# 🐣│cohort-voice
- Session: b7f1c22f-b6a4-4c0e-8179-6cbe5111212a
- Channel: Discord #🐣│cohort-voice
- Started: 2026-06-11T17:32:36.123Z
- Ended: 2026-06-11T18:10:15.563Z
- Participants: takekek, duckanbro, ECWireless, Kerp, Aphilos • Pharo
[00:00:00] [takekek]: Nice. Is that like wasn't there like a rate my professor site? That's what I was thinking about. Does that still exist? Yeah, let's let's send you a bunch of reviews.
[00:00:01] [Kerp]: Uh uh I get like student feedback at the end of uh every semester. Um like how's it on the course but also what's it talking to me as an instructor in the course. Um Yeah, that's yeah sort of that doesn't do so much, but there's internal I mean they don't they don't use it as much as we used to on students. Partially because the schools looked at that and they said, Oh, we should move all of this feedback shit online and they did, and so now the students engage over there. Um it's not published in the same way. I don't know, I haven't looked myself up on the my professor and they seem to like me, so there's another one called Course Forum that people have talked about before. That one there, everybody reports their grades and then it calculates the GPS so students look on there to see if your class is easy or not. Sure. Which is funny. Um Yeah, for sure. Uh Kirk around these charts. Um Adam Kerbleman Smevelman. Uh professionally these days. I am a professor at the University of Virginia's uh McIntyre School of Commerce, which is the undergraduate business school there. Uh I teach information systems, I also teach entrepreneurship generally, and then my skills of study I've learned these days, I should say, is applied AI. So I have a lot of thoughts on this. Um yeah, I mean that's just playing around with all the tools because I can't help it. And they were like, Yeah, you should call it this and you should write some papers. And then everyone will take you seriously. That's like great. By everyone they mean the other professors. Um take everyone else, so uh yeah, otherwise uh I mean I guess context that you know everyone in this room knows, but uh for anyone prefer to go or whatever. Uh also I guess I've been hanging around long enough to say OG member of Ray Guild. Uh I still not as not as OG as some of the rest, so every time I say that I'm like, yeah, there's a picture to the people. Um it's the same as any time I say an expert about something, or I'm like, yeah, but any expert also goes twenty people that are know more than them instead of you know, I immediately am like, well, I'm not an expert. But uh yeah, um I guess it's worth worth mentioning remaining context just for anyway, but you know, uh this episode
[00:00:06] [duckanbro]: Oh that's hilarious. Well uh so let me tell you a little bit about what we're doing, you know. It's like we set up these we're trying to schedule a bunch of these f protocoling fireside chats, um notes from the edge or something like that. Um it's kinda funny we're trying to say that like that. Um basically we're trying to talk to members and uh um other people to understand how they're using AI in their daily work if they're dad or designer or if they're doing something outside like a beautiful professor. Um we're not only trying to target members, we're trying to kind of expand out even a little more and find other people that might be using AI in different interesting ways. Like it could be as much as just like using a chat app to uh to ask questions, or maybe it gets as deep as like creating uh like a personal content management system and context storage and memory, you know. Um But what we found is that uh everyone's doing a little different and it's really interesting, and we actually didn't know. We don't know how everyone's using things. So it's just really casual conversations to see how people are doing this stuff and really like how they're adapting, you know. Um basically every field potentially um being totally disrupted by this stuff. As a professional, how are you like adapting to that and incorporating or branding or whatever it is. So um so yeah, that's basically it. So for the sake of the recording, you mind introducing yourself, you can just like uh send us up. That's awesome. For sure. Okay, we're gonna Yeah. So it was good. Uh so you're even talking about it like in your kind of relationships with your students, you're already having to like kind of adapt in some ways. Uh kind of interested in standing on that a little more, but also um even the second way you're doing research or the way you're you're building service or or you're uh engaging and consulting, you know, like maybe uh those are two threats to pull on there. Uh that's amazing. I love that. Yeah. Uh-huh. That's really interesting and it's almost like the way that I always use uh chat sometimes, it is it's almost like through the process of trying to learn about something or really to do something with it. I accidentally learn a lot, you know. And it's so
[00:00:26] [duckanbro chat]: What are you working on right now?,
How has AI or automation changed your workflow?,
What tools, prompts, agents, or processes are actually useful?,
What still feels broken, risky, or confusing?,
What should other builders learn from your experience? (https://discord.com/channels/684227450204323876/873247103231610931/1514683897282170891)
[00:02:29] [takekek chat]: who's your world cup choice + underdog (https://discord.com/channels/684227450204323876/873247103231610931/1514684411818410145)
[00:03:01] [Kerp]: committed to that job. Uh I'm also assistant director of student entrepreneurship. Uh like for the whole university. So I'm also involved in a whole bunch of initiatives, partially between both of those roles to like figure out how we use AI to help students like make new companies and stuff like that. Um as well as just generally and then uh also um before that I spent twenty years running my own businesses. Uh the first ten was like marketing companies, then I did a few startups and I ended up holding crypto land, which is how I or blockchain land, which is how I ended up in Ray Guild. And then uh after that was sold in house to do marketing for start uh data startups it was required and that's kind of what got into career momentum to do whatever I wanted to do next. And so I picked hang out be back in TMF, which has turned out to be a lot of fun. Because I have the flexibility to just work on whatever goofy projects I think are interesting and as long as I train this research and you need to answer questions for students or other nerds uh they let me keep the job. So it's pretty great. Yeah. Yeah, well so i in terms of like sort of where it hits the student piece, the the cool thing about being inside of the university right now is that everyone is really being forced to reckon with the reality of the emergence of the AI the tools. Like we can hide our head in the sand because like you know I'm I'm at a top university is it's competitive and the students that get in here are competitive by nature and so they are using whatever performance enhancing substances or tools they can find. And so you know they are they immediately adopted Chat GBT the second it works for helping them answer questions. And like so so we're all in the instructors have to reckon with the reality of things like just having you write or read like a like a reflection paper like doesn't work the way that I think about it anymore. Um or the way we used to think about it. And there's a there's a term they've been throwing around the last like couple of months to call them like proxy collapse which is like forcing educators to realize that we have used the like a format of assignments and things like that forever. Really as a proxy for showing that you know the knowledge gained in the course has actually happened. And like the the AI breaks the thing where it used to be too costly or whatever to be on things. And now it's so easy that like write a paper as a proxy for understanding of a topic has collapsed now and we have to like reassess how we do things. Um which for me has made it a great time to be an SA guy that's like, yeah we use the agents to do it too. And so like I keep hitchams where you
[00:03:06] [duckanbro]: That's kind of like uh you know, like as it's asking you questions, it's like, well, no, teach me, you know? Yeah. That's incredible and uh kinda cool to do. Uh uh. But why why are we doing this? Like it makes sense to be what it's not just like agreeing with everything, right? And it's like yeah. Yeah. Well Yeah, it's almost like you have to kind of like be able to it's almost like a new skill, you have to like notice like that it is just telling you what you want to hear, kind of and know how to get around that. So, uh you well, I guess you mentioned a couple things, you know, with your student relationships with your lectures, what other kind of things are you using the work is really absent. That's awesome. Yes. Oh yeah, I mean that's previously have uh open source uh if it's on GitHub or whatever, or you have a random you want to share, share it in chat. Um yes. That's cool. Right. So what what uh are you like remote tools? Are you mainly using Plac Code, it sounds like or are you using other tools or uh and stuff? Oh right. So you you were using grandma to kinda spin up fast uh and host through that, so yeah, so um uh s is a big enough because he's working on project outline right now. Yeah, let's not uh get track yet. Um so you know, it's not like s you know and the sex question is about uh maybe so if it's broken and like pain, it's not like context management was one of the one of the big problems and like it's mentioned over a whole lot. But everything's like model uh model orchestration which skills that I'm but uh not great. A lot of the local AI guys um are doing that in ways like uh they'll be into a hosted frontier model when they need it, but trying to do it not circle. Um yeah, I mean that's that's definitely a good one. Um then last point is just kinda like what do you see with the future? Like in six months we can see your profession changing, where you see how you see AI changing, like in specifically in in free profession. Like what's the Doomer take and what's the like more optimistic take, you know? Question answer you know, sorry. Yeah.
[00:06:01] [Kerp]: just get interviewed by Claude based on a a skill that I built out that will run an interview that will assess if you learn this stuff in the course. And essentially we get to bring back oral exams, except the oral exam will be run against the Claude instead of me. Because I my class size is seventy and I have three sections sometimes for a semester. I can't possibly do oral exams for that many students. But the agent's good so yeah. Like I Yeah, I I mean I say that to you know students and even other like colleagues all the time, which is like part of the paradigm shift that we're not used to is like the idea that like I use Claude Code as the example. It's easy to think of it as an engineer and then it does the engineer things. But also it's an infinitely patient tutor about the engineer things so like if you went to your senior engineer and constantly said, Yeah but why'd you do that? Like I want to understand the methodology and the and the libraries and and the blah blah blah you're you know your senior engineer is gonna roll their eyes and be like you are not I'm not I'm getting cage you have to teach you how to code. But like the models don't care. And so my working with Claude Code has caused me to learn all kinds of things that I was a little bit new but now I'm better at around you know database schemas and you know structurally coherent ways to build an apple in terms of you know, you know like what should go where and what should go this this way for security reasons, whatever it is um it's a thing that's I think that people need to learn, right? Like they need to learn that they can say if you explain your logic behind that to me like at any time I'm a I'm Yeah, right. I mean yeah Right. And it's a way to force you to not agree, right? Explain your li you know, for a while I was like what used to call it chain of thought, right? It was like explain your chain of thought. Like I'm talking about as I think past that a step, which is like once you realize you can use it to understand the thing then it it to me it's a really cool opportunity for someone like for someone like me who really learns well by doing. Right? Like I I don't do so great with uh the irony of me preparing lectures for students is like I don't consume how to do things very well via lecture. Like I think I'm a pretty good storyteller, so I think I put together pretty good lectures. But like if people just lecture at me, I I have to go do the thing before it sits. And now like having models for that combines that in a way that is really you know super like again, maybe I'm falling in the the significance trap. But like I feel like I know so much more since I started working with Claude for every second.
[00:06:06] [duckanbro]: Interesting that we took two uh uh few of the new s have startups about Yeah right. Yeah. I think we're gonna yeah we're done with what it's a little just uh quick note we did hit the third mother mark so if you had to drop anything for that I have one real quick Okay. So are you using any of the um like workflow um automation management tasks, you know things like that for the Uh-huh. Okay. That's awesome. Yeah. Yeah that's awesome man. Um I don't know if anyone else had any questions or uh comments and stuff for for jokes that are hanging here. Um people at the university when you say we have like access to Q this um do you have the other like access to uh research on JSON campaigns or yeah Yeah yeah So yeah well actually here let me uh stop the recording because I wanna um talk about something here uh so uh oh yeah shit uh yes by the way thank you for um for coming
[00:07:05] [takekek chat]: how are your students feeling about the future - whats the general sentiment? (https://discord.com/channels/684227450204323876/873247103231610931/1514685571245670590)
[00:09:01] [Kerp]: in my life than like I used to Yeah. Right. Uh yeah, well so the so the super you know interesting that's just like you know uh it's also just happen sense or like so I pivoted to a completely new profession like right at the time of you know it's actually like three point five or whatever, you know, it was working just it just started working well enough to be a thing where I was immediately like, I have no idea how to do a lesson class. Uh but this thing over here does and it's a thing that even at the 3.5 level he was really good at because there's all kinds of public stuff already in the knowledge like in the training day there about basic ass, you know, classroom experiences that are good at teaching students stuff. Which is the endless blog posts about that kind of stuff. Um and so early on it was very much me just sort of like inheriting a syllabus for a class where I know I need to cover a certain topic but I want to find a way to do it with students that's compelling. And I could do ten minutes of ideation with GT and get like a good forty five minute interactive experience for students in class. It's way better than what I would have made up'cause like I don't have the background to I don't have a toolkit to draw on for, you know, there's twenty different ways to do a lesson. That probably was compelling for students. Um so that's a way I still use it a lot like syllabus planning, like it's really good with okay I've got twenty eight classes and this many topics, and how should I split this up and you know blah blah blah. Um but yeah also the last year I'm really falling down a rabbit hole of like now I can build all of the as have a heard somebody say it this way the other day and I really appreciate it. I'm already the kind of person that sees the world as a series of like product shaped problems uh that could be solved for money. Um and so I have notebooks on notebooks of things that I'm just like if I had an engineer on staff, here's a thing I would start building. So and so I've been able to hack a pack at those like crazy. And then that started to and then that started to overlap with like so we use Canvas which is the learning management system that like all the grades and the assignments and everything live in prison. And I found out the damn open API it's already open source software and our school allows for the instructors to have open API access to it and so I'm able to generate a key and then literally talk to my can do anything in my Canvas instance. So I immediately started building a software layer that helps me run my courses uh way more easily and effectively and lets me do cool stuff and like you know some of it is as simple as just
[00:12:01] [Kerp]: managing my grades or whatever. Um but then uh building out all the cool things like a hitch tool that's not bad or it's it'll waste too much time if I start pitching the thing, but I've ended up building an entire suite of things that just make my life easier or that I was otherwise doing with like little third party tools that cost twenty five bucks a month or that the school is paying for a site license for or something like that. And instead of living in ten different SAS apps to like be able to do polls, you know, with my class in real time where the results then show up on the screen so then I can use that to facilitate a discussion. Like, you know, I mean I'm talking to a room full of engineer types, you know that's a thing that's not hard to build but like is an example of a thing that university is probably paying a hundred thousand dollars a month to have for all of its instructors at the scale of UVA. That's crazy, but like I didn't realize we had a version of a thing that does that. I just built my own and then later found out about the one I could have used based on her site license. But like Yeah. Yeah, for sure. So like that's an example but then I've been hacking my own things. Just like emergent problems. Like I started working on this thing I've been calling hyper context which is so like I have two different because of my two different jobs I have two different offices, each of which has its own computer and I have my computer's home. So like keeping context united across all of the different agents and chats that I'm working with is a giant hassle. Uh so I ended up building a thing that like it's a local client but that things to cloud that lets me dump context at the end of any workday and then take it back up like pretty easily over it. I know this is like a I know this is a problem that a lot of people are trying to solve and yeah basically. Um also sort of an archive tool. Like it it honestly started at the thing where it's just like I just want to store my like core documents about things like writing style and preferences so I don't have to keep telling the agents like or like ingesting markdown files anytime I go somewhere to like get it to write in a way that sounds like me or to understand the context of the project or or whatever. Um but uh Yeah, I've been transitioning to using cloud code and then like you know just kind of a CI CD five time that I built with clug uh code. But the last year I've mostly been using replic. Um it's got some drawbacks but it's yeah yeah exactly. It's just it I I find it to be the most like it's it's it's like it's a little more technical than like the lovables and the you know V Leroes and the whatever's um but still gives me the ease of like I can just swap up the thing and then especially when I'm just building dumb little apps for myself like you know, I don't need to go crazy with this thing has to be like perfect and you know performant. I just need it to
[00:15:01] [Kerp]: on a server so I can access it from the other computer or whatever. Um also like there's another interesting or thing that y you guys might find it was interesting that happened recently where so I saw a tweet about it literally had like three lines in it and I have screenshot and I was immediately like oh okay I could do this and it was it's a it's a local app that lets you plug in uh API keys for your LMs and then sets two agents against one another negotiating uh a contracts and I was like okay this already aligns with a bunch of the stuff I was building back in the jurist days with my legal tech and like legal engineering stuff and so I took all of the old code from all of these from these three different versions that existed across a bunch of different pivots from jurists including the like blockchain dispute resolution layer and then later a version a piece we built about documents and like uh document problem box and like hashing and stuff like that and then another layer and just kinda dumped it all in and and ended up building a new thing that like because I still had the old URLs I just slapped up on the old URL and like I don't know if I'm gonna do anything with it. But the idea that I was able to take this pile of old IT and like in forty five minutes make it like live again in a new and modern way was really a mind blowing thing to sort of just do a experience Yeah context management and and uh context management and then memory those kind of go hand in hand. Um which like on one hand I feel like I'm just kinda holding my breath because I think both of these probably end up getting solved because there's money in MR Hills, you know, for the model providers um but uh the one that I flagged when you gave me this question earlier though is like I'm I'm consistently aware that I am throwing the beefiest model had problems that can easily be solved by a lesser model. And I don't know like which one to pick when I a little bit wish it would just take for me, but I also know that every time they say it'll pick for you it doesn't do that effectively. And so I don't know yeah model but like yeah yeah. I I've heard of open route I haven't played with it yet Yeah. So Yeah. Well so I I can also kind of jump into that one through the led to the question from chat from Ted about how the student how students are feeling about the future. Um
[00:18:01] [Kerp]: Yeah, the st the students are understandably anxious. Um they you know the justification for the investment in the university is you get a better job, right? And they are currently being told that AI is gonna replace all the jobs, and so I you know I think they're I think hopefully they feel better after they take my class so they come to office hours when I explain to them like I don't think that's right. I think right now that's a popular narrative is it because it's based in value that these like companies want you to believe is you know, tied into their IPO, frankly. Um which we'll see if that figures out in the next few months. Uh but like uh, you know my my answer broadly is I think it creates more work for uh like especially engineers. Um because it's gonna replace a lot of the like I heard somebody say this recently, which is like you know, uh code generation is a task that engineers do, but like it's not actually their primary function. The primary function of of an engineer is to like understand the processes and mental models and like you know, structures and things like that that an engineer understands to be able to do their job well. And then yeah, they generate some code. So like you're still gonna need the people that understand all those pieces. And then you're still gonna need the QA people to deal with quality assurance. I should not use sacrifice without explaining them. Um like check this 10x amount of code that is gonna be that's gonna be you know, like one way I say it sometimes it's like I think there's gonna be a shift to this idea that's more like like just you know, a title like filter, right, which is some kind of like amalgamation of product manager and coder. And those builders are gonna be empowered to just skip prototype and just build features like crazy. It was gonna be delivering these like almost finished features left and right. And somewhere somewhere someone in every org is gonna have to absorb that 10x output and make sure it's like they put in the app that they are legally responsible for not breaking. You know. Um and so like I think jobs shift a lot, but I tend to be in the canon that that believes that like all of these technologies drive the need for some version of reskilling, but in the end they result in more work than they undermine. Um I think that's true in a lot of spaces. There's some interesting spaces like law firms or like marketing firms where
[00:19:56] [thekerp chat]: https://jurisproject.io/
[embed] Juris — Negotiate with confidence - A local AI negotiation client. Research preview. - https://jurisproject.io/ (https://discord.com/channels/684227450204323876/873247103231610931/1514688805901307905)
[00:20:09] [thekerp chat]: https://hypercontext.me/
[embed] Hypercontext - Universal AI Context Layer - Connect your digital identity. Power every AI interaction with personal context. - https://hypercontext.me/ (https://discord.com/channels/684227450204323876/873247103231610931/1514688857852088402)
[00:20:38] [ECWireless chat]: OpenRouter? (https://discord.com/channels/684227450204323876/873247103231610931/1514688981126873139)
[00:21:01] [Kerp]: Like maybe they do get massively disrupted and the whole model changes to things that are like sort of outcome based instead of like billable hour based, and I don't know what that disruption looks like exactly. Um Yeah, well well they're the two where I'm comfortable making proclamations like that and somebody wants to say, Back that up and I can talk about the bullshit they do all day that's not worth what they are. But like I said law terms are interesting because they're in a very interesting like cornered market space for part of what they're up against is they're wildly overpriced for what they're actually doing. But then there's this other side of that because anyway, it's a different rabbit hole, but there's a saying of like, okay, how much of that is about absorbing liability and how much of it is actually about a lot of right, like how much of it is about just having somebody who's overseen your contract, so you can say, Well, my lawyer handled it and we're gonna sue them if it didn't work out, right? Like um, it's uh yeah, kinda like my broad take on like where this stuff goes. I have I'm actually working on a book outline now uh that I'm trying to sell like tentatively called something like everything is vibes. But whatever, like I have this I have a broader broader theory that like the emergence of these tools, particularly in programming, but I think across all fields, are gonna cause the the costs of execution and and like execution both in terms of building but also just like executing a marketing campaign. So like falls of basically commodity price on, which means all that's left is people like with tastes and judgments and like access to distribution. And so it in terms of like they it used to be a significant hurdle to just say, So you have an idea? Well uh you know like i here's a way to say investors used to say, like, do you have a do you have a technical co-founder? That was one of the questions they'd ask you, right, before they throw money at your your project. Because the first thing you're gonna need to do is spend a few hundred grand, possibly like building out the thing that you just pitched us. Right, and you want to make sure that you have co-founder and look at that work and make sure it's it's worth the money and blah blah. Um that's not a barrier, if that's not a barrier to entry anymore. If everybody can do that part, just the execution piece, then like, what's left? Right? And like some of it is knowing how to execute and like what sequence to execute them. But then like some of it is just the X factor where like why is some art like more popular than other art, right? And it's and it's kinda just this like ten percent freak shit thing where somebody with an opinion plus a version of taste plus some version of judgment, like like, you know, uh says, I don't know, this is how it should be and then the world goes, uh we agree, here's money, right?
[00:21:05] [thekerp chat]: https://coursetools.uvafoundry.com/
[embed] Course Tools - Making Canvas Work for Class - Enhance your courses with peer feedback, prediction markets, surveys, AI study tools, and syllabus management. Seamlessly integrates with Canvas LMS. - https://coursetools.uvafoundry.com/ (https://discord.com/channels/684227450204323876/873247103231610931/1514689094616350730)
[00:24:01] [Kerp]: So um I think I like what I'm excited about is that I think it expands the design space. Like in a way that like we used to talk about this with blockchain a lot. Like this is it was one of Chris Dixon's like favorite with the courses was like the emergence of of of like workable digital scarcity. Just creep like widens the design space for what kind of cool like real stuff is doable. Right, like the stuff that we've lived, right? Like like DAOs, like collectively like Ray Guild, right? Collectively owns operations like this fit into that design space of stuff that wasn't as do was really as easily doable before we had the tools to you know, the crypto economic tools to make it work. Same thing here, like in a world where I can rip up this negotiation app in forty five minutes, you know, based on some previous research. Like, what does that do to opening the design space for all kinds of little things where maybe they wouldn't have gotten traction before because maybe they aren't monetizable like by themselves at a sustainable level. But it might be clustered with twenty other random things that it that a dev shot could build also at the same time. So, you know, this should be one that's um I'm sure close to the hearts of many waiters around here. You know, what the like like how many of the goofy little public good things that we've always imagined we could build that costs, you know, twenty bucks a month that we need to cluster together to have a sustainable conglomerate of you know, uh of goofy little little tools but that solve problems for people at a level that they you know like I think that I think that art was super exciting. Yeah. No, I'm good. I'm good. I got to have uh yeah. Yeah, yeah. I've done a couple of like we'll talk about a few things that would sort of qualify. One of them gets to one of my like current academic interests, which is funny'cause it overlaps with all the jurisdiction and the negotiation stuff. Uh so we at the the in my role as uh assistant director of entrepreneurship here, we oversee a a yearly uh start up competition. And so over last year we built out a platform that uses any L L M that you want to plug in to do the first round judging. Um which I presented the workflow'cause it's just a multi step, you know, process of having multiple different judges with different personas look at a uh like the pitch uh materials that are submitted by students and then score it and then it ties together a lot of scores and then gives us a report and then we use that for the first round to get like from the round of two hundred down to the twenty semi finals. And then we use the humans to do that. Uh that's downstream from a bunch of research I've done on like
[00:27:01] [Kerp]: how good LMs are at that sort of judgment. And it turns out really good. Which is cool. Um to that end I've also built out a Claude's skill that I've used for two semesters now to do all of my grading for my midterms and my finals. And so that one actually is very workflow because it's it it puts it out so it could be used by any other instructor. And so it has an onboard process where like they have to upload their answer key and then it asks them a bunch of questions about the answers so they can get justification for how it should apply the grading logic to the student uh student replies. And then it runs through a process where for the first twenty tests it checks every score against the instructor and then has the instructor explain why the instructor would have created something differently and it uses that to harden the rubric for the grading process and then it'll run the rest of the tests itself after that. Um it's super effective. Uh it's it's uh it's a workflow built up of computer vision because a lot of these tests have handwritten answers um and then yeah a bunch of like there's like 20 Python script in there for different things that are happening to double check. And like things like that. But then yeah just it's it's like data you know JSON files with all the grades and stuff in'em that I can then put into my course tools app to have a dashboard where any time somebody comes to office hours and wants to know how they did in the test I can just pop up this dashboard and be like, here's what you got wrong it's all pretty um so that's one that's you know kind of in the space of workflows. Um that I have a bunch of other skills uh happy with you keep jamming if you you know I don't want I don't want to eat anyone else this time that um I have a bunch of skills for filling out dumb forms I have to deal with every semester, like my TA requests is a pretty standard form that exists and the answers are almost always the same. So I still use browser controls is still that far in California every semester, so I don't have to deal with it. Um I have a skill that automatically puts that automatically does research on and then puts into click up any guess that I flag that I want to have on my podcast. I'm rebooting this engineering uh podcast that I used to do. Uh so that one's cool because I can just drop it somebody's name in and it does a research class on like what they're up to, puts it in the description, creates a thing for it. That's um I in theory I have access to stuff like that. I just haven't really poked at it. There's a lot of like the thing I really need to I'm gonna start working on this next year is they have a super cluster here. Um in their like research compute division that I think might make it so I can run all these tools without having to pay.
[00:30:01] [Kerp]: Which would be cool. Like it would be like running a local model or such I would have a you know, GI GPU cluster to to run it on. Um but uh I should call out one other one that's one one other thing I've been using it for, because this is maybe different from what all people here might have been using for, but like I spent a lot of time uh like pitching projects and putting together decks to try to or I should say it this way, I spent a lot of time emailing people more powerful and more busy than me to try to get traction for things that I want to do around here. And so if I've been using club design like crazy to make deaths that like they think are super impressive, but I know it are super easy to make like, you know, an easy five-page readable, you know, kind of kicked thing that I can send to somebody to try to explain, you know, whatever idea I'm working on. Um and along with that I have a skill that actually the buddy of mine made that that I actually can share if anybody wants to use it. He calls idea writer, but it's basically just a skill that like anytime you have an idea about a thing, it'll run an interview with you that'll go as long as you want it to help you flesh out like what's really behind that idea, how coherent it is, like you know, how it fits in a broader ecosystem of ideas and stuff like that. Uh I've found myself using that like crazy to flesh out things where I'm just like I am about to talk to a guy about this thing. Uh let me see if this thing I'm thinking is, you know, actually a good idea or not. And it's and and uh that has helped me get a bunch of so like I mentioned the book outline I'm working on. Uh as you might imagine from listening to me talk, I've tried like maximum AI in terms of how I'm working on this book. And I actually have another book that's we're in like final edit now is gonna be fine out about marketing and technology uh through the business press over here. But like I don't I'm still not I can't talk about it too openly because there's still some bias over here. But like I wrote that whole th Claude broke that whole thing for me. I mean not really, like I've I provided an outline, I spent hundreds of hours talking to Claude and or Chat GPT to like flesh out the ideas, so it's all still my ideas and my constructs and my you know frameworks. But like it did all the drafting. Like you did the part that would have taken me an additional thousand hours of sitting there and and to my like analogy with the engineers and code generation. It did the text generation part, which was great. I was I'm bored by that part these days. But like I I did all the idea stuff, so before you stop it, thanks for having me.