportunity for teachers to learn as well. And so we’re excited to cover this topic. It’s a little bit scary though cuz we’re pulling back the hood a little bit here on on what we do in our work.

And so it’s not yet quite available, but we see it as a really powerful opportunity for us over the upcoming year. To your broader question though about how teachers engage with Quill, we provide them with a platform with access to more than a thousand activities. They get to assign activities to students, view their results.

They get to see all the writing the kids are doing on the platform. One of our tools, Quill Lessons, is a multiplayer tool where the teachers and the students are learning together. That’s one of our most beloved tools.

Kids get to share their answers with each other. They get to debate the answers. That’s a a real-time tool for students teachers.

And so we’re trying to really intentionally think about how to create tooling specifically in the K-12 context and a higher ed context where teachers are empowered to engage their students and be partners in their learning. And there’s a a couple of big design decisions. Again, things like these are bite-size activities that are layered into the classroom.

They’re used a couple of times per week. All these decisions are critical towards Quill being a partner to teachers and helping to advance their their goals as opposed to something that lives apart from the classroom and tries to replace a teacher. And so we think that by making all these smart decisions, teachers feel really empowered by Quill.

We also have great training and webinars and all those things that connect teachers together. But the heart of it is really the sort of intentional design, which is designed for teachers and really to be a partner to them. Yeah, I mean that’s like that’s the golden goose really is like it can we use AI and technology to actually enhance collaboration and you know, it’s sort of like those these like human durable skills that students need to build alongside like yes, critical thinking and the knowledge base itself.

And I think in the right hands, AI can absolutely make, you know, because like not every teacher knows how to create a really effective project-based learning activity around like any topic, right? And so you know, it’s it’s it’s definitely not so simple as like oh, AI is going to is is is good or it’s bad, you know, for use in the classroom. I do think that teachers are really well placed to look to organizations like yours that have literally been obsessed with this question for a very long time because it’s not necessarily like something that can be turnkey.

And so but the good news is there are free products like Quill available and also, you know, AI Tutor sort of has a very similar approach, modular, bite-size, not trying to be the curriculum. It’s hard to imagine what the curriculum would be for AI readiness. It’s actually more about how do we get more organizations to sort of like start to adapt some of the practices that you’ve you’ve taken like this very sort of like self like introspective approach to you know your your product design and like being really intentional about when not to use AI, when to use it, and role of teacher, etc. I think the headline here, what I hope everybody is doing or trying to do, and I think we’re on the precipice of this moment, but it’s to really engage in deeper and more active learning.

I think that when I look at EdTech, Quill’s mission statement from day one has been disrupt multiple choice questions. My very first grant application to the Gates Foundation was just about how multiple choice isn’t the best way of learning. I remember as a student doing so many multiple choice questions and ABCD, select the right answer.

You’ve got three wrong answers, one correct answer, and so you kind of just could guess. Like this is clearly a wrong answer. I’m going to go with A or B.

And you’re not building an argument, you’re not expressing your own idea, you’re not building something. And I think that’s a really powerful way of learning. And that as we think about AI in the future, using a thin wrapper tool to generate multiple choice questions for you isn’t really advancing learning forward.

You’re taking us something that we’ve been doing now for decades and just making it a little bit faster. But the more exciting thing is how can we go from multiple choice to writing and to project-based learning and towards collaborative learning and things that are very hard to do well, right? There’s been a big push for many years for project-based learning.

It’s very hard to implement in the classroom. It’s hard to get 30 students all working on projects, but you can imagine a number of ways in which AI can serve as a partner to the teacher in a way that previously was not possible. And I think all those opportunities lead to deeper, and richer and more effective learning.

So I think that’s the name of the game here is how do we reimagine learning and what are those opportunities that we can now pursue that were just hard to do in the past. And that’s where we should be applying our effort as opposed to just using AI to automate what we’re already doing which is not the most effective and deepest form of learning. Yeah, I mean I couldn’t think of a better way to close it.

It’s like do the status quo is clearly not working. And when if if AI just becomes a way of allowing us to sort of like cover resource gaps to maintain the status quo, we’ll have failed. But there is this opportunity if AI can actually it’s almost like a the Trojan horse for these sort of like much longer you know sort of very very old and frankly boring conversations that have been had for decades, right?

Like 21st century skills, you know, digital readiness, if you know, project-based learning, critical thinking. Like none of this is new and I think there’s actually power in that because there’s a lot of disruption happening to schools right now, you know, at the national level, at the state level. It’s not necess- you know, I don’t know that educators respond well to like more disruption and they don’t see it necessarily as a positive.

But I think there’s there’s sort of subtle but really intentional ways that the technology can actually just make it easier for teachers to start to implement some of these practices that you know, are not necessarily intuitive, but but can be turnkey with with amazing tools like Quill. Peter, anything else that we missed that you want to that you want to share before I let you go? I know it’s relatively late on the East Coast.

Thanks for making time for me today. Yeah, it was a lot of fun. We we covered a few really big questions and really excited that this is just again the early innings of this of this world, you know, that AI is going to be here for the rest of our lives.

There were a lot of things to figure out and I hope that we can spend more time trying to get ahead of some of these questions that it’s, you know, hard to imagine what the world will look like 10 years from now, but we can certainly see certain trends. We can see that we’ll have more information than ever. We’ll have deep research queries that are feeding us 100-page documents and that we’ll need to be able to think critically, to be able to parse them, to be able to have our own points of view.

Education will become more important than ever. And that if we do it really well, if we make it active, if we make it joyful, it will be a really amazing opportunity for kids. But if we don’t get it quite right, I think we’re going to feel a somewhat scary world where we’re all a little a little bit taken aback by the world where AI sort of is something that lives beyond us.

And happens to us. Well, and you know, who better to help us answer these questions than the students themselves who are going to be both a part of that world and also building it. Yeah.

Absolutely. Peter Gault. I’ll see you in I guess a few days, right?

Are you going to be at ASCD? Yes, I’ll see you next week. Okay.