Couldn't write the summary. The backup AI service has used its tries.
Nov 11, 1:02 PM Finished Nov 11, 1:00 PM Finished Nov 11, 1:00 PM Transcript ready Nov 11, 1:00 PM Call ended Nov 11, 11:00 AM Recording Nov 11, 11:00 AM On the call, not recording Nov 11, 11:00 AM Joining
Here we go. Hello. How are you?
This is my. My first day in two months back at my workspace. Really?
How's that feel? So you got to write about that.
Probably. I mean, but it's been mostly six months, so. Yeah. So excited to talk to you today. I want to do is talk to a smart person.
Well, that's what I'm doing. I don't know what you're doing, but thank you for sending your materials over, by the way. I love. I mean, it's so funny.
you follow it? It's like a dump.
Didn't read like a dump. My read was, oh, the workshop's pretty much ready. Like, this is what I'm talking. Like, this is the stuff I am talking about actively with. Oh,
daughter's calling me. I'm gonna tell her to call her dad.
okay. You sure? I mean, I can. I can pause for a second. That's fine.
No, she's an adult. She's fine. Hold on. We just need our time.
Once we start, it's not going to stop, you know, it's just going to be like, oh, tomorrow.
I know. I told Michael. I'm like, I got my headphones on. I shut the door. So how are you? First, how are you.
I'm doing all right. Yeah. You know, it's the. I'm hanging. I've hung up my shingle. It's mid being hung or the professional shingle hangers are hanging it. I don't know. There's a shingle. And, you know, I'm already almost done with my first engagement and. Yeah, helping a startup run focus group, end to end, and sort of engineer the context around the entire conversation, inclusive of all of the responses and reactions from people. And then I go deeper and I analyze, like facial reactions to demos and.
you doing that with. With the visual came?
yeah. So just with you can. There's two layers. I have two layers. Like, so you can see this note taker in here, in the call. So that's my notetaker. That can do a ton of things. Imagine if you had access to everything that calendly was capturing, but you could review it mid call. You could ask a question mid call that could pull up insights based on the last three calls that you've had. And so I've built a in person and now remote sort of full context
engine for not just recording the call and not just transcribing the call, but for me, most importantly, diarizing the call so that Bev said this, Danny said that, Michael said this. If you don't have that. Well, first of all, if you're using an agent or whatever and you've connected your note taker, whatever it may be, more often than not they're pulling the summary. All right,
yeah. I like the transcript.
Same. Like, that's where the sauce is at. And.
I hate the summary.
Yeah, it's always wrong and not how I would do it. Yeah,
And will you bookmark that thought? Because I have an unrelated resentment that I have to air, and I feel like you will be a vessel that will accept this story anyway. Okay,
so. So anyway, like, the diarization is the piece that is very. It's wanting. You know, you don't get that very often. And if you've got a group of more than three people, like, I'm sorry. Like, you probably need to go and review the whole call. Not just, oh, I've got. Yeah. And people don't do that.
the who said what Right,
Right. And so. And that alone. Right. And then the layers on top of that. So then it's able to, you know, I can augment what that little square is presenting. I can have that become the thing that I'm presenting. I could pin this in the conversation. And now it's streaming in live while I'm still here.
I think a lot about the infrastructure that no one thinks you have to care about in this role. And, like, you're doing it. You have a nice setup, good quality lighting, good camera, clean audio. Like, your first impression is going to be a lot better than 99 out of 100 of people that I get on calls with that are also like, oh, my God, no one cares. You know, like, they don't think you have to care. But if everybody has the advantage, like, everybody's got an AI agent. Nobody has the advantage. So everybody's got a webcam. Nobody appreciates that you can get a nice webcam. I think those tiny little tidbits just also are little professionalism signals. And
just like, you know, you're marketing like a. You're marketing a whole image. Like, I'm your technology partner.
I will a little piece of. You know, I'm always like, evidence based, right? So everybody makes fun of me. Like, did you read a book about that? And I'm like, okay, I did. Books behind you is the only according to the most recent research that I read, which may be wrong now is that books behind you'd build is the only credibility signal that it's been shown to work. So books behind me. There you go. Yeah.
moved my stack of books, but now, now I'm subconscious
You might want to put some books behind you. It's just books.
and that makes sense. Right. And so I think a lot about these intangibles in the same way I think about like, everybody can make a PowerPoint now. Right. I've spent the last 18 months probably most focus on artifact design. And so that's, that's a bucket of design engineering, which, like, I've been digging deep into a lot of the principles from. How were textbooks teaching quantitative information distribute display of quantitative information in the 1980s because that's when taste stopped.
What is the book? And I had a copy I bought. I bought the whole set at Stelter for the design department. They gave a shit list, but I liked it. It's the guy that did all the work on how to visually present information.
Yes. So it's tuft.
Do you know him?
Yeah, it's tuft. Edward r. Tuft, the visual display of quantitative information. And I've actually got here once. I am.
I loved that. That was a night I probably 10 or 15 years ago kind of blew my.
Yeah. And like, because everybody doesn't know what to look for and what they did before, which was probably no design, right? Like having the default cloud output. Looks good, right? It looks great. What, no notes, you know, oh, wow, look at all this information. But the infractions that those make, right? I mean, just as someone who's just kind of perusing, looking for a low cost whatever, like, you know, they're not going to notice, they're not going to care. It's not going to impact them. But if you really want to land something, like, meaningful, you got to be able to demonstrate that there's, like, you have taste. And so I distilled his book into a few different lab scenarios. So he's all about
the data ink ratio. And so if I slowly erase, you'll notice the data ink ratio shifts
at what you're doing.
and you can see like, okay, so here we now have like reciprocity across all of the bars. There's no infraction on. One looks higher or lower. If we drop that back, the axis alone can make that confusing. And so, like this stuff. The other one, the one I really like is this one. The lie factor. This is what every AI lab, everybody who's presenting quantitative information today, this is what they're doing. All right, let's go to the 3D cube. Here's a 70% data change, right? This is a volume X.
on the. Boo on the volume. Yep.
Right? And so if we keep going, here's 200, right. This is supposed to be 3X, right? 27. And this is almost like this is the pattern people are jumping on or here, Right. They'll split the axis. Wow, look at this increase. Right? Oh, but there's a little blip here. It starts in 900. Like it's the it. You think it may not matter.
did you train your AI, your visual piece on these principles?
So I kind of have a. I have a few different visual analysis tools and pipelines. One is entirely local. And so that's a. It's a combination of very small language models. Some that are really good at looking fine detail, others that are better at text. And so they work in a council. And that's what I had. Extract the entire kind of design principles from the copy of the book I had. From there, I then have a profile that I've made called Tuft. So if I want to apply the Tuft Tuftian principles, it's just kind of one ask and then I have my own profile. As the other one which just my tastes, my patterns and anti patterns.
And like to close that thought. I think why that matters is like. Everybody has stopped realizing they are standing on the shoulders of giants. In fact, you can abstract the entire reality that you are standing on the shoulders of giants, of engineers, of the evolution of sales, of full technological infrastructure changes. We're all connected now, you know, like, no one really. They think it's just, if I get the Claude subscription, if we, if our company gets co pilot, we're going to the moon, it's going to work. It's just going to be so great. It's like, okay, so you've added something. What did training look like? You know, like, who's measuring success? And, like, are we granular?
Yeah, there's so much. And Clyde makes work. You know, how do you. What do you deal with it? How do you deal with all that?