Dreams and Delivery
Real stories from tech pros who deliver at work and dream beyond it. We cover the tech landscape in an age of AI with layoffs and opportunities that come from change.
Dreams and Delivery
Aaron Kagan: GraspingAI - Exploring the Human-AI Connection
Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.
Exploring the Human-AI Connection
This conversation goes places most AI discussions don't. We get into why most companies don't have an AI problem — they have a human-AI coordination problem — and what that distinction changes about how organizations actually adopt these tools. We also talk about whether AI can dream, what the word "hallucination" reveals about our assumptions, and what it looks like to leave a staff-level role and build something completely your own. Plus: what parenting a two-year-old in the age of AI is teaching him about the things no model can replace.
Aaron Kagan has spent his career studying the gap between how people actually think and how the technologies we build assume they think. At Google, Meta Reality Labs, and Native Instruments, he was asking the same question the whole time — and going independent with GraspingAI gave him the space to finally pursue it on his own terms.
Bonus: the rubber duck analogy that perfectly captures what AI actually does when we think it's being creative.
About Aaron Kagan
Aaron Kagan is founder of GraspingAI, a human-centered AI research and strategy practice, and co-author of An Introduction to the Embodied Mind (Routledge, 2026). He held staff-level UX research roles at Google, Google X, Meta Reality Labs, and Native Instruments, studying how people think, decide, and work with complex technologies. He now helps individuals and organizations build better relationships with AI through research, strategy, and open community workshops.
Connect with Aaron: linkedin.com/in/aaronkagan | graspingai.com
About your host
Christine Kahn is a Staff Technical Program Manager with 12+ years leading AI platforms and program delivery at Intuit, OpenTable, 20th Century Fox, and Realtor.com; and a musician and soundbath practitioner outside of it.
Connect: linkedin.com/in/christinekahn | christinekmusic.com
© 2026 Dreams and Delivery LLC. All rights reserved.
Welcome to Dreams and Delivery, a podcast for tech professionals, real stories from people who deliver at work and dream beyond it. I'm your host, Christine Kahn. By day, I'm a staff technical program manager in big tech. By night, I'm a musician and a sound bath meditation practitioner. This is a conversation about the human side of AI, how people actually think, create, work, and make decisions with these systems. We'll talk about AI dreaming, what it really means, and we'll cover what it's like to go from a staff-level role in big tech to building something completely of your own. Today I'm joined by Aaron Kagan, founder of Grasping AI, a human-centered AI research and strategy practice, trained in philosophy, cognitive science, and UX research. He spent his career studying how people think, make decisions, and work with complex technologies. He's led research and AI-related initiatives at Google, Google X, Meta-Reality Labs, Native Instruments, and Articulate, and is co-author of An Introduction to the Embodied Mind, Thinking Outside the Head. Today, he helps individuals and organizations build better relationships with AI through research, strategy, and education. Welcome, Erin. Hey, thank you so much.
SPEAKER_00I'm psyched to be here for having me.
SPEAKER_01So nice to see you. Erin and I worked together at Native Instruments years ago. And it's just great to connect with you years later and just see everything that you've built. The work landscapes really changed a lot. I I feel like compared to five years back when we were just leaving Native and now different playing
A Different Playing Field in Tech
SPEAKER_01field.
SPEAKER_00Yeah, it's a it's a different space and it's you know, things change. None of this stuff can last forever, and you see things grow and take off, and it just everyone kind of has to adapt in their own way and sort of see where where the next iteration of things is is taking taking us. And right now it's a really tumultuous period of transition, which is exciting and terrifying at the same time.
SPEAKER_01Yeah, I think with AI in the playing field, it's just it's making a lot of things easier, but also a little bit different. And we can get into it more in the convo today, but wanted to give you the chance to just intro yourself and how you ended up working at the intersection of cognitive science, philosophy of mind and AI is such a unique center to be.
SPEAKER_00I appreciate it. It's always kind of been the same thing for me, but I'd say, you know, I I got into philosophy because I really wanted to understand the mind, like perception and consciousness and all those sort of big questions were really interesting to me. And then in uh, you know, university and academia in general, I was like, that's it. That's the thing to do, and I'm I'm gonna do that forever. And then I I pushed for that, and you know, lots of different bumps in the roads and setbacks and things happened. And when I got towards the end of the PhD, I was like, you know, I I kind of want to go into some of this other stuff. I was always interested in like how the things worked in in with neuroscience and biology, and that kind of led me to cognitive science and some of this embodied cognition stuff, which we're I'm happy to talk about. And that led me to UX research, where, you know, the theory kind of meets real people and real practice. And I think throughout all of it, I've spent my career studying that gap between, you know, how do people really think and how do technologies that we are making assume people think? And, you know, AI being right now one of the most important technologies that forces us to revisit these questions. So the stuff that I've been interested in is now critically relevant in in a lot of different ways.
SPEAKER_01Could you tell us maybe the short version of what is embodied cognition? Why does that matter in AI?
SPEAKER_00Yeah, sure. So the the UX folks in the in the audience might understand this from an affordances standpoint. People talk about that with like UX design and like is a thing sort of graspable or clickable and cognitive offloading is another term that often happens in tech that people understand. But
Embodied Cognition and the Human - AI Coordination Problem
SPEAKER_00um, embodied cognition is kind of at the base of all that. It's it says intelligence doesn't really live inside the head. It emerges from, you know, brains, bodies, environments, tools, and other people working together in the right sorts of ways to basically uh talk about where cognition happens. It's not just happening in the head, right? And and that's sort of the one of the biggest lessons there. And because of that, the biggest lesson right now in in AI and and what we're developing here is that a lot of organizations are having a problem with collective knowledge or understanding things or people, you know, cognitively offloading in different ways. And they really don't have this AI problem, I'd call it. It's more of like a human AI coordination problem. We coordinate together easily, talking about different things or sharing memories and stuff like that. But we don't do that as well with artificial systems that work like this. We do it pretty well with calendars, right? I I didn't remember what time this meeting was, but I knew that I could look it up on my calendar and like that memory was was offloaded because I knew I could access it quickly and easily. But that happens with easy stuff like notebooks and you know, stuff that is is pretty reliable.
SPEAKER_01Yeah. Maybe we could get into a little bit about AI strategy, because I've seen AI transformations happen at companies where they'll say, all right, here's the tools, go do it. Become X percent improved in your efficiencies without being very intentional about it.
SPEAKER_00Yeah. Yeah. I think
How to be Strategic with AI Transformation
SPEAKER_00the the biggest the biggest takeaway here really is that like it's for a lot of us, at least for those of us who aren't actually building these models, it's not really about building smarter models. It's about building the relationships between humans and these models or these systems. So what I've seen happen in, you know, or most organizations, which are when I say most organizations, I mean not like the ones that are building a lot of these things, is they they end up treating the technology as a problem or as a product, right? AI is is just like saying the internet, right? It's it's too big of a target. Like what is it exactly? Are we talking about large language models? Are we talking about different kinds of you know, robotics or different kinds of predictive processes? And people end up chasing different kinds of tools rather than the capabilities, and they focus on models like, oh, this one's smarter than that one, and sort of ignore how people are naturally working now and don't invest as much in AI literacy or even judgment about this sort of thing, because it's very easy to output polished, seeming stuff very quickly, and people don't necessarily know how to use them unless they're like really heavily invested in this stuff. And we see this today. I mean, more recent things I've seen are a lot of companies who invested heavily in one product or another for the organization have already burned through all of their credits, like in the first quarter, and now they've got to buy more. And that isn't necessarily the most efficient use of the tooling, right? I don't think a lot of folks who understand how this worked would go through it, the those that fast. And what it actually is changing. Recent thing, one last nerd thing I'll say about this is like even Nielsen Norman put an article out a couple of days ago about they interviewed it wasn't a huge study, but they interviewed a lot of people who are not software developers but use AI for hours and hours a day. And what
The AI Usability Problem
SPEAKER_00they found is that they actually are not learning anything new. They're learning how to get better outputs, but the skill of like, okay, this is a thing that could potentially teach me or walk me through something, they're just sort of getting outputs over and over, and that knowledge about how to use it is not increasing, and nor are the things that they should necessarily be getting better at. So it's uh it's a really tough, tough one right now. And I think for us in tech, we are this is a use usability problem. And I don't just say this because I'm a UX person, but it it doesn't matter what the technology is if it doesn't if the user doesn't interact with it correctly, everybody has a bad time.
SPEAKER_01Yeah, so true. And things are evolving so fast where, say a year ago, the focus was on prompt engineering and just making sure you're prompting, whereas now we've moved more into feedback loops where you don't want to just prompt, you want to let it continue to run without you and then give you an output that is really thought out well. I'm
Can AI Dream?
SPEAKER_01curious because I keep hearing lately that AI can dream. Can it really? What does that look like?
SPEAKER_00Yeah, I I think, you know, AI, I'd say, you know, from a from a very open-minded and and deeply like theoretical philosophical perspective, I would say that it maybe can, but not in the way that humans do, right? It doesn't, this technology doesn't sleep, it doesn't have a body or desires or emotions or like lived experience, right? Those are the things that are usually all prerequisites for dreaming, at least for as far as we know. But it can generate combinations and associations of things that can almost feel dreamlike to us. And so the question doesn't become necessarily whether it dreams, it's like, why do we experience these outputs as meaningful or creative or surprising, right? And we we even use terms like hallucination. And and like that could almost be a dreamlike thing. You're sort of in a fugue state and you have a hallucination for, for example. But what's fascinating about that term is that it causes us to think that the system is doing this very human-like thing, like maybe they hallucinated it or they're they they took mushrooms or something like that. When in actuality, what it output is not a hallucination. It's it's as we know from from tech, and as we say, this is working as intended. The system is doing literally what it's designed to do. So it's not actually doing anything wrong. It's doing exactly what it's designed to do. So that's different from saying, you know, you have a calculator and you say what's eight times eight, and it it outputs like 8,020, and you're like, wait, that the calculator is broken. That's not how it's supposed to work. But it's gonna, because it's gonna output something very specific every single time. We don't have that with hallucination. It did what it was programmed to do in the way that it does what it does. And part of that opaqueness from even an engineering standpoint is interesting, and part of that opaqueness from what we expect it to do and how we even talk about it is is very, is very interesting. It's like it's misleading in some ways, but it also helps us use it faster and more efficiently. So it's it's a very nuanced kind of thing. And I'd say that the dreaming thing is is a great way to kind of get into that conversation.
SPEAKER_01Yeah, because when I think of dreaming, like as someone who's creative, makes music, I'll just have ideas pop up in my head on Ben versus when I think of AI dreaming, it's more of a retrieval, just in a way that is a little bit more in-depth.
SPEAKER_00Yeah, I'd I'd almost look at it as like the dream, uh dream enhancer or a dream organizer. Like you can they used to call like back in the human factors days, I had guys in the lab of you know the term rubber ducking, like that they use sometimes in incorporate. Yeah, but guys in a lab are trying to work on a problem, and before they bug somebody else, they got to go in, you know, some corner somewhere with a rubber duck and like talk it out. And a rubber duck obviously doesn't do anything, but it allows the person to work through whatever it is and they may have solved the problem without interrupting anyone else, right? And this is kind of an even more interesting and automated way of doing that. So there are times where I'll I'll try to work through a gnarly problem and just be walking and talking through that. It says something back to me, and I'm not thinking that it's gonna solve my problem. It actually is a reflection of my thinking or a reflection of myself. Say, oh, okay, I seem to be indexing on this or going through that. And it can bring together interesting associations and like we do, you know, in music or any kind of creative endeavors, are you sort of experimenting and playing and throwing things around and seeing what it sounds like. I mean, you it that's an easy analogy, is like you know, if we were working in a DAW or some sort of audio workstation and you started like clicking around and say, okay, what happened if I move this over here and this and that, hit the spacebar and it plays, and you're like, oh, that's what that is. It's a similar kind of thing. It's more like that than I would say doing its own dreaming or or kind of creativity.
SPEAKER_01Yeah, it's like still following some sort of sequence that was either defined by you or whatever skill you've loaded. So yeah, curious to see how it evolves over time.
SPEAKER_00Me too.
Going Independent and Creating GraspingAI
SPEAKER_01Shifting a little bit into how you went independent and built your own thing, could you tell us what was it like leaving corporate and creating grasping AI?
SPEAKER_00Yeah, leaving corporate was was interesting. I mean, the things are growing and changing so quickly. There's a lot of there's less stability, there's less predictability of what's happening. And I saw that, you know, industry is changing rapidly, AI is changing, how knowledge work kind of stuff that we do happens. And in starting this, I initially said, okay, the work I was doing could be leveraged even more so if I went independent. It's almost like we're going through a platform shift. Similar thing happened when the internet came out, right? Every business thought they needed a website. Some of them did, some of them didn't. And and usually it's in between. It's like a shop just like a bagel shop just needed a static HTML page with the information on it, while as other ones might have needed like e-com and customer service and all kinds of functionality. But that means every business has a website, but not in the ways that you'd think. Same thing happened with when we went mobile, right? Needed to be mobile friendly, or everybody needed an app. That's sort of something that we all saw. This transition feels similar to that. So companies can use this technology. They can leverage it, but they need to do it in the right ways or ways that are good for them. And they also can potentially shoot themselves in the foot by over-indexing on or over-leveraging AI. And so this was a chance to help businesses do that, but also pursue questions I have been thinking about in general around human AI collaboration, what expertise means, decision making, cognition, and sort of the future of like working together. And these changes are happening across a number of disciplines in an accelerated way. And as we know in corporate, it's kind of hard to collaborate in certain ways, do interdisciplinary work and pursue things experimentally in these kinds of environments. And so that's kind of what accelerated that decision to go independent and then to go with the sort of 2.0 that's that's kind of happening this next couple of days with the group of folks that I've been building this with.
SPEAKER_01Exciting.
SPEAKER_00Yeah.
SPEAKER_01Yeah, actually that would be a great chance to share
GraspingAI 2.0 - Human Centered Practice and Community
SPEAKER_01more about the 2.0 that you're building.
SPEAKER_00Yeah, yeah. It's it's it's kind of turned into a human-centered like research and strategy practice, or almost like a studio, where on the one hand, we help organizations build better relationships with AI, not just like adopting tools, telling them how to learn and think. But there's also this the front of this really is about community. Um, it's not your typical sort of consultancy or some of these other cottage industry things that have popped up. This is about large-scale dialogue and understanding for everyone to leverage this technology well and experimentation and sort of everyone contributing in this kind of dialogue. And that information then lends itself towards things that people can use. So speaking of prompt engineering, there are kinds of ways that you can become a better consumer, for example, of all these different products. If you have your own product agnostic set of prompts about yourself and throw those into the different products, and then you can have an Apple Hubble comparison. Say, oh, I heard an advertisement that this new model came out and it's smarter. Well, how is it going to know you if you have some cost with the other one? Right. So this is one of the things we're doing in the next couple of months is having open workshops where people can learn to do these things. And if an organization wants to help that, then we can help them do some customized and proprietary kinds of things. But this is about experimentation openly and working out in the open with a community of people first, then sort of bringing that into organizations and into business.
SPEAKER_01I love that because really having that cross-section of people from various disciplines being able to form some point of view around AI. And I know responsible AI has been one topic that's been coming up a lot for everybody. I could definitely see that being discussed in that sort of forum.
Tips for Starting Your Own Business
SPEAKER_01For for somebody who's starting out thinking of building their own thing, what kind of uh advice would you give them? Little tips to get started.
SPEAKER_00Yeah, I'd say some of the latest today is that it it is in some ways it's never been easier to do and less scary and less risky. And in other ways, it's it's much harder because of the landscape. The the tooling, the ability to sort of hang out a shingle is is quite easy. What I'd say is don't I'd say, you know, follow, follow that like the what you are passionate about and try to realize that, but also stay very conservative into how much you over-engineer the stuff. So there's a lot of like, it's easy to sort of cosplay corporate stuff and sort of recreate some of the things you might be used to from uh larger organizations that isn't necessary until you have uh the need for that. Sometimes it can slow you down. It is incredibly freeing to be able to try any tool or to go through anything without any restrictions and not have this uh muzzle or constraints, which is wonderful. And I'd say just to stay with like-minded folks, figure out the cheapest minimum viable way to go about this and stay motivated and stay making progress every single day, you know. And it's almost like similar to writing my my dissertation or a book or like larger things, it's just sort of brick by brick, you know, building this sort of larger thing. And I wouldn't be as worried about things like, you know, funding and uh all the sort of traditional startup models these days, because you can get away with, depending on what you're offering, you can get away with a lot. And it stays your own and it stays uh less scary in terms of a big push before you know whether this thing's gonna work or not, and just try and experiment as much as you can.
SPEAKER_01As someone who has a child,
Parenting in the Age of AI
SPEAKER_01I'm so curious of how you're thinking about incorporating learning in this new age of AI.
SPEAKER_00So Pardon that we talk about how we're gonna uh how he should be engaging with technology, right? We uh we're always thinking about screen time, that was that was one thing. But then, you know, we've we've had these uh we've got these systems that are frictionless that can serve up information. I think about things like necessary friction involved in learning and growth. Like these things don't feel good. How do we instill these kinds of of things in in a person so that they get that that thirst for knowledge and growth and and learn it in the right sort of ways, right? And I also think about what what these technologies mean for education. There's stuff on either end, right? I it and there's things that we can control as parents and the environment, and there's other things that are going to be out of our control. And one of the one of the best phrases I heard from my my cousin-in-law early days when my wife was pregnant and we were sort of like when you have the time to sort of think about all the intellectual stuff you want to do before you're, you know, sleep deprived and just going on basic primal, like primate instinct to keep the kid alive. He said, you know, you don't prepare the path for the child, you prepare the child for the path, which I really like because there's only certain things that you can do. It's about instilling those kinds of instincts. I think there's going to be some stuff around the next level of learning. We had a lot, you know, in our sort of earlier generations with information literacy and like everything seems because it's in typeface, everything seems authoritative or like, you know, fake media and interesting, you know, stuff like that. This is almost to a whole other level where you have things that are psychophantic and can get you in your own further echo chambers or really preventing you from learning. You say, oh, I could just ask AI to do it. Why do I need to, you know, uh go ahead and learn this thing myself? And and trying to figure out where that sits. My hope is, you know, there's going to be two in uh in about a month from now. And my hope is that, you know, maybe when it's even more critical, some of this stuff will have figured itself out in terms of like policy and regulation and you know, computers and things in the classroom and how we go about that sort of things. All this is to say I've I have no idea what I'm doing because as any parent should honestly say, they they they don't because we just don't know. It's just gonna it's gonna be what it is. We go on instinct and and we try to do the best we can for the stuff that we can control as a as a parent.
SPEAKER_01Awesome. Where can people find you and get in touch with you?
SPEAKER_00Yeah. Um graspingai.com, g R A S P-I-N-G-AI.com. Find me on LinkedIn, uh Aaron Kagan. And uh yeah, there's plenty of places to sign up and and connect with me from there.
SPEAKER_01Thanks so much, Aaron. Yeah.
SPEAKER_00Appreciate it. Thanks for having me.
SPEAKER_01To everyone listening, if this resonated, subscribe and share it with someone who needs it. I'm Christine Kahn, and this is Dreams Hunt Delivery.