220: Embracing AI with Dan Wilson
In this week's episode, the team gets into it with special guest Dan Wilson, an AI & tech consultant that has recently created free and paid courses on using AI in daily workflows. They also dig into the ins and outs of creating Udemy courses and their viability in the tech space.
Dan's courses:
- Master ChatGPT: Build AI Assistants That Know Your Business (FREE)
- Generative AI for Developers: How To Use AI In Your Workday (Use code quackquack for a special discount)
Follow the show and be sure to join the discussion on Discord! Our website is workingcode.dev and we're @workingcode.dev on Bluesky. New episodes drop weekly on Wednesday.
And, if you're feeling the love, support us on Patreon.
With audio editing and engineering by ZCross Media.
Transcript
Spot an error? Send a pull request on GitHub.
[00:00:00] Highlight
[00:00:00] Tim: and so maybe CIOs will become more AI focused, but there might be a role where there's like a, like a, a, a Chief AI officer, you
[00:00:10] Tim: know, a CAIO.
[00:00:13] Dan: I have seen that title pop up and um, there's people selling, there's people, there's universities selling curriculum for this.
[00:00:41] Intro
[00:00:41] Adam: Okay, here we go. It is show number 220. And on today's show we have a guest. Dan Wilson is here. Some of you may know him that know Dan's club. say, hi, Dan.
[00:00:50] Dan: Hi everybody.
[00:00:51] Adam: well that was not your, your usual voice. That's all right. We'll come back. You, you'll, you'll do plenty in your normal voice anyway. So, today we've got our guest, Dan here.
[00:01:00] Adam: We're gonna talk about his new course on Udemy and, AI stuff because that's what the course is about. And that's is a law. Now, you can't have a podcast episode where you don't talk about ai.
[00:01:10] Tim: Yeah. Apparently. Or, or, or, or earnings call.
[00:01:15] Adam: but first, as usual, let's start with our triumphs and fails. Tim, it is your turn to go first, my friend.
[00:01:20] Tim's Fail: Health Scare and Personal Update
[00:01:20] Adam: What do you got going on?
[00:01:21] Tim: All right. I'm going for a fail here guys. Sorry. you know, at the end of the show I always had the little thing, you know, your heart matters.
[00:01:29] Ben: Of course.
[00:01:30] Tim: Yep. Yep. So my heart matters a lot to me.
[00:01:35] Adam: Yeah.
[00:01:36] Tim: I, I mean, almost as much as you guys' heart, but so, uh, about 10 years ago I had, uh, a cardiac ablation procedure, but my heart was running too fast and it was, you know, so it could lead to large heart syndrome.
[00:01:49] Tim: So they go and they just electric, they could cattle prod, basically shove it up your leg and goes up to your heart and, you know, reboots it,
[00:01:58] Adam: The butt bones connected to the heart
[00:02:00] Tim: exactly, you know, the song. so I just, it's been feeling weird. So I just went back to the doctor and I'm like, you know, let's, let's check this thing out again.
[00:02:09] Tim: And so I'm wearing a hot martyr right now. I'm gonna flash you guys, so I'm, oh, there we go. A little. Yeah, yeah, there you go. so I'm wearing a hot monitor, so I get a call a couple days ago. They're like, so 4:30 AM there was a something your, your heart monitor trigger alerted. What, what were you doing?
[00:02:26] Tim: And I was like, sleeping. Were you symptomatic? I'm like, I don't know. I was sleeping. I just told you that. Right. They're like, so your heart was beating 40 beats per minute, which is like, it
[00:02:37] Adam: really, really
[00:02:38] Tim: the opposite, right? So it used to be, my heart would, like when I'm sleeping would be 190 beats a minute, like
[00:02:44] Ben: my
[00:02:45] Tim: running a marathon full out.
[00:02:47] Tim: Right. Full sprinting. And, and, but now I guess they rebooted it so good. It's slow. So, but anyway, the doctor's like, you know what? He's like, as long as you don't, do you ever feel that way during the day? I'm like, no. They're like, uh, it's, they're like, we're not gonna do anything. It's, it's fine, but it's just kind of, yeah.
[00:03:03] Tim: It's kind of weird,
[00:03:04] Adam: Did you swap out for efficiency cores or.
[00:03:07] Tim: right? I, I guess so. I don't know. I don't know. So that's, that's a little worrying. 'cause I mean, if you have a very slow heartbeat that could lead to a stroke or things like that, so,
[00:03:17] Adam: Oh my goodness,
[00:03:17] Tim: you know, that's why I'm keeping, keeping track of it.
[00:03:20] Ben: But during the day, you're saying it's beating normal speed.
[00:03:24] Tim: They, they did, uh, like a ultrasound on it. They're like, it's my, my, my flow. Like the flow rate is super good, like 50. they gave me some number, like basically the amount of blood that you're pushing out should be around 50. And I'm at 65. They're like, yeah, you have really good, yeah, my heart is matters.
[00:03:41] Tim: Right. It's pumping. but anyway, that's, that, that's what I'm doing. And then, like, you know, the personal front, my, my wife's out of, not out of town, but she's up in Atlanta right now taking, uh, a bunch of teenage girls to go see, um, this k-pop band called, stray Kids
[00:03:56] Adam: Okay.
[00:03:56] Dan: You had to miss that.
[00:03:58] Tim: Yeah. I had to miss that just for you, Dan.
[00:04:00] Tim: I was like, I'm like, no, I'm sorry. I'm ba I'm like, Dan Dan's coming on the show. I really, really wanna see Stray Kids. You know, I love K-Pop.but you, you go with a bunch of teenage girls and Yeah. Deal with that. And, uh, I'll, I'll stay home and hold down the front.
[00:04:16] Dan: The sacrifices we make.
[00:04:18] Tim: I know. Yeah. But anyway, that's me. How about you, Adam?
[00:04:21] Adam's Triumph: Automation
[00:04:21] Adam: I'm gonna go with a little bit of a triumph. so I mentioned a couple of weeks ago some stuff that we were doing to improve automation and survivability, outages, that sort of thing. We came up with another one. Um, we actually, kind of recognized that we had this potential hole, uh, or like a potential point of failure.
[00:04:40] Adam: Uh, so we use one password to manage our passwords for, you know, individuals and we have like system passwords and shared vaults and stuff. And then we have. ways for our applications to load secrets from one password at runtime. Sometimes on demand. Sometimes it does, like file interpolation stuff.
[00:04:57] Adam: It's pretty neat. But, it means that it's a simple point of failure. If one password goes down, then like, if we reboot our apps, they're just not gonna come back up, which is problematic to say the least. We, we realized that and then like two weeks later they had an outage and fortunately we clocked the outage pretty quick.
[00:05:15] Adam: We, you know, we monitor their status. Uh, RSS feed. We noticed that they were having an outage, so we, you know, we were like, no deploys for the rest of the day just until we find out what's going on here. and we, we've survived that one, no problem. But if we had rebooted an app, it probably would not have come back up.
[00:05:32] Adam: and so, uh, that kinda lit a fire under me and, I figured out, I mean, it's a very simple strategy, basically like all the. All the secrets interpolation that happens at startup it, ces that to an encrypted bucket that like nobody has access to a S3 bucket. and then if during app startup, if, uh, if the interpolation that we do from one password fails, then it just goes and grabs the latest from, from S3.
[00:05:59] Adam: So it's always constantly backing up the current values to S3 and then recovering from there if necessary. And it's been, I mean, we haven't had any outages, but, you know, knock on wood here, so far so good. Been been happy with it. We're all knocking on our skulls. quick, little, little, uh, survivability improved.
[00:06:17] Ben: what's the, I mean, increasing my nines, is that what's going on here? Availability, resiliency.
[00:06:23] Adam: resiliency, that was the word I was looking for, so.
[00:06:26] Ben: I, you know what? In, in all the times that you've talked about using one password to do your key storage, I don't think I realized that it was making a call to something that, that could go down. I, I thought it was somehow always just a command line thing you were running.
[00:06:42] Adam: Hey, even the command line, you know, that's gotta access your vault and, you know, stuff, stuff can break.
[00:06:46] Ben: I, no, totally. I just, I, I don't think I ever quite connected the
[00:06:50] Adam: I mean, it, it's funny, we have like, I don't know, five, somewhere between five and 10 different ways that we use one password. We do have a command line thing that like, you know, so for example, the aws-sdk or I'm sorry, the aws-cli right? If I wanna deploy a Lambda function from my local machine, you know, if you use the aws-cli to do that, you know, your AWS Lambda publish whatever the command is.
[00:07:13] Adam: if you have environment variables set for your username and password, it'll use those, right? But I don't wanna have the environment variables set. On my system because then somebody, you know, if I leave my laptop unattended accidentally, then somebody could just go look at those. Or that could be exfiltrated by something, who knows.
[00:07:28] Adam: so instead we, the one password offers this utility, it's a command line utility where you can kind of like do, single execution environment variables, right? So instead of AWS Lambda publish, it's, uh, our, our command is OPIQ because IQ is us and OP is one password. Their, their utility is op, and then we have a wrapper that simplifies it, whatever.
[00:07:51] Adam: but uh, yeah, so it pulls secrets out of one password from a, a very specifically named place, makes them single execution environment variables and then calls the whatever command you run. So yes, we do CLI stuff, but there's like 10 different ways to, to interact with one password, with your apps and, yeah.
[00:08:10] Ben: Dan, you'll, uh, may notice at some point that Adam is by far the most adventurer of us, adventurous of all of us programming wise. He likes to live on the edge.
[00:08:21] Adam: I do.
[00:08:22] Ben: Throwing himself outta planes
[00:08:24] Tim: likes to wrap himself in a
[00:08:25] Ben: Yeah,
[00:08:25] Tim: that's covered in cold
[00:08:27] Ben: I like a good weighted blanket
[00:08:28] Tim: yeah.
[00:08:30] Ben: weighted under decades.
[00:08:32] Tim: and I just like, I just like to manage people. So
[00:08:35] Ben: Well, very cool man.
[00:08:36] Adam: All right. Well, uh, that's it for me. So what do you got going on, Ben?
[00:08:39] Ben's Triumph: Nerd Sniping
[00:08:39] Ben: I will also go with a small little triumph, which is that, uh, on Friday I had got an email someone asking a question about cold fusion, and it was a relatively innocuous question. they wanted to, they basically, they had a cold fusion query that was the result of joining several tables together.
[00:08:57] Tim: I read that this morning,
[00:08:58] Ben: And they, yeah.
[00:08:58] Ben: And they wanted, thank you. I appreciate that. Um, and they wanted a way to, to kind of co collate the data. Uh, and it just wasn't obvious to them how that should be done. So I, there, there's a couple of relatively simple things that you can do, but it's sort of just a, I I, I refer to it as nerd sniping. I didn't make that term up.
[00:09:17] Ben: That's a, that's a term people use basically when you give someone a problem and it tickles them in a, in a way and they just have to go off and try and solve it. So I went down this rabbit hole over the weekend to see if I could come up with a sort of declarative syntax to say, Hey, I'm gonna get this query back and it's gonna be all of these columns based on a join.
[00:09:37] Ben: But what I want you to return is a structure that has a structure in it, that has a structure in it, that has an array in it, kind of a thing. Uh, ba if you can think of it, almost like how I took inspiration from TypeScript, how it defines interfaces and you can say there's an interface for this object and then this array is an array of this type of interface kind of a thing.
[00:09:55] Ben: anyway, so I just, I had a lot of fun and for the first time in months, I would say I woke up in the morning. I was kind of excited. And that sounds like awful, right? I mean, I'm, it's not like I woke up with a sense of dread, but it used to be that every day I would wake up and I would, my mind would immediately return to whatever problem I had been thinking about yesterday.
[00:10:17] Ben: And you know, as I've talked about on the show before, I've been kind of in a funk, a little depressed and, but I was, this weekend I was just, I woke, I would wake up Saturday, Sunday morning, Monday morning, and immediately my brain would go back to this. And it just sort of felt like I was back in that, in that groove of,
[00:10:34] Tim: it a comeback baby. He's been here for years.
[00:10:37] Ben: just, I don't know, I felt, I felt kind of back in my element for, for a while, which was really nice and um, kind of dovetailing with the show. It never once occurred to me to say, Hey, I have this problem. I'm trying to solve it and let me see if AI can solve it for me. And I think that's really interesting. No, but I think it's really interesting because the joy. Solving this problem was the joy of thinking about it. To me it's like, this is a problem.
[00:11:05] Ben: I'm trying to find it. I, I went down a whole bunch of wrong paths. I went down, I had to like restart it a couple of times and like, that was, that was the joyous part. I mean, the fact that I got something working is obviously satisfying, but it, you know, the,
[00:11:20] Tim: you gotta work in two different ways.
[00:11:22] Ben: yeah, yeah, yeah. So, so I don't know. I, I just, it's interesting 'cause I'm, I know for me personally, getting to a place where AI is helping me with a lot of work is gonna be as much an emotional hurdle as it is a technical hurdle. 'cause there is a, a joy of artisanal, handcrafted, you know, hops rich, brooded cold fusion.
[00:11:47] Tim: Yep. Paul, Paul from Little House on the Prairie. Mm-hmm. Yeah.
[00:11:51] Dan: We should revisit that sentiment a little later. You'd be surprised how much I agree with you, Ben.
[00:11:56] Tim: Yeah.
[00:11:57] Ben: right. I'm, I'm, I'm down to clam.
[00:11:58] Tim: I, I will say that I was surprised. I read the article and I did not see, I had in my mind a way to solve it. And I, I, when you got to the CF output group by, I'm like, I never would've, I've never used that the entire time.
[00:12:12] Ben: Yo, I've literally never used it. I,
[00:12:14] Tim: I've never used it, so, so how did you, how did you come to that conclusion?
[00:12:18] Tim: How
[00:12:18] Adam: use that thing all the time.
[00:12:20] Ben: so, okay. So in Lucy, it turns out, and I discovered this as I was writing up the post, the, the reason I've never used it in cold fusion is because when I write my cold fusion templates, I have a, an open CF output at the top and a closed CF output at the bottom. And everything else is just regular template variable interpolation. When you have a CF output with a group attribute, it cannot be inside of another CF output. Otherwise it complains. So that, like, like, like 15 years ago, I was just like, all right, well then I can just never use that.
[00:12:51] Adam: I'm never using this feature. how much extra work you did over your career.
[00:12:56] Tim: I know, right.
[00:12:57] Ben: Wait, what'd you say, Dan?
[00:12:58] Dan: Once Loop had a group attribute. I never looked back.
[00:13:01] Ben: Yeah. Yes. That's, uh, that's what, that's what, uh, Tim's referring to, but, um, Lucy does not have that constraint. You can have a, a high level CF output, and apparently the group works.
[00:13:13] Dan: Are they bending space and time to make that happen?
[00:13:17] Ben: I, it's all magic to me, man.
[00:13:20] Dan: Wow.
[00:13:21] Ben: But, uh, yeah. So, I don't know. I just had a good weekend, and uh, I just felt
[00:13:26] Tim: art, yeah. The article was very enjoyable. I, I, I really, I, I learned stuff,
[00:13:30] Ben: I'm going double triumph just on the fact that anybody read it, so. Awesome. Alright, that's me. Carol, as Adam mentioned is out. So we're gonna bring it home with Dan. Dan Wilson.
[00:13:42] Dan: clock Carol as a, as a fail because she's sick. I really doubt that's a joyous moment.
[00:13:47] Ben: Yeah. Fo show.
[00:13:49] Tim: She would, she went back home to her mother died while she was out of town, and so then she went to her grave site and salt all her family and everything, so all her nieces and nephews got her sick. So little
[00:14:01] Dan: Price we pay.
[00:14:03] Dan's Triumph: Book and AI Course
[00:14:03] Dan: So I'll, I'll give a bit of a triumph. I'm, I'm putting the final polishing on a book.
[00:14:09] Tim: what?
[00:14:09] Adam: A book.
[00:14:10] Dan: know, I know. Like back in the day, the book writers were always thought of as like, top of the food chain, you know, and I was like, one day I'm gonna write a book, but I never, never really cared enough to get it going.
[00:14:21] Dan: So I talked to Luke Kilpatrick, who you guys probably know from the Adobe communities. He lives out in California and has been writing books with his daughter, the Mia King Tide series, if you're down with that, it's about ecology, marine life. It's pretty neat. And, uh, he's like, Hey, why don't you take your, your marketing course and turn it into a book?
[00:14:42] Dan: And I'm like, why don't I not do that? Because like it's a course, right? It's not a book. Like the ideas are book ideas, but it's not formatted in a way that lends itself to the perspective of a someone reading a business book. And honestly, guys, the thought of refactoring all that, just to get to like a base point, I just didn't want to do it.
[00:15:06] Dan: Now, there was a lot more I wanted to say in that course. Fun fact about Udemy free courses, they have a hard limit of less than two hours. Like mine is a minute 59 only 'cause I spent four hours whacking three minutes out of it to get it below. I'm like, I'm not doing this again. It's
[00:15:24] Adam: Trim, trim. Two seconds here and four seconds there.
[00:15:27] Dan: a few moments late. I don't if anyone
[00:15:30] Adam: Yeah.
[00:15:30] Dan: the course will see that. So anyway, I, I stuck it through chat GPT and gave it a deep research query that had it flip the perspective from course instructor to someone writing a business book. It probably got me 40% of the way to the book, but, but what's important, and Ben, this kind of calls back to what you were saying a moment ago.
[00:15:50] Dan: Like, I wouldn't have enjoyed taking material I already put out and then switching the perspective. Like that isn't fun work
[00:15:57] Ben: Yeah. Yeah.
[00:15:58] Dan: I don't want to be consistent. I don't want to do the drudgery, I wanna automate it. I, if I could write code that would make it happen, I would do it. And so like the parts of the book that are interesting to me is like the learning path and like what am I actually saying and how am I guiding the reader?
[00:16:14] Dan: Like that's what I get excited about. That's my handcrafted hops blended whatever you said. And I wouldn't wanna give that up. But some of the basic things or the repetitive things, please, please AI do it. So ideally this book is released really soon. Luke and I are meeting tomorrow to come up with a cover.
[00:16:33] Dan: And I don't know what the publishing process is, but I'm gonna have my name in lights here soon. Boyo.
[00:16:39] Tim: Oh, we knew you win, buddy.
[00:16:43] Dan: Buy one. I got kids to feed
[00:16:47] Adam: Seriously.
[00:16:48] Tim: So I, I don't quite understand. So what's, is it, it's a fiction book.
[00:16:52] Dan: non. Mine is nonfiction. Luke writes fiction books.
[00:16:56] Tim: Okay.
[00:16:57] Dan: Were you asking about what I'm doing or what Luke
[00:16:58] Tim: Yeah. What you're, no, what you're doing. So he's sort of mentoring you, but he's is his,
[00:17:03] Ben: He's taking his AI course and, and,
[00:17:06] Tim: okay.
[00:17:08] Ben: it into a book
[00:17:09] Tim: Gotcha. Gotcha. I feel, I feel like a total loser here. 'cause you know, there's four people right now on this, this show, and I'm the only one that has written a book.
[00:17:17] Adam: we'll get on it.
[00:17:18] Tim: I've written one chapter in the Learn c CF in a week.
[00:17:22] Ben: You could write a, uh, you know, more testicles than you could shake a frying
[00:17:26] Tim: That is true. I, I, I am pro, I am definitely the, the resident, the testicle cooking king.
[00:17:34] Dan: One has to question the motivation for writing the book. Honestly, had I known that I would rewrite every chapter in paragraph 12 or 14 times, I don't think I would've done it.
[00:17:43] Adam: Yeah.
[00:17:44] Dan: But like in ai, we get a lot of content for free, and it doesn't matter if it's good or not, it looks impressive until we start diving in and we're like, I can do this.
[00:17:54] Dan: It's like, oh wait, there's a lot of work that has to be done. So I've just been chipping away at it for about a month or so, but I think it's gonna be good. You
[00:18:02] Ben: So, so you're gonna, who are you publishing at? Like what service are you using to publish?
[00:18:07] Dan: probably just Amazon, Kindle. I'm gonna just kind of phone it in on that one. Luke's a big fan of this, and he says if you publish anywhere else, Amazon kind of depresses your book from showing up. And I'm like, you're not gonna fight Bezos dude. Like, you're just not gonna fight him, so I'm not gonna fight him.
[00:18:24] Ben: Yeah. I, I try, so I had, when I had written my book, I, uh, was selling it on Gum Road. I think Adam had recommended Gum Road. And then to get, to try to get a physical book, I started to look at Lulu press, lulu.com. And I, I made it as far as trying to design the cover to the specifications of the paper size and the paper stock and the bleeding margins.
[00:18:51] Ben: And I just didn't have the patience for it. I, so I was like,
[00:18:54] Adam: do you have five bucks?
[00:18:55] Ben: yo, so I, so I did go to fiverr.com and I tried to find, and every, so. Everybody who I've had found advertisements for, you know, postings for, they needed you to send the book to them as a Word document. And I had written the book in markdown and I was assembling it with cold fusion.
[00:19:18] Adam: Of course you were.
[00:19:19] Dan: there's this thing called Pandoc you can use to go from mark down to Word. It works pretty good. I
[00:19:24] Ben: Oh. So I was using Pandoc. I, I had a Pandoc docker image that was taking the, the ebook and turning it into a PDF. So I didn't know that it, I, I did the least amount of research that I could to get that step to work. I.
[00:19:40] Dan: Yeah, I wrote all of mine in markdown kind of sort of too, because at the end of the day, like word baffles me and when something is slightly tweaky and formatting it, I just don't know how to deal with it. And so in markdown it strips away a lot of the nonsense. And I find that like my whole MO with ai, with content, with everything is get rid of the fiddly bits.
[00:20:03] Dan: And word formatting to me is a fiddly bit. And so if I do it all in markdown, at least I have a consistent set of things that can be styled. And it's not gonna be like a random button I didn't know I pushed in Word. And you can never un push it.
[00:20:18] Adam: Yeah.
[00:20:18] Tim: Yeah. Yeah.
[00:20:19] Ben: well I was, I was just really surprised 'cause you, you go to five and you look at these people's advertisements and it's, you know, I've been publishing books for 15 years and I can publish in 47 different formats and I can do all this and all that. And I reached out to several and I said, Hey, I can't give it to you as a Word document, but I can give it to you.
[00:20:38] Ben: Either as a PDF that I'd like printed on my own with like a Save as print A PDF, or I can give it to you as an epub. And everybody was like, no, I can't do that. And I was
[00:20:49] Adam: Well, I mean, why? Yeah, I don't know. We don't have to get into that. That's not the top of the,
[00:20:53] Ben: Yeah, yeah, yeah, yeah,
[00:20:54] Adam: but let's, let's save that
[00:20:55] Ben: way off. We're way off.
[00:20:57] Adam: I feel like you could have just done the cover art through the, the fiber person, but
[00:21:00] Ben: didn't know. I didn't know I was so, I was so, it was like the very, you know, it was that last mile and I just didn't care anymore.
[00:21:06] Dan: the hardest mile dude. It really is. Like, and it's the least fun one 'cause it's usually the tactic you don't like.
[00:21:13] Adam: Yeah.
[00:21:13] Ben: And it was, it was a big spotlight on all the parts of the workflow that I didn't know. You know, it's like once I could do the writing, at least I can know how to write. But then once it was the mechanics of everything else, I'm like, oh,
[00:21:26] Dan: Well, like when you publish a course on Udemy, it's like, Hey, you have a course that's great. Do you have a course image? I'm like, no. Well, well you probably need one of those. I'm like, I don't do images, right. So I go to Invato, I grab a background, I put my face on there when I had to do my next thing, I use the same face, which no one can tell 'cause they don't really show up next to each other except in my admin panel.
[00:21:48] Dan: And I have like two of me facing the exact same way with the exact same question. And I'm like, for creativity, I'm gonna give myself a two, definitely a two next time what I'm gonna do? 'cause I'd actually know how to do this. I'm this good. I can flip the way I'm looking and then I'll have a whole different, you know what I mean?
[00:22:07] Dan: Like I'll be crushing it.
[00:22:08] Dan's Udemy Courses
[00:22:08] Tim: Well, that's a good topic to get into. Let's talk about, so you have a Udemy course, huh?
[00:22:12] Ben: Oh yes, please. Tell us more.
[00:22:14] Dan: just like to throw that in everywhere. Like
[00:22:16] Tim: so, so I, I'd like,
[00:22:17] Dan: course, please.
[00:22:18] Tim: so I'd like to get into it. So first, kinda wanna talk about what your course is 'cause I enjoy taking it, but then we kind of want to get to the sausage making part of Udemy.
[00:22:28] Ben: Wait, wait.
[00:22:29] Dan: However you want to go.
[00:22:30] Ben: but just to be clear, Dan has two different Udemy courses, right?
[00:22:34] Tim: Yeah, a free one and a paid
[00:22:35] Ben: Yeah, I took the
[00:22:36] Tim: I'm talking about the paid one
[00:22:37] Ben: but I have not taken the paid one yet.
[00:22:39] Tim: cheap. Well,
[00:22:40] Dan: tst.
[00:22:41] Tim: end of the show, he'll have a discount code for you. If you, if you listen to the end of the show, he'll give a, a, a code for you. So, so hang on buddy, we're there for you.
[00:22:49] Dan: Yeah. Don't worry, Ben, I got you covered my man. I gotta pay you back for that great stand mixer you gave me for my wedding.
[00:22:56] Ben: That was like 15 years ago.
[00:22:58] Dan: I still have the stand mixer and I still have the same wife, so it must have been a good luck thing I
[00:23:03] Ben: Uh,well, you're welcome.
[00:23:06] Dan: Yeah.
[00:23:06] Tim: On both counts.
[00:23:07] Dan: Courses, right? So I think like one of the things I haven't been doing lately is presenting places.
[00:23:13] Dan: And I really miss that. I think it's fun. And like my whole motivation always was a, Hey mom, look at me. I'm important, right? We all have that. But the most important part was I've learned something and I'd like you to know it too, at less pain than it cost me.
[00:23:29] Adam: Yeah.
[00:23:30] Dan: that was one of my main motivations. And if you walk through all my presentations, you'll see there's a common thread there of Ouch, that hurts when I do that.
[00:23:38] Dan: Here's how you don't do that. And so AI to me has been very interesting because I went on the same journey everyone else has gone on. Ben, you're kind of in the early skeptic stage. It sounds like some of you are a bit farther along and there's no wrong answer here. It's really just, you know, your experience and what you're willing to do, and you know what your expectations are.
[00:23:58] Dan: And as you put it, how you're holding it, you're holding it wrong. And so I put together a course. Because I've been helping a bunch of smaller businesses and the way to help a small business grow faster is help them do the thing they're not good at. And most of the businesses I work with are usually technology founders who don't know marketing or worse, they think they know marketing. And so by putting this course together, what I wanted to do was skip over the first like month or two of working with 'em and give them something where they can get reliable outputs to power their business without having to hire people or go get Fiverr people. It's pretty low overhead and it kind of took on a life of its own.
[00:24:45] Dan: taking away five jobs, is what you're doing
[00:24:47] Dan: I think I'm actually creating fiber jobs, frankly, because like what I'm doing is, yeah, I mean, this gets into my philosophy. There's a couple ways you can think about this and I'll promise you, if I could predict the future, you'd be talking to me from a Caribbean island. If you look out my window, it doesn't look very Caribbean
[00:25:05] Tim: No. Mm.
[00:25:06] Dan: But my theory on all of this is as it becomes cheaper to do things, we want more of it. Like there is no fixed amount of work that to be done and then we automate it and there's no one that's needed to do any work. And we're like, take Adam. He's a scruffy looking nerd herder according to his description.
[00:25:24] Dan: But what he actually is, is a guy with a job and a guy with a side hustle, and I think he'd like to do his side hustle if it was only the fun parts, right? And if it was the fun parts, he would have a business and he would sell stuff and he would put money in the economy and hire 10 Fiverr people to do fivey things from time to time.
[00:25:42] Dan: Because frankly, the business isn't the product. The product, there's, there's this concept of the whole product, not just the bits on the page or the buttons you push, but it's the market. It's the way you reach the market. It's your customer interactions. It's, it's a lot of things. As a builder, we don't like doing things we, we don't see as building, right?
[00:26:02] Dan: That's overhead or admin. Well, what if the machine can do the admin for you? Now you can have all the fun you want and still grow your business doing the things that a business needs without having to do it the hard way. Hire a person or chit itself, itself out. So there's gonna be more atom businesses out there that suit a need.
[00:26:24] Dan: And so in some ways, I see myself as a job creator.
[00:26:27] Ben: Well, and it's interesting because I liken this to when Docker and Kubernetes were sort of at peak hype cycle. Not to say that they've fallen out of favor, but you know, there was kind of the newness and, and I feel like there was a lot of, we don't need DevOps people anymore because now you can just spin up a container and, and I, I was always of the mindset like, well, I don't want to become a Docker expert.
[00:26:52] Ben: I want to use Docker to do my work, but I don't wanna now become a Docker expert and a Kubernetes expert on top of all the other stuff that I'm supposed to be an expert at. So I connect with the idea that by making the marketing parts easier, it doesn't necessarily mean that the people who don't do marketing are now marketing experts, but it becomes less expensive to, to get other people who have become now specialized in that at a lower barrier to entry.
[00:27:18] Dan: Right. And if you look at my generative AI course for developers, I really intentionally have built that course to be, here's how you'd use it in your day right now. And like I
[00:27:28] Tim: what I liked about
[00:27:28] Dan: look around the room, you hear people trying to sell you stuff, whatever stuff is that are gonna tell you, AI can do all this great and wonderful things, not today, but come in real soon, you better watch out.
[00:27:39] Dan: It's gonna replace all your developers, automate everything, walk on WA water, eat bullets, poop out ice cream. Like who knows what it's gonna do? Does, does that get me a quack? It should get me a quack. I want my quack.
[00:27:52] Tim: didn't say
[00:27:52] Dan: The code is quack. Quack. So come on now. I gotta get two quacks in this podcast. But anyway,
[00:27:58] Adam: gonna have to be more profane than that, sir.
[00:28:00] Dan: I can do it.
[00:28:01] Dan: Really, I've been working hard eating, right? Anyway, the, the way I kind of see this thing is the people that aren't trying to sell you stuff are trying to do a thing. Like Ben has a very high personal set of expectations for his code. We've all seen the way formats, his code, like, it's
[00:28:18] Ben: I have, I have, I have opinions. I
[00:28:20] Dan: the amount of effort one puts into formulating thinking through and administering those opinions.
[00:28:26] Dan: It's outsized. Most people don't put that much effort into
[00:28:30] Ben: I, I
[00:28:30] Dan: So I think, you know, there's the theoretical and the practical. Like I'm gonna tell you, I'm not gonna tell you, but people will tell you that AI is super capable, but it falls short. It falls short because there are limits. There are expectations.
[00:28:42] Dan: There are standards, and it is not good at all the things people think it is. So they extrapolate, well, if it continues on this rate, it'll be, but that's not how it works though. You know what I mean? Like at some point we're all gonna realize we're spending $200 in electricity to get my email formatted.
[00:29:00] Dan: Like we really have to pay that cost. And it's not just Silicon Valley Magic money making it all work. We're not gonna be doing it this way.
[00:29:08] Perpetual Learning
[00:29:08] Ben: Well, you, you know what's interesting? So when I was watching your Udemy course on the marketing stuff, I think I was taken back by how iterative the process was, especially when you were, you were taken, like here's the, the marketing materials and what does a, what does a 10 word blurb look like, and what does a 20 word blurb look like, and what does a 200 word blurb look like?
[00:29:32] Ben: And you were sort of wi whitting things down and then blowing them out again and like whitting them down. And I'm, when I was reflecting on seeing that and then reflecting on how I think about stuff, I think if I could articulate. Like the rut I have fallen into, and I don't necessarily mean rut in a negative or positive way, but I think I have, I have landed in a world where I almost prefer predictability over efficiency.
[00:30:02] Ben: So when I need to say, start a new module in an application, I will just go and copy and paste whatever previous module I worked on and then edit the things. Because to me, in my mind, like that's very predictable. I know exactly what I'm getting. I know more or less exactly the stuff I have to cut out.
[00:30:20] Ben: Whereas if I were to say, go to a command line and say like, generate this new module for me, it feels like now I gotta fight with the CLI to make sure that I know all the commands and that it's generating the right stuff. And, and I don't know, I guess I'm, I'm living in a world where predictability has maybe for historical reasons, become an outsized value in the way that I think about work.
[00:30:46] Ben: And yeah,
[00:30:47] Tim: a man needs predictability.
[00:30:49] Ben: and, and I think it's trained me to think in this sort of one and done mentality, and I sort of have to reframe my outlook on the world to be one and done. Just isn't gonna do it with these new technologies that I really do have to commit to getting my hands dirty and embracing the idea that I'm gonna make a lot of mistakes and I'm gonna find a lot of, you know, crinkly edge cases that I don't quite understand and they're gonna release new model and suddenly everything that I knew before isn't gonna make sense anymore.
[00:31:19] Ben: And I don't know, like I don't know how to embrace that world yet mentally
[00:31:23] Dan: Well, there's a chapter in front of me right now and the title is Adopt the Perpetual Learner Mindset.
[00:31:29] Ben: I'm trying.
[00:31:30] Dan: are skill stacking experiment, often, first principles, curiosity, but like Skip all that. If you have a process that works for you, then you have a process that works for
[00:31:43] Ben: But Dan, I'm gonna be left behind.
[00:31:45] Dan: you that you're doing it wrong because you're not using what they think is right. You see what I mean? Like number one, the definition of code that that's good is it works. So however you got there, right? But let me add another point here. Let we talk about ai. A lot of times we mix our words and metaphors and we don't really know what we're talking about because it's all so new. Most of the AI in my courses is large language model-based AI and large language model-based AI is non-deterministic.
[00:32:17] Dan: And so for marketing it's marvelous because marketing doesn't compile. Right, but like code does. And so a lot of times when you ask an LLM to give you a thing, it'll take the long way around because it kind of feels like it gets better scores for itself if it produces a lot of content. Like how much, how much do we correct Our junior developers when they take the long way around, like write code for readability.
[00:32:43] Dan: Be concise. Don't repeat yourself. Like the large language models don't do that. They don't want to do that. They want to be creative and exciting and like use new and exciting ways to do things well. We want predictability in our applications. It's our pager that goes off at 2:00 AM Are there still such things as pagers?
[00:33:02] Ben: Yeah, Adam. Adam got page during our last recording.
[00:33:05] Dan: There you go. And so like. I don't think you're off base at all. I think maybe you're a little early in the AI journey, but like if you take the course that you're gonna pay, I'm gonna give you a special price of 10 million. It's gonna give you like the realistic ways to use AI and not these sort of post on LinkedIn.
[00:33:23] Dan: Look how magic I can vibe code. Like I wonder if any vibe code's been in production, right? I mean, I can make a prototype sure thing and it looks great, but like Carol was saying last week, she's got security reviews, she's got DevOps, she has to work through, she has all these sort of chains of commands.
[00:33:42] Dan: Just 'cause it looks good doesn't mean it is.
[00:33:45] A Tangent on AI Hype
[00:33:45] Ben: Oh man. Sorry. If I can go off on one quick, minor, minor tangent. I was, I was listening to an interview.
[00:33:51] Tim: never stopped you before, Ben.
[00:33:52] Ben: I think this was the Change Log podcast. I know, I know some of our listeners listen to that. So you may have heard this, there was an interview with a guy who was talking about how he does all of his agent based coding and he has like five different agents running in the background at his tabs at all times.
[00:34:07] Ben: And he'll go to bed having prompted it and then he'll wake up in the morning to see what it's done. And he is talking about just like all the amazing stuff. And it sounds, it sounds both unbelievable and also incredible. And like 45 minutes into the interview there's some question about, well, like how can people start doing this at home?
[00:34:24] Ben: And he says, he's like, well you can't do this at home. You, if you program like this, it'll cost you a hundred thousand dollars a year in tokens.
[00:34:31] Adam: Yeah.
[00:34:31] Ben: Yeah. Like, oh, okay. I wish said in intro.
[00:34:37] Dan: There are limits to everything, and I think we will find the limits with AI soon enough, but you're gonna have to wade through this. Hype swamp. That's, you know, being perpetuated by people that want to sell you something, whether that's a vision or a product, or you to invest in their latest thing or click, like on their LinkedIn post.
[00:34:56] Dan: Like it is, it's embarrassing the hype. I haven't seen this amount of hype in anything since the web came out.
[00:35:02] Ben: It's very hypey. And that I was saying to, uh, the guys and Carol, I think last week, that I feel so overwhelmed by the sheer volume of perspectives and the variance in perspectives that I, I feel like my BS meter doesn't. It's, you know, it's like I'm in the middle of a, of some sort of solar flare and my compass is just all over the place.
[00:35:25] Dan: Yeah, I know what you mean, man. I wish I had advice
[00:35:28] Ben: Yeah.
[00:35:28] Dan: you there, but it is, it is. It's like, it, it's, it's like you're being assaulted constantly and at some point you just start conceding little parts of yourself and you're like, eh, it's fine. You know, whatever. Maybe Ben will start indenting his code different now.
[00:35:41] Dan: I.
[00:35:42] Tim: No, he is got, he's, he's gotta fit it on a CVS receipt. He just, he wants a big long CVS receipt with his code.
[00:35:50] Generative AI in Daily Work
[00:35:52] Tim: So, so, so, Dan, I gotta tell you, I, I am, I really enjoyed your course about generative ai, the paid one, not the free one that Ben took. 'cause he's cheap. Um,the, I'm not gonna, I, I don't want, I don't give too many spoilers.
[00:36:01] Tim: I don't give too many spoilers, but the two best for me was basically about the morning box triage. Right. And I love how you like, made it a day, like it's a day and a life. So you're going through, uh, a workday with a developer and how he uses it differently, he or she differently throughout the day. So the triage, so I started doing that, like, so many meetings and it's like, I come out of a meeting, and I think I mentioned this in the last show, it's like, you know, sometimes the meeting gets a little heated.
[00:36:29] Tim: You don't really get the points. So, but AI filters all that out. So I'll take, I'll take the transcript from teams and then pop it into chat GBT and say, you know, give me a formal meeting review and it does, and I'll send that to the entire team. Now we have our action items, we have all that filtered out.
[00:36:46] Tim: We don't worry about that. One person got upset because they said security was bad. Right? And then, and then the other one was fixing agent code. And I love that you had, I love that you had an example that, that we are doing right now, having to go through a really old 20 something year old code base that's been on cold fusion since version two.
[00:37:06] Ben: Oh my
[00:37:07] Tim: Uh, and, and SQL injection is, you know, gotta make sure that that's thing. I'm like, why are we, why are we hiring? People to like a, a company to come in and like parameterize all our SQL queries, we'll just have AI do it. And so the entire week I've been using a little bit of AI for the stuff that it's good at.
[00:37:25] Tim: And so those two things, just being able to recap a meeting well and then being able to just take very simple, Hey, make this, you know, parameterize these, these queries so they're not SQL injection vulnerable. And so those two things are awesome.
[00:37:41] The Role of AI in Coding
[00:37:41] Dan: Yeah, I think I intentionally went through multiple LLMs for that one to show like they all have their own little take on things, didn't I?
[00:37:49] Adam: Yeah, Gemini and chat
[00:37:51] Ben: you, you used a phrase there, Tim though, that I wasn't familiar with, you said, Mor morning box is.
[00:37:56] Tim: The morning inbox triage is the
[00:37:59] Ben: Oh, inbox. Gotcha.
[00:38:00] Tim: I'm sorry if I, I slipped my words
[00:38:02] Ben: No worries.
[00:38:03] Tim: Yeah, so you get an email in the morning and it's like you don't really understand what the boss is saying, and so you pop it into chat, GPT
[00:38:09] Tim: and it helps like summarize the email in a way that makes sense.
[00:38:11] Tim: Well, I'm doing that with like the meetings that we constantly have,
[00:38:15] Dan: Well, you wanna hear something funny? So last week, a certain person named Adam Tuttle said, why would you ever have to tell chat GPT that it has to be a great software developer? Why would you ever tell it to be a garbage software developer?
[00:38:29] Adam: hate that guy.
[00:38:30] Dan: Yeah, he, he, he has some zingers. Well, I'll give you the answer why that is so.
[00:38:34] Dan: When I wanted to make a garbage email, I basically fed in the tasks I wanted it to do and said, you know, act like a hared distracted boss and give me a long-winded, confusing email that I can then feed into an LLM and have an extracted task list because I'm not making all that up, dude. Like what? Like this is an example.
[00:38:54] Dan: I'm not, this is not a masterpiece. And I just, I left it as it was. I thought I did a great job. So that's why you might want to tell it to be, you know, incompetent because sometimes you need incompetent.
[00:39:07] Adam: sure. Yeah. Yeah. If you're generating bad data or something. I guess my point in that statement was more like, why should I have to tell you you're an expert, right? If, if it's not otherwise specified, you should be,
[00:39:19] Dan: Yeah, I
[00:39:20] Adam: s tier.
[00:39:21] Dan: I'm not sure, right, like if we're, if we say too much to it. But I think, like, the way I understand LLMs is they've been trained on about everything that's ever been written. And they have context as well of what's good and what's bad. So if you go to Stack Overflow, you can find plenty of examples of bad code.
[00:39:38] Adam: Yeah.
[00:39:39] Dan: And so I think by telling it to be an expert coder, you're, you're narrowing its whole universe into coding good expert coding. You know what I mean? And so it,
[00:39:50] Dan: it's gonna make less mistakes for you because it, it really needs to boil down its context into what you want and how you want it.
[00:39:58] Tim: don't gimme the code that was broke. Gimme the, gimme The guy who gave the check mark answer,
[00:40:02] Adam: Yeah.
[00:40:03] Dan: Yeah. 'cause I mean, look, I mean, it's gonna give you an answer. It will, it will give you an answer. It just might not be the one you want.
[00:40:10] AI vs. Google for Problem Solving
[00:40:10] Ben: Let me ask you a question, Dan, because you clearly have embraced this stuff, and one of the things that I hear a lot of people say on podcasts is that they either have entirely stopped using Google or they just rarely use Google. Do you find that to be the case for you? Are you basically in the LLMs all day?
[00:40:31] Dan: You know, I'm not sure how I feel about that. Um, let me kind of talk through it. So I was at my father-in-law's and he says, Hey, we broke a bike. I'm like, oh, what? Well, what did you break? He goes, I think we broke the frame. I'm like, oh, that's
[00:40:44] Ben: not good.
[00:40:45] Dan: You're like
[00:40:46] Ben: That's kind of an important part.
[00:40:47] Dan: right? But let's go out and look at it.
[00:40:49] Dan: 'cause all I'm doing is watching a hockey game, looking at my phone. So we'd go out and we look at this thing and I don't know a lot about fancy bikes, but I have derailers and all that. I just ride 'em. And so I go to look for some video that talks about. Broken derailers and I find a video and I start watching it and the guys tell me about his mom and his grandparents and like, he's not really answering my question.
[00:41:12] Dan: So five minutes in, I go find another video and, and I do this for like 15 minutes. Then I go to Google and I'm typing in stuff and like you can guess how this is going. Once again, I'm waiting through content looking for something very specific. I've been doing this for 25 years. This, this is what programming used to be, remember?
[00:41:31] Dan: So I finally got pissed off at it. Did I get a duck yet?
[00:41:34] Tim: No,
[00:41:35] Dan: You guys are not regulated by the FCC.
[00:41:37] Dan: Are you? Okay? All right, all right, all right, let me warm up here. So anyway, I go to chat GPT and I said, Hey, we think we broke the frame on this bike by the rear derailer. And he goes, oh, that would be terrible, but you know what?
[00:41:51] Dan: There's a derailer bracket. Are you sure you didn't break that? I'm like, didn't know there was such a thing. So I'm like talking to, we're trying to get the chain off one bike that's not the broken one has a master link. Oh no, that works. This other one does it. I'm like, maybe we can't find it. chat. How do I get a chain off if I can't find the master?
[00:42:09] Dan: He goes, well, it might not have one. You need a chain tool. It looks like this. I'm like, he has one of those. So we get that. We take the chain off, we take the Derailer thing off, and you would, you believe it is the derailer bracket? I didn't know that such a thing existed. So then I'm like, okay, let's get another one.
[00:42:26] Dan: I said, it's this kind of bike. I need a derailer bracket. How many can there be? There's a thousand of these things. There's a site called Derailers slash whatever, and it had all thousand, and they're all exactly alike except for the most minute distant. The little difference you can't even tell. So I'm like, well, I don't know what to do here, so I take a picture of it.
[00:42:49] Dan: Now I want you to imagine this in your mind. There's this little piece of metal. On a brown granite countertop. So it looks like I have a camouflaged background, right? It's the worst possible picture ever. I said, well, here's the derailer bracket. Which one is it? He gave me the exact model number, multiple places I could buy it
[00:43:09] Ben: This is in chat, bt, or this is in like a Google. Okay.
[00:43:13] Dan: And then I said, okay, well we wanna replace the derailer 'cause it's bent. Uh, I looked on Amazon for the exact part number, it's like 90 bucks and they're out of stock. I'm like, so I asked, he goes, oh, that was, you know, there's an improved one that you can get and it has these upgrades and, and I take that part number and I go to Amazon and it's like 22 bucks.
[00:43:33] Dan: And I'm like, my problems are solved. And I don't know, I mean, I would've got there through Google, like I've solved every problem in my life with it. But it's sometimes you have to wade through an awful lot of garbage. Either someone's not answering your question or they're trying to rank for keywords.
[00:43:52] Dan: They just wanna tell a story. And I don't want a story. I don't care about your keywords. I want this bike to work. And so I find that that a more AI directed workflow gets me to my specific answer faster. But I use Google a lot still.
[00:44:08] Tim: So that's interesting. You, you brought up something I've never thought about. I don't even think I've read about it. So the key words, if AI is smart enough to figure out like what the actual content is versus what the SEO or whatever, I mean, that really could change the way people publish things on the web.
[00:44:27] Dan: Oh, yeah, yeah. There's actually, you've heard of, um, robots, TXT that says Stay away, or, okay, so there's now LLM txt that sort of helps LLMs know what you're doing and where to look for it and all that. And so you're gonna start to hear about this more and more as some folks try to optimize for SEO and other folks defend against it
[00:44:49] Tim: Hmm.
[00:44:49] Dan: where they don't want their sight scraped.
[00:44:51] Tim: Because people spend so much time. Anguishing and handwringing over SEO and trying to get those hits. All they care about is those hits that you know, and hopefully a small percentage of them turns into action. But it's like, at the same time, if you're a person who's looking for something very specific, you don't wanna wade through all those possible hits.
[00:45:10] Tim: You want the actual answer like you were doing with, with this bike part.
[00:45:13] Dan: Yeah. And I don't know how to optimize and make the LLMs give you your answer. It's funny because there's one camp that's like, don't use my data. There's another camp that's like, please use my data. And so it's so early in all of this like, and the people who train these LLMs, they confess all the time.
[00:45:29] Dan: They really don't know how they work or why they're making decisions the way they are.
[00:45:33] Ben: I'm finding that, uh, just going back to the Google stuff for a second. Because I, I use chat, GPT and I use Google, and I'm still quite a heavy Google user. And I think what I'm finding is the chat, GPT, feels like I get a good answer when it is a very scoped question with a very, provable answer. And if I want to have more of a philosophical question answered with Chat GPT that I find, it just doesn't, I, I I, I don't know, like, again, if I'm just holding it wrong or, or what, and like, that's the kind of stuff that I'll still Google for.
[00:46:14] Ben: And I'm a, I'm a freak, in this aspect that I will actually go through several pages of Google search results, because sometimes I want to go through like 40 or 50 or 60 links and not necessarily read them all, but just see if any of them catch my eye. You know, I'm looking for this combination of words and suddenly, here's a.
[00:46:33] Ben: An article with a title that doesn't necessarily make sense and, and like, oh, okay, maybe they're to coming at this from a totally different way that I didn't even know to ask the question in that, in that regard. And I still feel like Google, the kind of serendipitous experience of Google still feels in, in some ways better than the Chat GPT experience.
[00:46:55] Dan: Well, in product design we've always had these different user arche types, and one classic one is like browsers versus searchers. And when you're doing that sort of scrolling through multiple pages, you're being a browser at that moment.
[00:47:09] Ben: Yeah.
[00:47:10] Dan: And there's nothing wrong with that. There's no judgment to be had.
[00:47:12] Dan: It just, it's the, the way you're choosing to educate yourself. And it sounds like you're doing sort of a survey of what's out there. I'm not asking for the definitive answer of two plus two. So I mean, once again, I feel like you're very defensive in your use of ai. I
[00:47:26] Ben: I'm very defensive, Dan, because I don't, because I don't wanna be left behind. Everybody is telling me I'm gonna be left behind.
[00:47:33] Dan: It's very okay to be judicious about your use of ai. It's very okay not to use it at all. If you want to handcraft, I can say for me, like I've had a number of like Adam Tuttle type business ideas where I'm like, I know a great application I would build. But then I think to myself, I'm like, well now I've gotta like start a database and like write the login scripts and like do all this ceremony before I get to the part I care about.
[00:47:58] Dan: And honestly, I don't wanna do it. Like this is a four hour thing for me, not a two week thing for me. So I'll just pass. And that's why I once had like 62 or 63 domain names that all could have been billion dollar ideas. I'm pretty sure.
[00:48:11] Ben: Yeah.
[00:48:13] Adam: How many are you down to now?
[00:48:14] Dan: I have just a few. I have five, like I was leaning over the toilet throwing up saying I will never buy late night domain names ever again.
[00:48:24] Smart Devices vs LLMs
[00:48:24] Adam: so I wanna go back to something you'd said earlier. We talked, I forget who asked you about like Google versus LLMs, and getting answers to your questions. That question reminded me a lot of asking the smart home speakers questions. You know, I, we've, we've had a couple over the years and my experience with them, like 90% of the time or more, is like, you ask it a question and either you get like a nothing answer that that's unhelpful or you get a.
[00:48:52] Adam: A five minute diatribe and you're just like, shut up. Shut up. You already gave me the answer. Shut up. Like, why? Why are, why is that so bad at giving, giving me the answer? But then somehow the LLMI can get, I can get it to give me a one sentence response that just answers my question. Hmm.
[00:49:24] Dan: you know, to be clear, if we're talking about LLMs, what you're really talking about is a lot of statistics, pattern matching and like a corpus of, of content that it.
[00:49:33] Dan: Quote understands it doesn't understand anything It, it's math, basically. Math and triangles is what it's
[00:49:41] Ben: That's the thing where I keep coming back to this idea. People are predicting artificial general intelligence, a GI within like the next five years or something, and, and I just,
[00:49:53] Dan: to you by a company selling a GI.
[00:49:55] Ben: yeah, like there's, there's zero evidence to me that any of that is even anywhere close to being a possibility.
[00:50:02] Tim: I mean, this is the same people that says actual, you know, energy cold Fusion is just the next 10 years away. Uh, it's been for the past 70 years. So let's get to the sausage matching
[00:50:13] Tim: part
[00:50:13] Ben: that sausage, Dan.
[00:50:15] Tim: Let's make that sausage.
[00:50:16] Creating and Selling Online Courses
[00:50:16] Tim: So, so give us, I mean, so making a course, I mean, how much work is it, how much, I mean, what's your goal?
[00:50:23] Tim: Is it a moneymaking goal? Is it just like a name branding goal? Is it like, I noticed your company name or the company where, I don't dunno if it's the company you work for, if your company, but, uh, we just want a little bit about, you know, what, what if we were thinking about doing a Udemy course, you know, are we gonna get rich and retire off of it?
[00:50:40] Tim: Or, you know, is it worth it?
[00:50:42] Dan: Yeah, so my first itch to scratch was a couple of things. One, I thought, can I make a course? I don't know. I think I can, I think I can dunk a
[00:50:52] Adam: it looks so easy from the outside.
[00:50:54] Tim: Mm-hmm.
[00:50:54] Dan: easy. Let me tell you. Uh, the second thing is, like I said earlier, the, the kind of companies I help need this information, they need to figure out like, who are they, who are they selling to, what do they say about themselves?
[00:51:08] Dan: And I can do those exercises with them, but they're gonna pay me a lot of money to do something that really, I'm just, I'm just being a therapist at that point. Pulling information out of them and helping them frame it. And the real work that I can do for them that's worth money is helping them build the business, find the customers, refactor their products.
[00:51:28] Dan: And that's why they should pay me. They shouldn't pay me for all this preliminary stuff. And so if I can help a person, like say in Adam's point, spend a couple of months on his own and get this thing to where now he has real problems, let's work on those. So that's why I built the course. And then, you know, I kind of like, I'm, I definitely like to ready, aim, fire.
[00:51:51] Dan: A lot of times, but this was really ready, fire, aim. Like I say, it was really surprising to me how restrictive they are with these courses. Would you believe I can't even message my free course people that I have a paid course, even though the paid course is on demy, it's
[00:52:07] Adam: I, I believe it. It's, it shouldn't be that way, but I believe
[00:52:10] Ben: Yeah. That's weird.
[00:52:11] Adam: don't own your audience there. Yeah.
[00:52:13] Dan: You know, and then I got so annoyed them, I'm like, I'll just make it a paid course. But then I thought, well the reason why we're doing this and blah, blah, blah, blah, blah. So if you're gonna make a course, you have to start with number one, what are your motivations? Because like all projects, the first 80% is fun, and the second 80% kind of sucks.
[00:52:31] Tim: Second 80%.
[00:52:31] Dan: Yeah. The second thing is you gotta ask yourself, are you a do it yourself person or are you gonna get someone to do the parts you don't have or don't want to do? And the most arduous part is the video editing.
[00:52:43] Ben: Yeah.
[00:52:44] Dan: I obsess over that a lot because I feel like if you're gonna give your attention to me, even in a free course, I.
[00:52:50] Dan: I'm not gonna waste it. And so I go over and over and cut and trim and you notice how my chat GPT is so much faster than yours.
[00:53:01] Tim: You
[00:53:01] Dan: I'm nine xing some of that stuff for you. If I cough, I take it out. You know what I mean? Now, maybe you're not gonna be as specific as I'm, I'm a little Ben on this thing, and there's ways to do it a lot faster, but if you don't know basic video editing, it's gonna hurt you
[00:53:17] Tim: Yeah.
[00:53:17] Dan: even if you have a video editor, you know what I mean?
[00:53:20] Dan: You need to at least be able to cut and trim and assemble.that's probably the skill that I think you'd have the most trouble with unless you already have that skill.
[00:53:31] Dan: You guys have all done presentations before. You know how to come up with a story and a message and content, and that's frankly how I built my first and second course was I'll present this and then break it up into chunks and then record the chunks.
[00:53:46] Ben: Does Udemy have a video editor recorder that they either provide or recommend or it's You can, as long as you can generate an MPG file, not MPG, uh, MP four or whatever.
[00:53:58] Dan: So I use Camtasia for my courses because like I use Da Vinci Resolve as a video editor. Do you guys know that one?
[00:54:06] Ben: I know Camtasia, that's
[00:54:07] Adam: I, yeah, I'm familiar with Da Vinci.
[00:54:09] Dan: Yeah, it does everything. But the problem is it has no opinions. It does nothing like out of the box, whereas it, Camtasia doesn't do a lot, but what it does, it's optimized for speed.
[00:54:22] Dan: And so I was like, you know what I, I could do this in Da Vinci, but even something like putting a call out on the screen or an arrow like in Da Vinci, you've gotta go on the fusion page and make an arrow and then. Dude, I don't wanna do that. I don't. I do not wanna do that. I don't. So 250 bucks to the Camtasia people and I, Hey, you can make an arrow like we've solved all the hard problems in life.
[00:54:44] Dan: So I would caution you, if you're gonna make a course, figure out how you're gonna do video editing first, and then see if you can make that part faster, because that it feels like that takes the longest,
[00:54:57] Adam: Interesting.
[00:54:58] Tim: So let me ask you, so your course, your paid course is how many hours or
[00:55:06] Dan: well, it's two hours and four minutes.
[00:55:08] Tim: two hours and four minutes. It went by fast. It was a great course, but
[00:55:11] Dan: That's a compliment.
[00:55:12] Tim: so how long, how. From start to finish, how long did it take you? Record and edit all of it.
[00:55:18] Dan: Well, first I did it as a presentation. And gave the presentation and that was probably a good first draft. And then I didn't like a section and I felt like I needed to add some more sections and so I added a bunch in. So I can't really answer that from a linear perspective, but I would say it was probably a heads down month where that was my, my 90% project that I
[00:55:41] Tim: So eight, like eight hours a day, five days a week. A month?
[00:55:45] Dan: I don't know how to answer that 'cause that's not how I work. I tend to work in bursts.there was a point for like, and I'm not counting this in the month part, but there was a point where I didn't do anything on it for 10 days. I just needed to kind of like let it all,
[00:56:01] Adam: Yeah, let it breathe.
[00:56:03] Dan: and then see, well, okay, what do I have here?
[00:56:06] Dan: What, what do, what do I want someone to learn from this? Like, do I want someone to look at vibe coding? Like what is the appropriate, what can I stand behind? And show off vibe coding. I don't want to be the hype man of ai. I don't like my job is to help you become better at what you want to become better at, not to sell you a vision that doesn't work. So I think coming up with that section was probably a week on its own of kind of goofing around and trying stuff and like just, what did I want to say? Right? Because once you record it, it's recorded, you know? And changing it is, it's so much easier to change a blog post than a
[00:56:47] Adam: is
[00:56:48] Ben: Yeah.
[00:56:48] Dan: published video. Did I mention the video editing part is the hardest
[00:56:52] Tim: Yeah, I'd imagine.
[00:56:54] Dan: one of us.
[00:56:54] Dan: I won't point any fingers. Went to clean up a bunch of old files in Camtasia. I cleaned up like 200 gigs last Friday. I also deleted my Camtasia project files and they're gone.
[00:57:04] Ben: Oh no.
[00:57:05] Dan: if I have to make any edits to the free course, that's a rerecord, but guess what? There won't be any edits to the free course.
[00:57:13] Tim: It'll be a whole new course.
[00:57:15] Dan: Hey.
[00:57:16] Dan: version two. This one costs a dollar. think I'm off track with your original question. You wanna ask it again?
[00:57:22] Tim: Just sort of ask like, what is your return on this? Is it monetary? Is it, is it name recognition? Is it.
[00:57:32] Dan: The developer course. I wanted to see, could I sell a course? Jury's out on that. Most Udemy courses will make about a thousand bucks during their lifespan. There are a few outliers that'll really sell a lot. It's a function of, and this is like, this gets back to product marketing. It's a function of your audience that you have and the audience that you can build, plus what you can tell that audience, right?
[00:57:56] Dan: So like, think of Adam's business. He's got a, he has a very finite market. It's any, any skydiving place
[00:58:04] Adam: Right,
[00:58:05] Dan: that want some kind of system to, to manage what they do. That's beautiful, right? You could make those phone calls today. What's the market for a developer course? Like do I start calling developers? Hey, you want a little of this?
[00:58:20] Dan: You can't do that. So what do I do? Like I have a LinkedIn audience. There's about, don't know, four or 5,000 people that listen to my nonsense on there. That's an audience. You guys have at least 50,000 people listening to this.
[00:58:33] Tim: Oh, come on. More than that. More than that.
[00:58:36] Dan: So here's the other trick about Udemy or a a company that sounds like ude me.
[00:58:41] Dan: I don't know if I'm allowed to say this or not, but if I send traffic to buy that course, I will get 97% of the revenue
[00:58:51] Tim: Really,
[00:58:51] Dan: if
[00:58:52] Tim: right?
[00:58:53] Dan: they purchase
[00:58:54] Tim: if, but if I find it on my own.
[00:58:56] Dan: If they purchase in 24 hours in one minute, it backs down to the standard 37% of
[00:59:02] Adam: Oof.
[00:59:03] Ben: How interesting.
[00:59:03] Tim: What,
[00:59:04] Adam: That's a huge cut in your, your cut.
[00:59:08] Dan: And why do you think that is? Because if I send the traffic, they credit me with owning the audience. And if they send the person they credit themselves with owning the audience.
[00:59:19] Tim: right? Hmm. So they're re, they're really trying to make the content creators become drivers to their site. So they're incentivizing. They're like, because maybe you'll like, have a good message and someone will come to your site and like, see your course. And like, I don't want that, but there's something else here.
[00:59:38] Tim: So they're like, we really want you to push people to you, Udemy. But if they're doing the work, they're like, nah, you don't get it, no cookie for you.
[00:59:47] Dan: I mean, you know, look, when I was a developer, I thought marketing people were idiots. I thought audience magically appeared like building an audience is hard. It's very valuable. And
[00:59:56] Tim: you, you know, I, I used to work in sales and marketing for several years and it, it's a tough job. It's, it's voodoo.
[01:00:03] Dan: Yeah, so I don't fault him for it. I mean, I knew the game before I signed up. You know, once again, I, I'm not gonna pay my mortgage on courses, you know, but I think once again, I've got something to say. I want to find ways to say it, you know, waiting to get accepted to speak somewhere. Well, that's someone else's control, fine.
[01:00:23] Dan: But what can I do that's in my control and publishing a course or putting a book on Amazon that's within my control? Let's see what happens. Most business books, they don't sell a thousand copies. Turns out Udemy courses don't either, especially at perishable topic like generative AI for developers like that changed the minute I published it. It's probably already,
[01:00:49] Tim: in 2025, it's gonna be completely different in the year.
[01:00:53] Dan: Once again, I walked into it with a, with a mindset that that was okay.
[01:00:58] Adam: So this is for you. It's more like, uh, putting in the reps so that you understand the process and the, if you have another idea for another course, it'll help you better understand whether or not to bite that off.
[01:01:10] Dan: Yeah. My next course is gonna be another businessy one For people, much like Adam, that have this kind of like solo practitioner business and need to get to the next level, there's, there's a quite a bit more that needs to happen than I couldn't fit in the two hours.
[01:01:25] The Future of AI in Business
[01:01:25] Tim: So I don't know if this is too fringe to talk about, but we can cut this if need be.so it seems like you've mentioned several times that you're focusing sort of on the small business side, but I'll tell you, I mean, you know. The company that I work for is a huge multinational corporation. They are very much looking for any sort of people that have good advice on how to deal with AI right now.
[01:01:52] Tim: So we've done a complete inventory throughout all of our companies and Constellation software of, of how we're currently using ai. And then they kind of restricted it to certain AI processes, or at least AI companies. They don't want us using DeepMind. So they have, we have an approved list, but we're also very much in the review of how, how can we make use AI to make things more, efficient.
[01:02:18] Tim: So, I mean, I, I applaud your, uh, helping small teams, but it's like there's a big market for, for big companies right now to have some sort of consulting to come in to say, you know, give, just give people a, A health check on. What are you using it for now? Why aren't you using this and you know, when this happens, how are you gonna react?
[01:02:43] Dan: I mean, I give stuff away to small companies, but I charge heavily to larger companies
[01:02:48] Tim: Good answer. Perfect answer.
[01:02:50] Dan: more than happy to speak with Constellation. If they, you know what I find a lot is that their first and second option is to go to Accenture, and then eventually they realize that Accenture has sold them billables and hasn't given them the outcomes they want, and then they're ready to take
[01:03:05] Tim: Or Gartner.
[01:03:06] Dan: me. Yeah. Or Gartner. But I mean, sometimes those firms nail it too. You know? It's a big world out there.
[01:03:12] Tim: Yeah. It's just, I, I sort of see a possibility that there could become a role within sort of corporate culture. There used to be a time when there was no CIO inside of a company, right?and so maybe CIOs will become more AI focused, but there might be a role where there's like a, like a, a, a Chief AI officer, you
[01:03:36] Tim: know, a chow.
[01:03:39] Dan: I have seen that title pop up and um, there's people selling, there's people, there's universities selling curriculum for this.
[01:03:48] Tim: Yeah. Really.
[01:03:49] Dan: I just don't know what the market uptake is yet.
[01:03:52] Ben: But this goes back to my point earlier in the show about when Kubernetes and Docker came out, and I didn't want to become an expert in that. I wanted to have the, the existing platform team do deep dives and become the domain experts so that they could allow the organization to build up around those technologies.
[01:04:10] Ben: I, I
[01:04:11] Tim: but I mean, come on Ben. I don't think that's a fair comparison because I mean, Kubernetes, that's a very specific narrow tool. This tool, LMS and AI are much more wide.
[01:04:22] Ben: No, no, no. I, I,
[01:04:24] Tim: pretty much anyone can use.
[01:04:26] Ben: I agree with that. I agree with your sentiment there, but I'm just saying that there, there's such a specialized, it's a.
[01:04:35] Dan: I think I know what you're saying. Do you want to keep talking through it? Do you want me to
[01:04:38] Ben: Well just, okay. It's like, uh, it's like when serverless architectures first became available and it was like, oh, you don't need people to manage your servers anymore. You can just hit a button and now you have infinite scale. And then we spent 15 years hearing horror stories about people's billing and things going down.
[01:04:56] Ben: And now there's more complexity and resiliency is degraded. And microservices that were actually just distributed monoliths, like just because you can click a button and do something doesn't mean that, that you have the domain expertise or you have the time to build the domain expertise to really leverage those technologies properly.
[01:05:15] Ben: And I feel like AI in some ways is the same thing that yes, you could, you know, become the expert on. Code agents and figure out how to build in all of the context rules and like the priming of the models and you know, all the various, I don't know, I don't even know the right words to say honestly. but like if that was someone's full-time job, I have to imagine it would be more beneficial to the organization than having, you know, every individual engineer trying to figure out on their own and coming up with different ideas on how to do stuff.
[01:05:53] Dan: that, yeah, that makes sense. And like you guys are coming at it from very interesting perspectives. I think the amount of AI a company should want to buy is zero. Now, they
[01:06:04] Dan: may
[01:06:04] Tim: expect that.
[01:06:05] Dan: they may want an outcome that is maybe accelerated with ai and that's what they wanna buy, but they shouldn't be buying ai. That's hype. I mean, so where you're getting at Ben is kind of like, you kind of want the outcomes of ai, but you don't want to have to deal with like the blood and pain of making it work yourself.
[01:06:26] Dan: You want to be a consumer of the outputs, not necessarily the person that's making the ingredients. Right? That's a very reasonable thing to want, and I think the market is so immature right now. You're looking at people that don't necessarily mind doing the hard work to figure all this stuff out, but that's not where it's supposed to go.
[01:06:44] Dan: And there's this whole world of agents that are supposed to allow you to tuck in
[01:06:47] Tim: Right.
[01:06:48] Dan: AI into what, you know, give me an input and I'll give you an output. And that's, it's just so early right now that you got people like me and you who are, you know, banging away with prompts and inspecting the work because it's so new, you know?
[01:07:02] Dan: But I'm gonna guess if you fast forward long enough, you're just gonna be receiving the outcomes that you want. You won't even know the AI is there. It's just gonna be like electricity. I don't ever think of my electrical provider unless it goes down.
[01:07:16] Ben: Yeah,
[01:07:16] Dan: want the light on.
[01:07:17] Reliance on AI
[01:07:17] Tim: So you got some AI agents that are just running, like you put 'em on your GitHub repository and you create a bunch of prs and they just go work them, and then you review. I, I got a feeling we'll all become code reviewers more than we are code creators at some point.
[01:07:32] Ben: I have
[01:07:32] Dan: I'm afraid of that world, honestly, because
[01:07:34] Tim: I don't, I don't want that world, but that's, that's kind of what I'm afraid
[01:07:38] Dan: My standards will slip if I'm presented with a large block of stuff that looks mostly right and I find myself reaching for AI to do trivial things. Like here's an example. I've been coordinating a team and a, you guys know anything about soccer?
[01:07:56] Dan: Is that a thing for you?
[01:07:58] Adam: Football, you mean?
[01:07:58] Ben: A little
[01:07:59] Tim: Food ball.
[01:07:59] Dan: there's this tournament called the soccer tournament. It's a seven on seven tournament and they just, it just wrapped up last night. It's like five minutes from my house, right? It's a big deal. People come from all over the world and they had a youth tournament this year and so I co coached, a team and the youth tournament and I was like the admin guy.
[01:08:18] Dan: And if you know me, I'm not an admin guy. It's not what I do. Well, so you know, what I would do is I'll turn on dictation mode, either in Word or Google Docs and say what I think I need to say in the email. I give it to a like a GPT project that knows about the tournament, it knows the rules, and I just say, clean this email up for the parents and it takes my ideas and reorganize 'em so that it flows logically.
[01:08:47] Dan: It takes long run on sentences, turns it into concise bullet points, and does it save me 10 minutes per email? Five minutes? I don't know. But it does save me something. And on one hand I'm a lot faster at getting these things done, which means I have a lot more five minute periods in my day than 10 minute periods, you know what I mean?
[01:09:07] Dan: But on another hand, I don't know if I'm good at writing emails anymore. I'm good at talking in the word docs
[01:09:14] Ben: Well, it's like
[01:09:14] Dan: good at handing it to chat GPT.
[01:09:16] Ben: heard the most terrifying thing in an interview just yesterday where someone was saying about pull requests, that the AI can produce so much code that it's really untenable for people to be doing the pull request reviews for the AI generated code. So they're like, so we're thinking about spinning up a different AI to do the reviews of the other AI's prs.
[01:09:39] Ben: And I thought like, this is just like, there's no way that ends. Well,
[01:09:42] Tim: But, but Dan has an, an example in his course where Ai, like code reviews itself and it, it does come up with changes, which is crazy.
[01:09:52] Dan: That's crazy. If you told me that happened, I wouldn't believe it, but I'm the one that recorded it and published it, so I believe it.
[01:09:59] Ben: Now I
[01:09:59] Tim: that, that's the in, that's the in deterministic part of it. It's like it creates this code and it goes, and you go, can you code review this and go, oh, this code could be better. You graded it. Why didn't you do it the first time?
[01:10:10] Ben: So okay to to, to bring it home here. If one wants to go watch this course, because I'm very, uh, excited to learn about all this stuff. Now you said there's gonna be a code. What is the, do you wanna share the code with the audience?
[01:10:24] Dan: There's a link we'll put out there and the code is quackquack.
[01:10:29] Ben: And to be clear, you were not just cursing,
[01:10:33] Dan: You'll never know.
[01:10:35] Ben: you just have to go try lots of combinations of swear words.
[01:10:38] Tim: You, you didn't, you didn't achieve it. Dan, you didn't get quacked.
[01:10:42] Dan: Yeah. The code is the, the code makes the course as cheap as I can make it, but if you're out there and you don't have a job and whatever price it's charging is just too much, like hit me up on LinkedIn, I'm sure I can find you a way to get in for free.
[01:10:54] AI and Job Security
[01:10:54] Dan: You know, I'm not gonna, once again, no Caribbean Island out there yet, so I just want people to find the parts of AI that work for them.
[01:11:04] Dan: Sensibly weave it into their, their workday, but most importantly, to kind of feel more reassured that there's other viewpoints than AI is gonna take your job. And, you know what I mean? Because I just, I don't see it. I don't, I I just think we're gonna be able to be a little faster and better, like all the rest of the technological innovations,
[01:11:25] Adam: Well said. Uh, you know, I don't think that, uh, that's, that's all that anybody needs to hear on a, I think that the whole internet can put that topic to bed now.
[01:11:35] Dan: I'll let them know,
[01:11:36] Adam: Yeah.
[01:11:37] Tim: We solved it guys. Good job.
[01:11:39] Adam: that's what we do here. so, you know, with that said, why don't we wrap it here. and Dan, are you, you got some time you wanna stick around for the after show?
[01:11:47] Dan: what does that mean? I'm probably can do it.
[01:11:49] Adam: Uh, that means we, you can get, you can curse all you want and we won't quack it, after, after the outro rolls. Alright, well, let me do my thing here.
[01:11:58] Patreon
[01:11:58] Adam: So, this episode of Working Code is brought to you by slipping Tim a caffeine pill at three 30 in the morning. And listeners like you,
[01:12:03] Tim: What? What? I don't get that joke.
[01:12:07] Ben: your heart.
[01:12:09] Tim: my. Yeah, yeah, yeah. Thank you. Yeah. Appreciate it. Okay. Makes sense now. Sorry.
[01:12:14] Adam: if you're enjoying the show and you wanna make sure that we can keep putting more of whatever this is out into the universe, then you should consider supporting us on Patreon. Our patrons cover our recording, editing and transcription costs, and we couldn't do this every week without them. Special thanks to our top patrons, Monte and Giancarlo.
[01:12:29] Adam: You guys rock.
[01:12:30] Tim: Woo. Woo.
[01:12:31] Adam: As previously mentioned, we are going to go record the after show that plays after the outro music, uh, for those of you who are patrons. Uh, so just keep listening 'cause you know, that's why you pay money. Uh, so hopefully you already knew that that was there.
[01:12:44] Tim: And you get it early, so
[01:12:45] Adam: Yeah. Um, in the after show, there's a, a couple of notes here.
[01:12:50] Adam: Uh, there's a, there's something called oic, I think, which I'm, I'm not gonna,
[01:12:54] Tim: os Odyssey.
[01:12:56] Adam: where I'm saving the explanation for the after show, if you wanna know what that is. you're just gonna have to find out. so we're talking about Mob land and, and I'm gonna ask Dan, what the heck is a
[01:13:06] Tim: Yeah. Thank you. Thank you.
[01:13:09] Adam: we'll, we'll, if you wanna know, there's only one place you can't ask an l lm, you can't Google it.
[01:13:13] Adam: You gotta come to the after show.
[01:13:15] Adam: Uh, and if you, so if you wanna find out what a derailer is, you gotta go to patreon.com/workingcodepod. Throw us a few bucks and you'll, you'll get to find out too
[01:13:24] Tim: yeah. I like how you spell it like, it's like, it's like a French word.
[01:13:27] Adam: that that's autocorrect. I think that's the correct way to spell it, because it, it got autocorrected that way. Anyway, uh, that's gonna do it for us this week. We'll catch you next week and until then,
[01:13:36] Tim: Remember everyone, your heart matters even if you spend four hours to whack it. Three minutes off, like Dan,
prev episode: 219: Potluck: AI Ego, Feature Flags, Customer Feedback