Algorithms + Data Structures = Programs - Episode 278: The Age of Ideas

Episode Date: March 20, 2026

In this episode, Conor and Bryce chat about the age of ideas, whether AI will lead to cognitive atrophy, some exiciting personal news and more!Link to Episode 278 on WebsiteDiscuss this episode, leave a comment, or ask a question (on GitHub)TranscriptSocialsADSP: The Podcast: TwitterConor Hoekstra: LinkTree / BioBryce Adelstein Lelbach: TwitterShow NotesDate Recorded: 2026-03-10Date Released: 2026-03-20Hank Green - This is Going to be Very Messyarraybox.devpodgod.cakp (Kernel Profiler)Intro Song InfoMiss You by Sarah Jansen https://soundcloud.com/sarahjansenmusicCreative Commons — Attribution 3.0 Unported — CC BY 3.0Free Download / Stream: http://bit.ly/l-miss-youMusic promoted by Audio Library https://youtu.be/iYYxnasvfx8

You've got the designer, the software engineer, and the product manager. And the short version of it is every single one of them doesn't think they need the other two now. Yeah, yeah, yeah. Like the software engineer thinks they can do both designing and product. But it's like, it's so true. Five or six or seven years ago when I was doing a lot more day-to-day engineering, I didn't know any of that stuff either. I don't think anyone ever did. If I needed to know how to use something, I would go and do some desperate Google search.

Product managers and, like, team leads, they use. They used to just do what we are now doing. Like they would delegate to employees. And now we can delegate to these LLMs. And it is intoxicating. And I mean, I've kind of buried the lead. I forgot that I was supposed to tell you this. So very exciting personal news.

Welcome to ADSP, the podcast, episode 278, recorded on March 10, 2020. My name is Connor. And today with my co-host, Bryce, we chat about living in the age of ideas. the ideas we have, whether AI is going to lead to cognitive atrophy, and also some exciting personal news. End of episode one, start of episode two. I hear people, like all these concerns. I listen to, I think it was a Hank Green.

I don't know if you want to call it podcast, but it was him kind of monologuing for 75% of it, but then he talked to a professor for a little bit. Maybe those percentages are off. But there were all these concerns that he had, and like one of them was like cognitive atrophy. And the idea that, you know, because we're not thinking, quote unquote, as much, because of these tools, we, the same way you can have muscle atrophy, which is when, like, your muscles degrade when you're in outer space or because you're, like, bedridden in a hospital, or maybe degradation is the word. I mean, atrophy is the right word, but I'm just

trying to find a word to explain it. But anyways, you don't use your muscles. They disappear. And so this is the same idea for your brain and I don't know I just I just I think that's kind of like wild that you think people are going to not be using their brain like you're going to use different

because on one hand I do agree you do lose the ability like I always have this analogy of dictionaries like kids don't look things up in dictionaries anymore unless if you're like a great parent and you have a physical dictionary and you get your kids to look it up but I don't think most

kids are using dictionaries these days. And you learn so much from looking things up in dictionaries, from the distribution of like letters. You know, like if you're looking up something that starts with a Z or Z or Z, if you're in America or a different country, you just go to the back and you start flipping the pages, right? Whereas if you are looking something up with an A, you still might flip a little bit. And like, where is F and P in relation to like the start and end? And then once you get close, you know, looking at the top left and right words. There's so much, like, you're basically doing a kind of binary search at a certain point. Kids don't do that now. You just, you know, go to type in the word and go define. And half these times, you just ask Alexa now or

Syria or whatever, like, AI, you know, or non-AI voice system that you have. And I do think that, like, that impacts the skills and the things that you learn. But do I think that, like, the youth are, like, doomed because they don't have dictionaries anymore? That's like, I don't know. It's, I don't know. Did you have thoughts? Like, have you heard of this, like, cognitive atrophy? Like, and people worrying that we're going to lose a generation of thinkers because we're,

we're now, like, dispatching all of this, like, knowledge work. And the reason this came up is because I don't do the typing of, like, CD dot dot, you know, CD command. And then, like, I don't type. And a lot of the times now, like, now that I'm working with not Android Studio, but, like, the ADB shell commands in order to, like, build APKs and launch the emulator, I don't know how to spell any of that. stuff. I just know that the models know. I don't know about you, but like five or six or seven

years ago when I was doing a lot more day-to-day engineering, I didn't know any of that stuff either. I don't think anyone ever did. If I needed to know how to use something, I would go and do some desperate Google search and then go look on some stackover floor or some other forum to find some magical invocation that would do what I wanted. I never, like there were some things that, like, I knew, but then there were also like a lot of things where I would have to go do the research myself on how to use the command or the tool or I would forget how to do something, I'd forget how to spell some flag

or something, or I would go like, you know, I would use the tool, but then I would do like dash-dash help and then like, you know, the help was not always like, you know, not always good. And like this is like true not just for like command line tools, but just like for like anything. Like, you know, I feel like it's just like more efficient now.

I don't buy that there's like a cognitive dissonance here. One, because we're still looking at a lot of code. I mean, at least me, I'm, I, you know, go and review the code. There's not, like, I don't review everything. If it's like a substantive change to like an API or something, like, yeah, I'm going to like review every line. If it's like, I've like written some script, you know, like, hey, like, you know, add a script to, you know, check the format of these Jupyter notebooks in this repo. I'm not going to go read that script because like I don't need to go and

understand all the details of that little script. That's like not load bearing. You know, that's like something that's going to run is like a GitHub action. What I need to know is whether that script is like doing the right job or not. And so like I'll like write the tests for it. I'll write the validation for it. But I don't necessarily need to read every line of that little Python code. Whereas like if I'm making a change to like my tutorial content or to like an API and like a critical library. Like, yeah, I'm going to read that a lot more carefully. But, like, I do feel like we're reading a lot more code. There are still, there are still times where I'm writing code, too, because one technique that I often use is if I know I need to make, like, a repetitive change

in a bunch of places in the code base, and if I know exactly what that change will look like, or even if, like, I know, like, I need to make, like, a similar change, I will go and manually type the change I want. And then I'll highlight. that context, I'll put it into the chat and I'll say, hey, go like, you know, I want, for example, yesterday I was changing some of the imports, like the ordering of the imports in some of my Python notebooks. And I wanted it, like, formatted like in a certain way and in a certain order. And so I went and I did it in the way that I wanted. And then I highlighted it and I told the agent, like, here's how I want it. Go like description of the change. Go order things this way.

here's an example that I just did for you, go apply it everywhere else. And like that was a lot quicker than me because that was on the third try where I had started it off on this task, which was like to change how one thing was imported. And the first time it had done the change correctly, but I did not like how the code was structured. And so I asked it to change it. And I described how I wanted it changed. And then it didn't get it quite right the second time.

And then the third time I was about to describe for it again how I wanted it to be changed. And I was like, you know what? I'm just going to give it the example. And I found myself doing that a lot of like giving an example or a pattern to follow or like pseudocode or something. So I definitely, I think at least for me who was in more of like an engineering leadership role and was maybe prior to last year, I was only spending like 30 or 40% of my time writing code, whereas I feel like I'm spending more time now doing. software engineering work, not necessarily writing code, but driving the writing of code. I feel like that means my skills are not atrophied, but I'm actually using them more than I would have otherwise.

Yeah, there's this anecdote that I've heard a couple times on a different podcast where, like, you've got the designer, the software engineer, and the product manager, and every, like, the short version of it is every single one of them doesn't think they need the other two now. Yeah, yeah, yeah. Like the software engineer thinks they can do both designing and product. it's like, it's so true. Like, I, you know, it's basically, you know, product managers and like team leads. They used to just do what we are now doing. Like, they would delegate to employees. And now we can delegate to these LLMs. And it is, it is intoxicating. And like, I was never good at,

you know, I don't know, go learning some framework. You know, I had done a bunch of different projects. Like at one point, I wrote a, my Scrabble training program called Hookstar. And I wrote it using Python and Pi Arcade. And it was, it never really looked that beautiful. And there was a bunch of things I wanted to add to it. Like it functioned the way I wanted to function, but it is, is it a good piece of software? No. You know, but like if I had started, you know, today with, it would be like a hundred times better.

It would look nicer, you know, it would probably be sitting somewhere in some Pi Pi thing because, you know, in order to bundle something up and package it and release it, like, I'd be like, sure, well, why not do it, right? And yeah, it is, yeah, this idea that, I don't know, maybe for some folks, for some folks that aren't curious that don't want to learn or whatever, maybe this is like an easy way to just call it in every day. But now I'm just like every single thing that I've ever thought of as like an idea I can now go and implement. And I guess, yeah, maybe this is the time to pivot to the idea guy. The, my buddy just mentioned me, sent me a link the other day of the chief imagination officer at some company. and he was like they finally have created a C-level title for vibe coders.

I think we should start with the great Michael Garland quote that you gave, which is, yeah, that I'll, he's your boss, I'll let you say. Insights are ideas worth keeping. But the larger context was I was saying, I got all these ideas, you know, code doesn't even matter anymore. Like we need to put these ideas, you know, in papers or out in the world. But like the actual implementation, like everyone, like, and that's, yeah, anyway. So it's just the ideas slash insights. This is like the important stuff.

And like that's why I'm so excited about this podcast player because like I was mentioning in the two episodes ago, I don't know anybody that listens to as many podcasts or spends as much time editing. And that that means that like just myself is probably better than like a team of 10 testers because every day I'm either in the car driving somewhere testing on Android Auto or you know, on a run or doing dishes or baking. And like every single time you're doing. that like you're running into different use cases of how you want this player to work.

And I've got so many good ideas of like, I used to tweet out these like quotes and I would just make them in PowerPoint. And then I'd screenshot it and I'd tweet it out on Twitter. And it would be like a quote by, you know, Brian Cantrell from Oxide and Friends about about something. And wouldn't it be nice if during while you were listening to a podcast, you could just like hit pause right at the end of like what someone said, like hit some button that brings you to a dialogue where you can drag back a slider to where you want like the quote to start and end, then hit another button. It just turns that into a visual sound bite with like, you know, the podcast art, a little wavy

thing that, you know, like every single time I send them out on Fridays. And then also to like, you know, transcribe the quote and then automatically put that as the content of like the tweet. And then like automatically do it to blue sky, Twitter, you know, all the different socials. That I guarantee you. You can do that in an afternoon. And I just am not at the point where I want to add like sharing functionality to the podcast player.

But, but yeah, like, ideas like that, I got, I got 100 ideas. And that's a great idea, folks, because you know what? That's a flywheel idea. You start getting people to post little clips. It highlights, you know, the great podcasts that are worth listening to for other people that are interested in finding podcast content. But at the same time, it'll say, listen to on Podgod. And then it'll get people interested in the podcast app.

Anyways, what are your thoughts? What are your ideas, your grand envisions for the world? Well, so it's interesting because, like, I feel like you. have all of these like consumer facing ideas and like my ideas of like all like like either like work or just like very like system softwarey maybe that says something about like the type of person we are well you don't really listen to podcast right like I spend so much time listening to podcasts that I just like I realize that like all the podcast players suck they suck there's ads everywhere and so maybe if you know I don't know you like a lot there are

some things, like, you have, like, inspired me that there were some things where it's like, you know, like, I don't have like a, you know, I've been searching for like a good software to, like, track, track my, like, investments. Like, I've tried a couple and, like, none of them really do what I want. It's like, oh, I could probably just, like, build something that does exactly what I want. But, like, a lot of the things that I've, that I need to build are things where I can just have an AI, build me the, like, I don't need, like, an app for it. Like, I don't, I don't need to build software for it. I can just use AI to, like, do the task.

Like, you know, like, the investment thing is like a good example. Like, I could build, like, a separate, like, app or a piece of software for, like, you know, tracking my investments. Or I could just tell it, like, build me a spreadsheet. And the, if I tell it to just, like, build me a spreadsheet, like, Excel probably has all the functionality to do what I need. And, like, that's probably fine as a platform for it. Or, like, if I need, like, a lot of the tasks that I tend to give AI are, like,

like generating some form of known, some form of document in like a known format. And like the podcast one was like a really good idea because like that it's like a concrete thing that like it's not like a known format of thing. It's not like you need like it's not like there's an Excel spreadsheet equivalent for listening to podcasts. It's like you're frustrated with the podcasts at players that exist out there. But then like you also you have all these other projects of like these various like sort of web front-end things like the array box or like you always have like a dozen of these that you're

building at any given point in time and you did this even before AI. I don't know. I guess maybe I have lots of those things. One thing that I could, that I do wish I could use AI for but I've not really been able to yet is for building my slide decks because like I have my entire library of existing slides in PowerPoint. I'm very specific about the visuals and how I want the visuals of my slides to appear. And the thing that I've still find AI struggles substantially with is generating me like pictographs or like graphics for slides that look exactly like what I want. And that remains like a, that remains a serious pinpoint for me. And like that, like I know I have to spend some time this week making slides for Nvidia's conference next week, GTC. And I know that that's going to be like

a huge time consuming thing. And I wish I had AI that could help me with that. So yes, let's let's enumerate. I mean, there's stuff that I don't even talk about that my newest creation is my AI trainer pace for my running goals. Oh, and maybe this is a good time. I mean, I've kind of buried the lead. I forgot that I was supposed to tell you this. So, very exciting personal news. It's so random. It's going to be in the middle of like episode 269.

Probably should have put this at the beginning. Shea and I have very exciting news to share with Bryce and the ADSP community. We will be expecting a baby. Oh, congratulations. Do you know if it's a boy or girl yet? We do know. We got a genetic test that test for healthiness things. I'm sure Shima would, as a doctor, have explained that a lot better than I did. But the key thing is, is that all of the things it was tested for genetically were very low risk. And also a part of that comes the gender. So we are having a baby boy. Oh, congratulations. That's awesome, man. I'm so excited for you. Wow. And you're probably thinking, why was it, we were talking about AI for like an hour straight.

And then all of a sudden, Connor remembered that he's having a son is that the reason I created my AI trainer Pace, I asked him to give himself a name, is because I am trying to run a final marathon and break 2.30 before the baby comes. A final marathon. There will be no marathons. Yeah, yeah. Well, first of all, I hate marathons. They're awful. Go listen to R4 my other podcast and listen to me whine about how bad they are. But it's so funny because I created this AI trainer.

I track my diet pretty closely and also my weight. And so I gave it like five years worth of weight data and a year's worth of diet data. And then was like, build me a thing to break these goals. And then after it did like a bunch of churning, it's like great news. You got good news and bad news. Good news is it's like you're very fast. The bad news is that your diet is atrocious. Its main line was you're a 242 marathoner eating like a university student at a bake sale.

And I told Shima that and she thought it was so funny. Anyway, so I'm eating a lot healthier now thanks to my AI trainer pace. But yes, we're going to be trying to do a lot because I will be having less free time to do all my AI stuff once the baby comes. Anyways, we got the AI trainer. We've got array box that we mentioned. We got Podgod, the player and also the editor. I have a financial dashboard that I built back in like the Claude 3.5 days that sits on one of my monitors pretty regularly. It tracks like the price of invidia, the price of the U.S. Canadian dollar, and then like all my different other classes of whatever things like RSP, etc.

Because I thought one day there's a company in Canada called Wealth Simple, which I think loosely is the equivalent of Robin Hood, which like you can kind of have all your your cash, your equity, blah, blah, blah. in like one account, and then it just shows you like a summary dashboard. I'm not with well simple, though, and I didn't really want to switch. And so I was like, I bet I could build my own like summary dashboard by myself, which I did. So we've got the AI trainer. We got a ray box. We should talk about a ray box because thanks to you, I was able to improve. And that's the thing is like it wasn't even me with the ideas.

You pointed me, pointed something out. And then I was like, oh, yeah, that's a great idea. Why didn't I think of that? And then sure enough, like a couple hours later, it was implemented and done. Then there's the book that I'm hopefully going to. That's the problem now is I got so many things that I'm working on. The book is kind of stalled. I created the title page and then that was about it.

And then also, too, I've got dashboards behind array box that show like what languages people are using more frequently. Anyway, so now I need more. I've got four monitors. I need more than, I should have gotten six monitors because three of my four monitors are dedicated to dashboards that are just constantly there. And you're like, do you need to constantly look at like the Wiiwa and APL and BQN, a little like ticks, like, you know, I have a little line graph that shows when when someone submits a BQN command versus a tiny Apple command, it brings me so much joy. You have no idea, like, to see people using something that you've built.

And like, and like even just like seeing the Podgod app on my phone, like every time I open it, it brings me so much joy for like 10 different reasons because it's like one, it's an app that works with no ads. It's just like, it's the best app that I've used. And two, I built it. Like, it's so, it's so cool to use something like usefully. Not that I had never built anything useful before, but like something so integral, like, listening to podcasts and like running.

Those are two things that are like such a big part of like my week. And now that I've improved like the quality of that, it's just such a, it's so meaningful for me. I don't know if that makes sense to listeners or to you. It makes sense to me. I mean, like, it is pretty cool. Like I have been using. I have been using it myself.

And yeah, it's like pretty cool. Like, I showed it to Ramona. And she's like, wait, Connor built this. Yeah, well, technically Claude 4.6 built it. I designed it. But admittedly, you're using a bad version. Like, there's a ton of fic. I've released to myself probably like 10 newer versions of the APKs because initially, like,

the biggest issue with your version is that how do you mark something as like finished and then add it to the history? Like if you only listen to like 2% of a podcast, should that go as like a listen to podcast? Because a lot of times I very aggressively will skip topics. Like there's podcasts that I listen to. Like a great example is the daily. I listen to the daily five days a week, Monday to Friday. But I could care less for the interview on Saturday.

And usually they have some topic that maybe one out of every six I'll listen to on Sunday. And so usually I just like auto skip those. Like I just remove them. So those don't clearly shouldn't count towards. the number of episodes you've listened to, but I do want it to count towards the total percentage listened to of the podcast. So, like, on the stats,

you've got number of episodes listened to, total time, and then also percentage. And that percentage is supposed to represent what percentage of the content that this podcast puts out, regardless if you even started the episode, do you listen to? Because I want to know, like, over a month,

what podcasts am I dedicating the most time to? And it actually makes me sad. It makes me sad that ADSP only has 30, you know, 40 minute episodes, because we're way down there on listening time. Now it is my goal to release our podcast in the future, as he just mentioned that he's having a son in September. Yeah. When are you going to have time to do that, Connor? Well, my goal is by September, the pod god editor is going to be like, is going to have a button called Safe Edit, which will remove the, it'll truncate silence, it will remove all the ums and Oz. It won't clean up very well

speaking over one another, but at least getting like tightening it up and getting rid of the ums is going to be, although now I've got to edit this and probably you're not going to hear those ums, unless if I had enough wherewithal to keep those in so the list is not confused right now. But yeah, it's, yeah, I should stop Ramley. The point is, is I should have had six monitors because I got so much stuff that I'm building. I don't have enough screens to monitor all this stuff. Yeah, see, I'm still, I'm still just like on one monitor. I've never understood the multiple monitors things.

It's like too much. It's too many distractions for me, I think, would be the problem. So, so. You're on one monitor? That's a nuts, ma'am. So my setup right now, I don't know if I've talked about this here, but, you know, a lot of my ergonomic issues have gone away due to two things. One, I just worked from my bed. And as I've told people on, like, at work, like, I'll often have my.

Wait, did you? I had checked my phone because I'm supposed to have a delivery in a couple of minutes. And I said, did I just hear, like, and so I was, my brain was catching up with the words you said. Did you just say you work from your bed? Correct. Well, not from my bed, from the second bed. And, you know, I had all these ergonomic issues for many years, like starting when I was at Berkeley Lab, and then I sort of fixed him when I was there, and they continued at Nvidia. And then during the pandemic, a lot of them went away, but I was working like more than ever. But during the pandemic, I was working from home. And my apartment in California, very small, did not have a desk at home because I was big believer in separation of church and state. No, you know, I don't work from home. But then once I started working from home, I needed someplace to work from, couldn't order a desk. So I just worked from my couch and I would always be like, you know, like this on my couch, typing on my laptop.

And for the listener, Bryce just like demonstrated lying down with one arm like holding his head up. Because clearly you caught that by the silence that was there on the audio recording. But so, so when we, I continued having ergonomic issues in New York. And when we were in our first apartment, you know, I was using this desk over there. And I've tried, like, every crazy keyboard. You know, I had the keyboards that, like, went vertical. Like, I had the split keyboards, the ones that go, like, vertical. I tried mounting them to my chair.

I tried pretty much everything. And I continued to having issues. But then at some point, I guess my chair, my office chair. Okay, audio file listeners will recognize the sound of Bryce's old blue Yeti because my little road wireless microphones battery died. Anyways, so resuming the story. So when we moved to our new apartment, my desk chair was sort of on its last leg. And the last time I had an ergonomic setup that worked for me at a desk was when I was at Berklee Lab. And it was with the specific, like, making model of desk chair.

And when I went to Nvidia, I, like, sourced the same desk chair. And it was tricky because they basically don't sell them to consumers. I had to, like, buy it from, like, a, like, a research. seller who sells mainly just to companies and I had to drive to like the middle of nowhere in California to pick it up. Anyways, I've had this desk chair for a long time and like the seed is like worn out and the way that I, like my latest evolution of how I was typing at a desk was that tray that would go on my knees and then the keyboard would go on that tray. But like the tray was a little bit heavy. So maybe that was part of the issue. But long story short, if I sat in that chair

for too long, it was not comfortable. The setup was not comfortable. So I would start, you know, just like chilling on the bed that's in our second bedroom that's right next to where my desk is. And I just started doing that more and more. And I started noticing that a lot of my ergonomic issues were going away. And I think that a large part of it is that they say that like the best thing for you for your like ergonomics is to be, is to change positions a lot. And if I'm on the chair, if I'm like sitting at a desk, I'm basically not changing positions that frequently. and I tend to lean forward a lot, whereas if I'm on the bed, I will rotate between sitting up, cross-legged, sitting up with my legs out, and then lying on one side versus lying on the other side and then like some other like crazy positions.

And so I was telling some of my coworkers this when I was visiting at the office in January. And I told them like, yeah, like that's why I have my camera off a lot during meetings because if I'm lying on my side like this looking like a, you know, a cover model for some magazine. You know, I don't think that's the most professional look. So I leave the camera off. But anyways, yeah, between that and the fact that I'm not typing as much as I used to or typing in different ways than I used to. My work set up right now is I work from the laptop on the bed. And I do have technically a second monitor, which is my iPad. And the way I use it is that if I'm calling into a meeting where I have to, like, be taking notes,

I'll have the iPad join the meeting so that I can look at whatever somebody's screen sharing with me while I take the notes on the laptop. But that is the only second screen that I really use right now. For the most part, I work solely from this laptop, which is still a special Lenovo anniversary edition 25 from like eight to nine years ago. I've been waiting a while to upgrade my laptop, but we'll probably do so soon. That is wild. Wow. I don't know how you can do it. I mean, it's fine for a week if you're on site somewhere at a conference.

But especially with Cursor and the fact that usually I want like a dedicated monitor for Cursor. And also too, like I picked up the fourth monitor and it's a 32 inch monitor. And that one is really nice because you really want more real estate for Cursor because you've got your, you know, the 5. folder view on the left, you've got your code in the middle, and then you've got your agent panel on the right. And a lot of times you want, you know, you got to be watching what the trace is doing. You still want to see the code. And just, I don't know, whatever, I'm probably, it's a privileged position to have so much real estate for my cursor window. But then also, too, like a lot of the stuff that I'm doing, whether it's websites or dashboards, or even, you know,

like, I have a kernel profiling tool. Like, that's a work thing. I don't want that, like, hiding the cursor window. I want that on one of my other monitors. But I guess everyone's got their different workflow. If a single screen on a laptop in bed works for you, who am I to judge? Yeah. Yeah, I mean, I don't know. It's been working for me. I can't complain. And you know, like, if the setup works, I'm just going to keep doing it. And maybe I could go back to the desk. But I'm very happy with my current arrangement. I suspect I could go back to the desk. I suspect that it's just the, the, the, The AI is the big change in my ergonomics more than anything else. Yeah, yeah.

I was going to say, comment in the GitHub discussion, folks, if you also work from bed. I'm guessing the number of folks out there that have four monitors is much higher than the number of folks that work from bed on like a full-time basis. I think so. But be sure to check these show notes, either in your podcast app or at ADSP thepodcast.com for links to anything we mentioned in today's episode, as well as a link to a link to a link to a podcast. get-up discussion where you can leave thoughts, comments, and questions. Thanks for listening. We hope you enjoyed and have a great day. Low quality, high quantity. That is the tagline of our podcast. It's not the tagline. Our tagline is chaos with sprinkles of information.