Episode 21 · Transcript

That's SO Last Model

Full transcript of the recorded conversation.

Back to the episode · timestamps jump to the video · lightly machine-transcribed, may contain errors

0:00 Kingsley

for a while I was saying skills are dead, stop using skills. I'm gonna change that and say the way you're using skills is wrong, Not you particularly saying, but

0:07 Sam Rogers

Me? No.

0:10 Kingsley

Yes, yeah. I was watching what you were doing. And no, no, no. I would say that

0:13 Sam Rogers

Ha ha ha

Welcome to Signals and Subtractions. I'm your host, Sam Rogers. Every week, everyone at the table brings two things: one signal worth watching, one subtraction worth making, that's the whole show. What to watch, what to drop. This week, who owns what your team knows? Most of us we have a team that knows something about doing its job, but where is that kept? Is it in training? Is it in tools?

Is it in the instructions that we write for AI? all of the above, right? A lot lives somewhere that we rent, though. So what do we actually own? What's worth owning? Joining me today are Susan and Kingsley, the husband and wife team behind Happy Alien AI. They built an AI that turns approved storyboards and slide decks and decade-old courses into new ones without changing a word that was signed off.

And Kingsley's new course authoring app, Xogogy it ships with no AI inside it at all. We'll get to why. Susan Kingsley, thanks so much for being here.

1:28 Susan

Thank you, Sam. This is an awesome opportunity to see you and so why not? worthy discussion.

1:35 Kingsley

Worthy discussion.

1:36 Sam Rogers

let's get to it. Kingsley, guests go first. What's the signal you're watching right now in your own work this week that you'd like to talk about?

1:45 Kingsley

long-form agentic workflows. we've had agentic workflows before to create an image and post it. But now I can fire up agents and give them all the tools they need to work all night long. And then when I wake up in the morning, they've not only created a thing, they've done stuff with that thing. we just launched a new tool.

possibly it's called Xogogy But

we just launched this new tool and it's a very full feature tool. And so we need to create all these tutorials on how to insert a trigger and how to update the style and all that stuff. And I'm frankly a little lazy when it comes to that. And I did not want to spend hours and hours recording and editing video.

And so I was able to set up a workflow where it uses rules and access to different tools. And so then I can just spin it up and it works overnight, and brings it all in. And then actually I pass it off to another agent to publish to YouTube because YouTube has some difficulties, you know, with AI interacting. And I found, I don't know.

if you've used it before, but Instinct is it's currently free right now. Which I guess makes us the product. if it's free, then you're

2:55 Sam Rogers

Ha ha.

2:56 Kingsley

the product. but it's a pretty powerful AI agent. for me it lives in WhatsApp and I can just say, hey, here's a Google spreadsheet full of videos that I need you to upload to YouTube. And it comes back to me later and says, There they are. so it's pretty cool. I'm not sure how they're getting away with it really. But

3:11 Sam Rogers

Yeah.

3:12 Kingsley

I was able to automate that whole process using several different agents. and overnight it makes like 10 different videos, and they all look pretty great. Now, could they be better to make them more perfect? Sure. But the key is I didn't have to do anything. And so

3:29 Sam Rogers

Yeah.

3:30 Kingsley

I mean I set up the original guidelines and like I said, rules for it, but

3:34 Susan

and then you used ElevenLabs to

3:38 Kingsley

Yeah, from the clone your voice. Yeah. I was a little sick the day that I cloned my voice, so it's not exactly

3:43 Susan

it's been going around. Yeah, yeah. Yeah,

3:46 Sam Rogers

Permanently sick. Yeah.

3:48 Susan

yeah.

3:49 Sam Rogers

so when the agent does all this stuff and it records a demo, I've been doing a similar workflow for a while as well. but there's sometimes it gets it wrong. So I'm curious like how it is that you're both finding that something isn't quite right, something needs modification, and well basically how do you find out before your customer does

and fix that?

4:10 Kingsley

I have a graph somewhere that I was gonna share to illustrate some of those. Yeah, yeah,

4:13 Sam Rogers

okay, screen share, yes.

4:14 Kingsley

as you mentioned before, it doesn't have any AI built into it. But we built it so that you can use your AI. And we'll talk about that later. But that also means that my AI agent that's recording the video.

can navigate the course itself or navigate the tool through that MCP. But also

4:33 Sam Rogers

Mm-hmm.

4:33 Kingsley

the agent has access to the code base. And so it can check to verify everything. I wrote a brief for each video. You know, in this video we need to show how triggers and actions work or whatever. And so it will do a dry run first.

And then compare that to the code to make sure that it has everything it needs to go through that. And if it has a problem, I've instructed it to either get around that problem. Like for instance, there was an issue on Mac. I guess it's option Q is supposed to be search, but it wasn't working in the actual application. And so

5:07 Sam Rogers

Okay.

5:08 Kingsley

it then I saw on the logs, it saw option Q didn't work. It went back and just clicked the search button instead.

And so I've given it some freedom to work around that kind of stuff But because it has access to the code and the app, it can bounce off those two. And then I have another agent well, two others that are evaluators. And so one evaluates the script before anything's recorded. and then it also evaluates the dry run that went through. And then third agent actually evaluates the video when it's done looking for things.

But then as you can see on my board here, anything that's pink is me in there. And so there are a couple of places where I need to go review it and make sure that it's gonna work well and

that it looks like.

5:54 Susan

And then I'm here to pipe in as well where I'm like, Hey, the first part of these videos keeps getting cut off. What's up with that? You know, and

6:01 Sam Rogers

Yeah.

6:01 Susan

so Yeah, yeah,

6:02 Sam Rogers

that's very important. Like you're the one that customers are going to probably hear from as well. So have you had

6:07 Susan

Yes, yeah, yes.

6:10 Sam Rogers

any feedback or input from people who are asking is this AI or how AI is it? Hey, it sounds like Kingsley maybe is a little sick. Is he still okay? like

6:22 Susan

Ha ha

6:23 Sam Rogers

Are people noticing, first off, are you disclosing that it's AI? are you getting any pushback

6:28 Susan

We have

6:29 Sam Rogers

from anybody about those kinds of support videos or anything?

6:31 Susan

You know what's really well, not necessarily about the support videos, but just feedback has been so awesome because it's been in the beta and we get insights that I haven't even considered. For example, the one about using it in very secure. yeah. Yeah. I don't know if you want to tell that to Sam, but I thought that was very well, I know but one of the

6:53 Sam Rogers

Ha ha ha.

6:53 Kingsley

beta one of the beta users was suggesting that because

As we mentioned before, we built the app to be really self-contained and not call home and stuff. They were saying that they work in a government facility in what they called a cleared room, where they're not allowed to have certain and so they were saying this is gonna be great for that. but I think your question was specifically more about like the video process. no, nobody has pointed out yet, hey, that sounds like AI. but I do hear what you're saying. Maybe we should put a little thing at the beginning that says.

This was completely generated by robots.

7:27 Sam Rogers

Well, for my own videos, you know, on signals and subtractions, we make one with every takeaway or support videos and things like that. I do generally say that also in part to credit Limited Edition Jonathan, who's been on the podcast a time or two, talking about the similar kind of process, you know, different products, but like being able to have a video kit that basically builds elements.

without as much human intervention. and I've been, you know, using it a ton. I've used it just, you know, three times today, customizing it. and what I love about it is that like for that government use case you were just citing, because it is actually

driving the video build,

it's using

AI in parts to like call out to do things, but because it is so self-contained, it actually works really well for my workflow, which is generally local first. So so that's that's

8:22 Kingsley

Yeah. Have have you tried

I'm wearing get off on a tangent, but have you tried it with like a local model?

8:28 Sam Rogers

I haven't done the full build with local models yet. I have been getting it down in terms of intelligence, which is, you know, part of what I wanted to talk

8:35 Kingsley

Yeah. Yeah.

8:36 Sam Rogers

about today was skills and stuff,

how much of it is outsourced to like local information. It's great that it's got all of the code base and whatever, but the actionable part.

of the skills, which could just be like a copy-paste prompt or whatever, but how is it that we're referring to that, keeping that stuff updated, and then how much of it is the model itself is something that I keep having to reevaluate as the models get better and better.

9:01 Kingsley

Yeah, yeah. And you've very wisely led me right into my subtraction for this.

9:07 Sam Rogers

Go for it.

9:09 Kingsley

I you know, for a while I was saying skills are dead, stop using skills. I'm gonna change that and say the way you're using skills is wrong, generally speaking for most people. Not you particularly saying, but

9:19 Sam Rogers

Me? No.

9:22 Kingsley

Yes, yeah. I was watching what you were doing. And no, no, no. I would say that

9:25 Sam Rogers

Ha ha ha

9:26 Kingsley

in my video workflow I have files that I call skills in there because that's how the agent looks for it's looking for a skill.md file. But in there, I don't just describe, here's how you make a video, because you're right.

AI is getting smarter every day. It knows how to make a video. And so what I need to do is give it those rules and guidelines. For instance, I need all my videos to look the same. So even though it knows how to make a video, that doesn't mean that it knows all the time how to make my video. And so in mine it says, you know, I need three beats, I need what the thing is we're gonna solve, how we're gonna solve it, and then show me how we solved it.

10:03 Susan

Much like instructional design, actually. Yeah.

10:06 Kingsley

and then also access to my Google Drive. So it knows where with the spreadsheet to put it on and stuff. Now, AI knows how to access a Google Drive, but my skill MD spells out this is the spreadsheet that you're going to access. So it's very

10:23 Sam Rogers

Mm.

10:23 Kingsley

bespoke. I love the word bespoke. It's very bespoke to me and this process

And so when I see, and it's so popular, people will post a list on like LinkedIn or something with 30 skills you must install. And I skim it and I'm like, it knows all that stuff, you know? Yeah. And one of the most popular ones has been the SCORM wrapper thing, you know. This, you know, install this skill and it, but if do you ever try it without the skill and just say and I've seen some people do it, it does a good job. It does a good job all on its own. And you might have a little back and forth where you're gonna say,

you know what, on this thing, on this slide, there's a complicated drag and drop. Could we, you know, that kind of thing. What makes it more advantageous to you, more useful to you, is when you start saying, after you've wrapped it in SCORM, I need you to upload it to SCORM Cloud. Here is a Python file that already has my credentials. And you don't want to give the credentials to

Don't write the credentials and the skill, but give it access to a Python file running on your box that it can then use to access the Review My eLearning API. to upload your interaction to Review My eLearning. when people get those 30 skills, you don't know what's in there. Yeah. Yeah. Yeah. Yeah.

11:36 Sam Rogers

Yes. Yes, that's very true. Yeah.

I've made quite a few things to solve for that and trying to circulate the security concerns around all that.

11:46 Kingsley

Mm-hmm.

11:47 Sam Rogers

but you are very wisely leading into my signal for this week, which is that the newest model made a bunch of the skills that I rely on just unnecessary. so back in July,

even before then, when Fable first came out, a lot of my skills were triggering security flags, left, right, and center. So like exempting all of those, I had like 54 skills left that I had made myself. I was on like a one-a-day kind of skills cadence for the first quarter of the year. and I was looking at them

from the perspective of like one question, does this still add anything for the best models that I use? And of those 54, 11 of them added nothing. Like just like your SCORM example. and we should define too for everyone who maybe isn't in that e-learning world. SCORM is one of the core standards that e-learning courses use to talk to the kind of content management system called an LMS

That they transmit back and forth to. anyway, of these skills, nine of them were partly obsolete. So, like between the both sets, there, that's like a third of my skills that actually weren't doing anything, and in some cases were making it harder than it needed to be, like burning more tokens to get to the same destination.

13:01 Kingsley

No. Mm-hmm. Mm-hmm.

13:03 Sam Rogers

and just this.

last week with the newest releases from OpenAI, from Anthropic. I'm watching it happen again. there's a basically only one kind of skill that decayed as much once I started looking at it, which is

step-by-step instructions for how to think about stuff. Stuff that we used to need to make the models like work at all a year ago for our workflow. And the facts that a model can't know. So, like in your example, Kingsley, of here's the direct path, you know, here's what you invoke for these credentials, all that kind of stuff. the scripts do the real work.

the mistakes or things that like I'm trying to write down and keep track of. so when a team is writing down its own know-how for AI, some of that's built to last, right? but some of it's built that's just like for the current configuration of things. And

I think it's a useful way to look at it from a from an agentic skills perspective, because what is built to last versus what's just built for this model? You know? in a temporary basis? Like this is going to be a regular recurring maintenance cycle. And what is it that's going to like live as part of our more

Permanent communication, say with other teams, or things like that. You know, like approval processes or something like that. Those might change in small ways, but we don't expect them to change in huge ways every day, right? But the

14:32 Kingsley

Mm-hmm.

14:33 Sam Rogers

what the AI is doing to enable those approvals does change. And how much approval do we need when, as the models get smarter and smarter?

actually does change in some not so subtle ways. I'd love to get your thoughts.

14:47 Kingsley

Yeah, yeah,

yeah. Yeah, yeah, yeah. I agree. how much direction it needs definitely changes the s the smarter it gets. I was wonder y when you said that you were evaluating them, how were you evaluating them? Do you have a methodology you're using or a tool or something?

15:03 Sam Rogers

Well,

as of today for this episode, thank you for leading into the takeaway yes, I've been doing it more or less formally, like just for myself. But what I have here now is something that I've made public to accompany this episode, which is showing exactly how to do what I did. So you can just point your AI to that page.

Which is sigsub.show slash takeaways. And you'll you can see all of our takeaways. What's scrolling through on the screen right now is the copy-paste prompt. And unlike so many of the skills that you'll see out there or the you must have, this skill kind of things, I actually made a security check as well as an open protocol, the GuideCheck standard.

Which can make sure that there's no like white on white text there and all that kind of stuff. Like it's basically impervious to the usual kind of injection attacks or security issues so that you can know when you point your agent at something that it's seeing the same thing that your human eyes are seeing. But for the tool itself.

That's what I'm using now. I've actually just been testing it a little bit. I sent it your way like late last night. I don't know if you got a chance to look at it yet, but but yeah, I'd I'd welcome your input as well on how to make stuff

16:17 Kingsley

I haven't seen it yet. No, no, no. No. It was s it was just serendipity that I Yeah. Yeah. Yeah.

16:24 Sam Rogers

like this better.

16:25 Kingsley

Yeah. Yeah. I'll look at it and see. Mm-hmm. Mm-hmm.

16:29 Sam Rogers

and maybe I should ask, like as a company, how is it that you think about stuff like that? Is that some of the reason you don't have AI in the new product?

16:37 Kingsley

yeah, I mean that's part of the reason is for security, but also I just you know having an ear to the ground with the community, and I'm not trying to, throw rocks or anything, but when Articulate said, if you're not part of the industry, Articulate builds one of the most popular e learning course development tools.

And they've been using AI. You have an option to put AI inside your tool so that you can chat with AI and ask it to make a slide for you. And it was an add-on option. But then they said it's no longer an add-on option. Everybody's gonna get it whether you like it or not. And here's how much more it's gonna cost. And there was a lot of pushback. Some folks just

17:14 Sam Rogers

Right. Yeah.

17:16 Kingsley

can't. Where they work, it's not, you know. And so I'm sure they have some ways to toggle it off for your enterprise or whatever.

But people weren't thrilled with it. And in my everyday world, using AI all the time, well, this week this one might be the best, but next week another one's the best. And then I want to go back to this one. And if you're locked into one, then that just means you can't enjoy all the benefits of all of them.

17:39 Susan

We like having control over the model we want to use. And some people like one model over another. Some enterprises can't use Claude for example. And so we want to give them that option.

17:51 Kingsley

So I wanted to be completely optional whether or not to use AI in the tool. And then if you decide to use it, I wanted you to be able to choose which model and how you interact with it and stuff. from a security point of view, yeah, I mean that is one of the things because when building the interface for your AI agent to work with this application, we have to watch out for to see, you know, what if

You're trying to use your agent to fake a license for the product so that, you know, then you can avoid paying me the $99 or whatever. You know, and so there are some things in there. that one not so bad, but when in a previous life, when I worked with it used to be called eLearning Brothers, but now it's ELB. I led the team there that created MicroBuilder, which is like a Rise competitor.

that had a lot of AI options in there. and that one did require a lot more because you have to look for prompt injection attacks. is somebody trying to get our instructions out? Is somebody trying to get into our instructions as somebody? And so you do have to think those things through. It's a whole new way of thinking about security

the one thing I want to add on to that is having built a product that's like MCP first, and then adding MCP to our other product, which has mature, been around for fifteen years, going back and adding MCP to that, I can see where you know people think that this software has been around, it would never die. But it's gonna be hard for them to engage in this AI world.

when you start out building a product from scratch and you're thinking AI first, it's gonna give you an advantage, I think. Even if you don't have all the best features as the other ones, if you can make it MCP first, I think it's gonna give some companies an advantage in the new brave world.

19:37 Sam Rogers

Yeah.

one more thing I want to ask just about your new position on skill files. how is it that you have been discovering what to stop putting in skills, things that or what maybe what made you stop

using skills as the de facto go to?

19:53 Kingsley

Yeah. yeah, I think I it just started by gut instinct kind of where I would see someone say, Hey, here's a skill that will, you know, make Claude animate stuff. And I would think to myself, maybe it could just do that now. And then I would ask Claude to make an animation of something, and it would. And so then that led me just further down that road of being more investigative of skills in general.

Also, and you mentioned this earlier the token bloat. Once you start throwing

20:21 Sam Rogers

Yeah.

20:21 Kingsley

a lot of skills in there. Yeah, yeah. And then ironically, we had a customer at Happy Alien, we build products, but we also do services and build products for people, clients and stuff. And we had a client that wanted a bunch of skills built. And as I was building them, I thought, hmm.

I don't know if the it needs all these skills. And so it was nice because we were able to trim back, you know, what they actually needed. and so it turned out to be a subset of what they needed. But when you start building your own skills, maybe that's when you start realizing you don't need skills.

20:53 Sam Rogers

Yeah, it's also good to remember just how fast this moves. I think it's important that we keep that attitude of continual improvement and being able to test.

Just exactly what we need it to do. Because as the models get smarter and smarter, they may need less and less of that specification for our own unique parts. We may be able to offload more and more of that. And just having that

21:15 Kingsley

Yeah.

21:18 Sam Rogers

as a expectation for how it is that we build,

well, for myself, I'll say. It changes what it is that I want to build in terms of product.

you know, of what's gonna get better as the models get smarter versus what's just gonna be a shrinking piece of the pie that they'll continue to consume,

21:35 Kingsley

yeah, yeah, Yeah.

21:36 Sam Rogers

Is there

anything that you guys are talking about around that? Like as business owners yourselves,

21:41 Kingsley

I got excited about an idea recently and it was based off of fear. Where

21:46 Sam Rogers

Yeah.

21:46 Kingsley

so they're starting to now be able to use AI to alter live video. to tailor to your audience. it makes it s

21:53 Sam Rogers

Yeah, yeah.

21:54 Kingsley

moves my lips and change. But way back in the day,

At a utility company, we were developing warehouse safety stuff. And I went to the warehouse, took a bunch of photos, and then immediately ran into a problem where at some warehouses they had certain safety equipment, and at other warehouses they didn't need that safety equipment. And so this argument ensued over it. And so I love the idea of that personalized on the spot in our case training, you know.

And I think we're headed there where I could take a photo at one warehouse and when you take the course, it changes the hard hat or the break on or off or whatever needs to be done for that, based on what office you work in or which part of the world you're in. Instead of having to take five photos, take one. Yeah, I think that'd be right.

22:40 Sam Rogers

Yeah. Yeah. Yeah.

22:43 Kingsley

I'm imagining a world now. I'm getting way off here, but I'm imagining a world

22:46 Sam Rogers

Yeah, go for it.

22:47 Kingsley

now where you have the let's say maybe the SOPs, the standard operating procedures of four different warehouses. You go take one photo. Right now, what people have to do is they have to take that photo and then go back and say, no, they wear a blue cart hard hat and then they have to ask AI to change the photo. But wouldn't it be cool you take that one photo, you have it automatically.

Go through those five different SOPs, find the color of the hats, and then just automatically f identify everything in the photo that's different for each warehouse, and automatically just create you five different versions of that. And then in your training material or your course or your SOP, it could just automatically switch based on your location. Yeah, it'd be cool. That'd be cool. Sorry,

we got way off from signals and subtractions.

23:29 Sam Rogers

Yeah. Well, it's not it's not

actually that far off. I think it gets to the same basic question of like what belongs in the model? What is the stuff that the model can

23:40 Kingsley

Yes.

23:41 Sam Rogers

know? And what is the information that we need to give it, like for the agents to be able to deliver the outcomes that we're looking for? what's the best way to specify that context

it may be a very smart model, but it doesn't know what color hard hat is in which place, you know?

23:58 Kingsley

Yeah, yeah.

24:00 Sam Rogers

so right now there's the places that we store information like that that we're renting, and then there's the places that we store information like that that we can own. and I would always advocate for the things like you're talking about in that idea, Kingsley.

Of being things that are on systems that we own and operate, because then we can plug in to anything we want. and if there's an outage or a billing change or, you know, all the stuff we were talking about earlier, like you can't use that model here, all that kind of stuff. Like it doesn't have to matter as much if we keep ownership over the parts that are truly ours.

And then we keep our kind of instructions out of it that can drag the model down or cause more tokens, but we can keep our specifications in it, you know, in that part

24:49 Susan

Mm-hmm.

24:49 Sam Rogers

that we own.

24:50 Susan

Yeah. Yeah. Yeah.

it's fascinating, right? I think about

all kinds of things like how much I use Google Drive, you know, for all of my communications. And is it truly secure? I'm sure it is. But then again, you know, like I don't even know anymore.

25:03 Sam Rogers

Well, it it it

may be secure, but it's not yours. Like that's a great example of something that we rent versus something that we own. So yeah.

25:10 Susan

Exactly. Exactly. Yes. SharePoint or yeah, like

anything with anyone else's logo on attached to it is not ours. Yeah.

25:21 Sam Rogers

Yeah. it's

not an either or either. Like I have my own NAS system here that's got a bunch of stuff on it that's like I own

25:28 Susan

Nice.

25:29 Sam Rogers

it. It's in, it's right over there. But I use Google Drive all the time. I use OneDrive. I use Share. Like it depends on what it is that you're trying to do

and what helps you achieve that result, right? It's not all one thing or another

25:40 Kingsley

yeah, it's about level of effort too. Yeah, like I could I build my own share drive system and stuff? Yeah, sure. But for seven bucks a month I can just have almost unlimited. But yeah.

25:52 Sam Rogers

Yeah.

That's great. thanks for spending a little time here with me today. if people want to follow up with you, Susan Kingsley, where do we send people?

26:01 Susan

Connect with me on LinkedIn. I absolutely love getting messages. I get all kinds of messages from interesting developers and instructional designers every day. And I just love keeping that communication going it's just always a pleasure to hear from people in and that's my go-to space, honestly.

26:19 Sam Rogers

And that reminds me, um, part of the reason I was reaching out to you guys was seeing this come up in my feed from Susan recently. Um

We last yes.

26:32 Susan

we're we're beautiful people.

26:35 Sam Rogers

Yes, this says DevLearn 2026 but that's actually a photo from 2025 with our lovely faces as well as uh Alaina and uh Jess Almlie

26:45 Kingsley

they thought to themselves, how can we advertise that we have the brightest, the smartest, the coolest people?

26:51 Sam Rogers

That's what it is.

for everyone who's watching, I would encourage you for this week to pick one piece of know-how that your team has handed to an AI and ask where it lives, who can change it, and whether anyone has tested it that it still helps. Because as things get faster and faster, some of those

Fantastic solutions that you used to need may not be needed in quite the same way anymore. And also that there are new solutions that have grown up around us, products like from Happy Alien AI, things that I'd, you know, we talked actually at DevLearn. I don't know if you remember Kingsley, we talked about

27:32 Susan

Mm-hmm.

27:32 Sam Rogers

this.

Product that you're developing and putting out there now. We talked about the whole problem and exploring all that. And it was just a dream to be able to think through that stuff a year ago, right? But now you've built it, it's there, it's live, people can use it and support you in doing that. I'm a happy customer I should disclose as well to Happy Alien AI.

27:55 Susan

you're a smart man, so you know.

27:57 Sam Rogers

Yeah, that's why my picture is up on the billboards at events. having

28:00 Susan

That's right. Trailblazer.

28:03 Sam Rogers

solutions like you can build now that just weren't really that possible before. Like maybe it would have been possible to make but like it's easy to do it now. So all of the things that we are building

all of the assets that we're delivering now. I think it's like a maintenance cycle kind of item that is definitely less than annual.

28:22 Susan

Mm.

28:23 Sam Rogers

I'm evaluating my skills basically once a month now and making sure that they work the way that they are intended or that they're needed. So I would encourage everyone to at least find one time when it was tested

And that you have those tests to answer your question earlier, Kingsley. Like, how did you know that

28:40 Susan

Yeah. Yeah.

28:42 Sam Rogers

that's actually better?

28:44 Susan

Yeah. Mm-hmm.

28:45 Sam Rogers

yeah,

thank you again both. really appreciate you being here.

28:48 Susan

Thank you, Sam.

28:50 Sam Rogers

we will send folks to Susan's LinkedIn as well as Happy Alien AI.

we do this live every Wednesday, and the newsletter is every Sunday.

And on Friday in between, this will come out on all the podcasts and YouTube and everything. If you're listening on audio right now, please tap follow. If you're watching live, open your podcast app, search for signals and subtractions, and follow us there so that the next one just shows up next week. Sunday is issue 78, I believe, of the newsletter. One signal, one subtraction, one analogy. It's about five minutes to read. Everything.

About the show lives at sigsub.show one signal worth watching, one subtraction worth making. Now go find yours.

Keep the next choice in view

One signal, one subtraction and one analogy every Sunday. Five minutes, no video required.