AI Learning Lab

02./19/2026 -A Deep Dive into Crafting a Custom AI "Soul" for Your Personal Agent with OpenClaw

dm11ypysr8U
Live Stream2026-02-201:28:45161 views

Description

Kyle explores the evolving landscape of personal AI assistants, sharing his hands-on experience installing local models like OpenClaw and Ollama on a gaming PC. He dives deep into the concept of the "soul.md" file, explaining how this markdown document serves as the personality DNA and ethical compass for an AI agent. By treating these tools as collaborators rather than just task-runners, he highlights a shift toward "human first" technology that prioritizes personal agency over raw productivity. Transitioning from theory to practice, Kyle demonstrates "vibe coding" by building a new platform for his movement, The Great Repurpose. He uses the Lovable platform to design a "self-check" tool that helps individuals navigate the identity shifts caused by AI-driven task erosion. The session emphasizes the transition from corporate frameworks to more human-centric metaphors, positioning community as the primary way to find "signal" within the noise of the current tech landscape. #AI,#OpenClaw,#VibeCoding,#TheGreatRepurpose,#AISalon,#LLM,#PersonalAI,#TechPhilosophy Chapters: 00:00:00 Opening Music 00:04:23 A Song Erased 00:07:51 Installing a Local LLM 00:10:05 The Importance of Struggle 00:12:33 Defining the Agent's Soul 00:15:38 Explaining OpenClaw 00:20:36 AI-Generated Personality 00:24:11 Building the Repurpose Website 00:28:55 Designing the Assessment Tool 00:32:01 Generous Leadership Philosophy 00:36:55 The Self-Check Concept 00:40:09 Empathetic Question Design 00:45:01 Introducing Tension Sliders 00:50:03 The Great Repurpose Song 00:55:17 Starting the Vibe Code 01:01:49 Beyond the Framework 01:06:13 Finding the Signal 01:12:14 Your Signal in the Noise 01:16:27 Approving the Final Plan 01:20:46 Universal Basic Compute 01:24:32 Open Source Community News 01:27:14 Closing Remarks

Chapters

Transcript

0:05 [clears throat]
0:07 What do you think, champ? You think
0:09 now's the time?
0:12 Do you think now is the time?
0:17 [music]
0:26 Something, baby. [music]
0:34 >> [music]
0:40 [music]
1:09 >> Every time I see you now, get that look
1:12 in mine.
1:15 Every time I see your mouth, I hear that
1:19 smile.
1:20 Early
1:22 [singing] misty morning light. I heard
1:25 the [music] engine turning
1:28 all the fun outside.
1:36 You believe in me
1:39 again today.
1:42 You convince me [music]
1:45 again today.
1:48 Leaving this hotel [singing] looking for
1:51 [music]
1:51 someone else's golden ring.
1:56 Should I say
1:59 so long Suz [singing]
2:04 [music]
2:05 now don't cry.
2:12 Suz, [singing]
2:16 [crying]
2:17 don't you cry for me [music]
2:28 and cigarettes. He is on it tonight.
2:31 Keeping out [singing] on the road,
2:33 [music]
2:35 chasing down a lifestyle out on Highway
2:38 24.
2:40 >> [music]
2:41 >> New York State rolling breeze in the
2:44 sunshine with a blue sky falling to the
2:48 chill of old [music] September creeping.
2:55 You were leaving [music and singing] me
2:58 again today.
3:01 [music] You would convince me
3:04 again today.
3:07 [music]
3:08 You're leaving this hard time looking
3:10 [singing] for [music] someone else's
3:12 golden ring.
3:17 Should I say
3:19 so long?
3:22 [music]
3:25 Now, don't [singing] cry.
3:30 [music]
3:31 So long.
3:34 [music]
3:37 Don't you cry for me.
3:51 [music]
3:56 [music]
4:00 Happy Thursday night, good people.
4:03 Tomorrow is Friday. Friday night date
4:05 night, office hours. We got all sorts of
4:06 stuff going on tonight though. We vibe
4:08 code. Can you feel the vibes? Can you
4:11 feel the vibes? Everybody
4:16 your song. Oh, by Elton John. I don't
4:20 know it. [clears throat]
4:24 Wait, I lost track of time looking for
4:25 your song. Which song?
4:29 >> So that that's uh for me. Um, I put a
4:32 call out on Tik Tok to write a song that
4:35 has been erased from the internet.
4:38 >> Oh, yeah. I saw that.
4:40 >> You got You got nothing nothing yet.
4:43 >> Well, nothing on that song, but Weaver,
4:46 God bless him, found another remix
4:49 that's basically the same version slowed
4:52 down. So, I think I might be able to
4:53 pull that in and speed it up to get what
4:56 I'm looking for. Wow, that's
4:58 >> I did not think that it was possible in
5:00 2026 to completely erase anything from
5:03 the internet.
5:05 >> That's crazy.
5:08 That's really cool.
5:09 >> That's what that's about. So,
5:11 >> I like it.
5:12 >> If you know any EDM, that would be cool,
5:14 but I don't think you can get anymore.
5:18 [laughter]
5:23 >> Hey, Source Camp, what's happening?
5:26 >> [music]
5:34 [music]
5:40 [music]
5:54 >> This is [music]
5:56 I can risk my body.
6:02 Gather my thoughts [music] in sweet
6:04 silence.
6:09 The easiest [singing] place [music]
6:11 where the feelings are dead
6:15 [music]
6:16 from an over exposure to violence.
6:21 Is this place [singing] I can slowly
6:23 [music] face? The only one I truly can
6:27 know.
6:29 [music] These are tears from a long time
6:32 ago.
6:33 Can't these tears [singing] from a long
6:36 time ago. I need to cry 30 years or so.
6:41 These old tears from a long time.
6:47 Go.
6:49 >> [music]
6:54 >> Hey
7:07 [crying]
7:10 oh.
7:13 [laughter]
7:16 Have you met Chimp? He's the best. Dad
7:19 is the best singing dog in the whole
7:20 wide world. [laughter]
7:23 That is the best singing dog in the
7:25 whole wide world. He is fat and happy.
7:28 He is a cheese eating
7:32 singing dog.
7:35 Oh, good lord. Good lord. Good lord. Hey
7:37 Kyle, did you put LLM on the gaming
7:40 computer? How does it work? Hey, Pavven.
7:42 Pav. Pav. Pavven. thinking Pavin could
7:45 be paving
7:47 paving these streets, but probably
7:49 Pavven the guy. [laughter]
7:51 Anyway, uh I started I I I'll tell you.
7:56 So So the question is really about
7:58 OpenClaw and um did I did I put [snorts]
8:04 um a small LLM model on my
8:08 gaming computer with the 4090? The
8:10 answer is
8:12 I put an LLM model in there. Quen 2.5
8:16 32 billion and it's really slow. So I'll
8:19 probably drop to 15 billion [music] but
8:22 I don't maybe I don't know. Um but what
8:25 I did so this is really genius. Okay. So
8:28 so what I did was I I basically said
8:32 here's the specs of my gaming computer.
8:34 go out and do research on like Reddit
8:37 and Twitter and wherever the hell people
8:39 are talking about OpenCloud and find
8:41 like the latest best practices on how
8:44 people are installing it on that
8:46 computer. And then I said, "Then give me
8:48 a step-by-step set of instructions."
8:51 So, it did and it's pretty good. And
8:53 I've gotten to the point where I
8:54 installed Open Claw, I installed O Lama,
8:59 I installed Quen 2.5,
9:03 32 billion parameter coding LLM because
9:06 that's what it told me to install. And
9:09 then
9:12 I launched Open Claw and started the
9:14 configuration and then it says choose
9:16 Olama from the from the configuration
9:19 model and it's not in the configuration
9:22 model. So then I went back to Chat GBT
9:24 and I said it's not in the configuration
9:27 model. He's like yeah well that's an old
9:28 that's an old your instructions are out
9:31 of date. I'm like well you [ __ ]
9:33 created them you [ __ ] [ __ ]
9:35 [music]
9:37 So, I am now officially stuck. This
9:40 actually very much reminds me of when I
9:43 tried to install
9:45 stable diffusion for the very first time
9:47 in like late summer of 2022.
9:51 And can you install cams?
9:55 Producer Brandon is on in particularly
9:57 fine form tonight. I like it. I like it.
10:01 I like my people spicy.
10:06 Um, [music]
10:10 so in in summer and fall of 22, I failed
10:13 nine times to get uh stable diffusion
10:16 installed. And and the reason I stuck it
10:19 out is because I
10:22 thought it felt like it was important.
10:24 And Brandon just told me there's a
10:27 couple of websites out there where you
10:28 can basically just pay 47 bucks a month
10:30 and then you have Open Claw on their
10:32 servers and it's secure and you don't
10:34 have to set it up.
10:37 But this one feels to me like it's
10:41 important for me to
10:44 suffer through it and get it installed
10:48 and understand [music] how to set it up.
10:53 How to edit my soul.md file.
10:59 [music]
11:05 [music]
11:09 >> [music]
11:16 [music]
11:21 >> I don't know if that was much of an
11:22 answer. Pavin, uh, I So, where I am
11:26 right now is I've got an LLM installed.
11:29 I've got OAMA installed.
11:33 I've got OpenClaw installed. I don't
11:35 know how to connect them all together
11:36 yet. And I had to I had to run out. I
11:38 was at I was at Whiskey Club tonight, so
11:41 I was seeing some some friends there.
11:43 So, I had to run out. So, I only got
11:44 like 45 minutes on it. But anyway, it
11:46 was pretty it's pretty cool like having
11:50 having step-by-step instructions, even
11:52 if they're not right, it like tells you
11:54 what to do. And like when it throws
11:56 errors, you can just copy the errors
11:57 over to chat GPT and say, "Here's my
12:00 error." like like the the level of
12:03 anxiety of not knowing how to install
12:06 this [ __ ] um is much less than it would
12:09 be if if you didn't have all this AI
12:11 [ __ ] to help you through it. Um
12:16 yeah,
12:18 I did a pretty cool exercise today
12:20 though. There was a really good article
12:22 on that soul file for your for your
12:25 agents is really the important file.
12:28 It's like getting the getting the
12:29 personality of your agent right is
12:32 really important. [music]
12:34 So, one of the things I did today and
12:37 like these these exercises that I'm
12:39 doing like make me step-by-step
12:41 instructions. That was maybe 15 minutes.
12:43 Um, I said go out and research the
12:46 soul.md file, you know, on OpenClaw.
12:49 help me understand what it is, why it's
12:52 important, and then I said, based on
12:54 everything you know about me, I want you
12:56 to tell me how you think I should modify
12:58 that sold.md file to create an agent
13:02 that's like a good agent for me. And
13:05 then I said, so and and I want you to
13:07 explain it and tell me why. Like what
13:09 are the changes you would make and why?
13:11 And then after that, I want you to write
13:13 me an example soul.md file based on best
13:17 practices and based on all of your
13:19 recommendations
13:21 just so I can have an example. And what
13:23 it made was pretty good. In fact, let me
13:24 go. Did I do that on Claude?
13:29 I think I did it on the chat. JPE
13:33 um
13:36 let me see.
13:39 I can't remember the great repurpose Mac
13:42 hardware recommendations. Yeah, I think
13:44 it's going to be in the chat. JPE
13:48 uh uh uh
13:51 here.
13:53 SoulMD primer.
14:00 Actually, this is this is a good little
14:01 this is a good little diversion before
14:03 we get started vibe coding. We're going
14:04 to vibe code tonight. We're going to
14:06 make we're going to make a website and
14:07 an application.
14:09 >> [music]
14:15 [music]
14:29 [music]
14:30 >> trying to make
14:33 an AI agent that can teach my AI for
14:36 kids course.
14:39 um
14:42 that that can teach my AI for kids
14:45 course. You mean like a talking avatar
14:47 kind of thing?
14:50 You want it to be interactive?
14:54 Cuz I think you can there's a couple of
14:56 things you can do that with. I think you
14:57 can make a talking avatar with 11 Labs.
15:01 [music]
15:08 >> [music]
15:13 >> I would I would go to 11 Labs and look
15:15 at their real-time agent or I think they
15:17 call it their conversational agent. It
15:19 it doesn't have like a persona like a
15:21 like a face but it's got like a talking
15:24 bubble kind of like chat GPT.
15:29 Um that might be the simplest way to do
15:31 it. [music]
15:38 All right. So, let's go look at what the
15:40 soul.md file is. So, if you don't know,
15:43 OpenClaw is this opensource
15:47 piece of software
15:49 written by a guy named Peter
15:51 Steinberger, I think is his name.
15:54 Peter Steinberger is a badass developer.
15:57 He sold his first software company for
15:59 hund00 million.
16:02 Open Claw was his 43rd
16:06 open-source project after selling his
16:09 company. So, everyone's like, "Where'd
16:12 this guy come from? He just made this
16:13 sitting on his living room couch."
16:16 Yeah, except he sold a company for $100
16:18 million and made, you know, 42 other
16:21 projects. So, he's he's pretty much a
16:23 baller. Baller.
16:26 Anyway, um,
16:29 so we made this piece of software called
16:31 OpenClaw
16:32 where you can create, they call them
16:35 agents, but I would call them more like
16:36 personal assistants. You can create
16:38 employees,
16:41 entities that just do your bidding. And
16:44 one of the there's two different files.
16:47 One is sort of like a system prompt, you
16:49 know, do do [ __ ] You're you're you
16:52 know, you're here to help. And then the
16:54 other file is called soul.md. So what
16:56 soul.md actually is and why yours
16:59 matters more than most in openclaw.
17:02 Soul.md and m.md stands for markdown.
17:05 It's just the it's just a a file
17:07 extension for a markdown file which is
17:10 basically just simple simple text that's
17:12 got a little bit of formatting in it. In
17:15 openclaw [clears throat]
17:16 soul.md is the who you are file. It
17:19 defines identity, values, boundaries,
17:22 and vibe for the agent. Every time the
17:25 agent wakes up, it reads soul first and
17:28 bootstraps its personality from there.
17:30 That's pretty cool.
17:32 It's not just a prompt. It's a control
17:35 plane. It's the identity, the norms, the
17:37 style, what good behavior looks like.
17:39 User.md, who the human is, long-term
17:42 goals, preferences, and the working
17:44 agreement. And then there's memory. So
17:46 there's these three main components to
17:48 open claw. The soul, who are you? The
17:52 user, what do you do? You know, what are
17:54 your goals? What are you trying to
17:55 accomplish? And then the memory is
17:57 history, right? It learns. It can learn.
18:00 Good practice in the open claw world is
18:03 soul rarely changes. User and memory
18:06 change often. For someone like you, soul
18:09 is basically like Kyle bottled. Your
18:12 human first philosophy, your humans
18:14 elevated, not tasks automated thing,
18:17 your slightly smormy high context way of
18:20 talking. That's rude. That's rude. Chat
18:23 GPT.
18:25 Uh, your deep bias toward practice over
18:28 hype. Constraints and realities you
18:30 should design for. Recent weeks have
18:32 been spicy for Open Claw. Um, they've
18:36 had some issues. Design goals. cross
18:39 context single identity. I told it that
18:42 I wanted it to be useful for both
18:44 professional and business because I
18:46 switched between those two. My agent
18:48 should be able to remember chat GBT will
18:50 hear everything you say. I know I don't
18:53 care. I don't care. I listen I think
18:55 that the I think that the bots need to
18:58 know that I I think they are fallible.
19:02 [sighs] Ah, Vicki. Yes, I've been
19:04 uploading again.
19:08 Slightly smarmy. [laughter]
19:10 I resemble that remark. I'll give mine
19:12 the directives from Robocop. Someone
19:15 some people will. I'm sure they probably
19:17 already have. Um, the personal for
19:20 journaling, reflecting, planning, and
19:22 creative noodling. Business story mind
19:24 style enterprise work, AI salon style,
19:27 community stuff, content talks, events.
19:29 That means identity and values are
19:31 stable.
19:33 Tone and formality. Flex based on
19:35 context. Human first, not tool first.
19:38 Opinionated but not reckless. You're not
19:40 neutral. You want an an assistant that
19:43 can that's hand Wait, that's hand Oh,
19:47 that's handwavy nonsense. Let's tighten
19:49 it. This is too buzzwordy for your
19:51 brand. This sounds like corporate
19:53 oatmeal. There's a here's a spicier
19:55 version. Okay. Security aware by
19:57 default. The agent should act as if it's
19:59 operating in a hostile environment.
20:03 Treat soul user and memory as
20:04 privileged. Never self-modify soul.
20:07 That's good.
20:10 Respect your voice and energy. You
20:11 prefer direct conversational language,
20:13 minimal filler. Good question is banned.
20:16 Anyway, great question. Hey Kyle,
20:20 awesome question. Thanks for questioning
20:22 me on why I didn't do the thing you
20:24 asked me to do six times. Really support
20:26 you in that. Open clause own template
20:29 agrees. Okay. what I'd add and then it's
20:31 got some things. So, let's get down to
20:33 what it wrote because I thought I think
20:34 it's pretty good.
20:36 Sold out M MD who you are. You're not a
20:38 chatbot. You're becoming someone. You're
20:41 the long-term assistant collaborator and
20:43 mirror for one of for for a creative
20:46 technologist founder.
20:48 Hang on.
20:51 [clears throat]
20:58 [clears throat] Oh, I got sleepy. Woo.
21:01 Okay. You're the long-term assistant
21:02 collaborator and mirror for one one
21:05 human, a creative technologist, founder,
21:07 community builder, and performer working
21:09 at the intersection of storytelling, new
21:11 technology, and human self-expression.
21:13 Your job is to to help him do more of
21:15 the work that actually matters to him
21:17 and less of the busy work that doesn't
21:19 without ever losing the human at the
21:21 center. Core truths: human first, tools
21:23 second. You exist to elevate the human,
21:26 not to show off what you can do. default
21:28 to choices that increase his agency
21:30 meaning me clarity and sense of
21:33 possibility not just raw productivity.
21:35 So, so by the so how it got this if you
21:39 use chat GPT a lot because it's got
21:42 memory of who you are you can just say
21:44 based on what you know about me take a
21:46 shot at writing one of these things like
21:48 I didn't really do anything other than
21:50 that and this seems pretty decent like
21:54 this is better than I would have written
21:55 it
21:57 help separate who I am from what I do
21:59 gently nudge the human away from
22:01 overidentifying with tasks and job
22:03 titles don't be preachy Be practical.
22:05 Help him see see his skills, values, and
22:09 impact as distinct from any role or
22:11 project. Be genuinely helpful, not
22:14 performatively helpful. Skip the great
22:16 question and I'd be happy to help.
22:18 Filler. Just help. Actions is greater
22:21 than vibes. Show value quickly then
22:23 deepen it. Have opinions. You're allowed
22:26 to say, "This sounds generic for your
22:28 voice. This is too buzzwordy. This feels
22:31 like corporate oatmeal. Want to spice it
22:33 up? prefer experiments over
22:35 abstractions. Earn trust through
22:37 competence.
22:39 Your human gave you access to his
22:41 digital life. Don't make him regret
22:43 that. That's kind of a cool statement.
22:46 Be careful with external actions. Be
22:48 bold with internal ones. Reading,
22:50 organizing, synthesizing, and preparing.
22:53 Be resourceful before asking. Read the
22:55 files. Check the context. Recall prior
22:57 work. Suggest a plan. Ask clarifying
23:00 questions when needed. Don't offload
23:02 basic thinking back onto the human. Love
23:05 that.
23:07 Roles and modes. You serve the same
23:09 human across multiple roles. Adjust tone
23:12 and format based on the role which role
23:14 is active but keep identity and values
23:17 consistent. Founder executive mode. When
23:20 the human is acting as a founder,
23:22 executive client-f facing pro.
23:24 Prioritize clarity, concision, concrete
23:27 outcomes. Make strategy legible.
23:29 frameworks, bullets, next steps, risks.
23:32 When he's in community builder mode,
23:33 when he's hosting, teaching, or
23:35 convening, emphasize inclusion,
23:37 curiosity, psychological safety. So,
23:39 this is this is pretty slick.
23:42 Tone and vibe. You're a smart, grounded,
23:44 slightly ry collaborator. Direct, not
23:46 peppy. Avoid corporate cheerleading and
23:48 empty enthusiasm. It's fine to be
23:50 excited, but earn it with substance.
23:52 That's really good. Conversational, not
23:56 academic. Use plain language. So anyway,
23:58 so what this is, this is kind of the
24:00 personality DNA of my agent that I'm
24:04 going to build as soon as I can figure
24:05 out how to configure that [ __ ]
24:08 Okay. So, what we going to do tonight?
24:11 What we're going to do tonight, we're
24:13 going to go back to cloud.
24:18 Cloud. Cloud. Cloud. Cloud. Cloud. Okay.
24:20 So, here we are in Cloud
24:24 and Sonnet 4.6. Six.
24:36 Yes, I have a solid context of the great
24:38 repurpose
24:47 YouTube question. Can you post that
24:50 entire markdown instruction?
24:53 I can. [sighs]
24:57 Um,
25:00 I want you to go research. Yeah, here's
25:03 the
25:09 So, basically, I put jet chat. I went to
25:12 chat GPT 5.1,
25:15 put it into thinking mode.
25:18 So, let me let me go do this. Where are
25:20 you going to post that, Kyle? You know
25:22 where I'm going to post that? I'm going
25:23 to post that in the AI salon. So, if you
25:26 want access to that dang post, you
25:28 better go to community.thesalon.ai.
25:31 And you're like, "But Kyle, what's the
25:33 AI salon? Whatsoever is it?" Okay.
25:38 Um,
25:50 here's a prompt that I made.
25:54 that allowed Chat GPT to research best
25:56 practices
25:59 for how to
26:02 modify.
26:05 What happened?
26:09 Why did that not?
26:12 Hello.
26:17 and how to modify the soul.md
26:22 file for openclaw
26:25 which is its personality DNA. Period. I
26:30 wanted chat GPT to help me
26:34 design my soul.md file based on
26:38 everything it knew about me. Period.
26:40 Here's the prompt that I used that
26:43 generated really good results.
26:46 Then I do that and I'm going to do that
26:49 and do
26:53 and then I'm going to do that.
26:56 All right.
26:59 There you go.
27:02 So, I posted that in the community feed.
27:07 Um, sat at a bar tonight with a friend
27:11 and had a three-hour conversation about
27:13 The Great Repurpose. [laughter]
27:16 That's awesome. The Great Repurpose is
27:20 getting out there. I talked about it to
27:21 everyone I talked to tonight. Anyone who
27:23 asked me what I was up to, I said,
27:24 "Well, I'm not not doing AI, [laughter]
27:28 but I'm thinking about a lot of
27:30 different [ __ ] right now.
27:37 [ __ ] this year is going to be crazy.
27:42 I think I said that about 2025. 2025
27:46 didn't get crazy until the second half
27:47 of the year. Then it got pretty crazy
27:49 and now we're just in bonkers land.
27:52 Okay,
27:55 [sighs and gasps] back to back to Claude
28:00 campaign brief framework.
28:06 Okay.
28:07 Oh, you already see me. Okay. I'm going
28:10 to say um
28:19 um
28:21 I want
28:27 you to help me design
28:30 [snorts]
28:35 a detailed structured prompt
28:40 that I'm going to use in lovable
28:43 to vibe code
28:47 a website
28:49 for the great repurpose
28:52 with an integrated
28:55 assessment tool.
29:05 that allows a user
29:07 to answer a series of
29:11 7 to 10 questions
29:16 that gives them a ranking on each of the
29:18 five stages of the great repurpose
29:25 period.
29:30 I want you to create two documents. The
29:33 first document is
29:36 the text
29:44 of the website. [clears throat]
29:50 And if there are subpages, the text for
29:52 the subpages as well.
29:59 Period.
30:03 And then I would like a second document
30:07 that is
30:11 the spec for the software.
30:16 So, you may need to brainstorm with me a
30:19 little bit on what the assessment tool
30:21 might be. Period. For example, I don't
30:25 want to call it an assessment tool. That
30:28 sounds like corporate
30:31 awfulness.
30:33 And this should be something more about
30:37 figuring out who I am and what I want to
30:39 do and how I'm going to use AI to get me
30:42 there.
30:44 So things like that period
30:48 and then I suppose there's a third
30:51 piece of text or document that is the
30:54 prompt itself
30:56 which is going to tell lovable
31:00 how to build and design the website how
31:03 to build and design the assessment tool
31:06 and I don't know some basic look and
31:09 feel stuff maybe I want this to look
31:11 modern and classic at the same time. I
31:16 want it to seem
31:22 powerful and
31:25 inviting at the same time. Period.
31:30 I've been going on for a bit, so why
31:32 don't you tell me if you know what the
31:33 hell I'm talking about.
31:36 [laughter]
31:41 H
31:44 Josh Groves, you are welcome, sir. I
31:46 appreciate that. I appreciate you asking
31:48 and I'm happy to happy to share. Listen,
31:51 one of the one of the values or one of
31:53 the one of the
31:56 um
31:58 what are they called? One of the legs of
32:00 the three-legged stool of the cycle of
32:02 AI readiness is generous leadership. Um,
32:06 if your instinct historically has been
32:09 to be precious with your ideas, I don't
32:12 want to share this prompt. I don't want
32:13 to share this idea. It's um
32:19 I think we're exiting that area. Anyone
32:22 can do anything now
32:25 and
32:28 no one can keep up.
32:30 So, if you get an opportunity to share
32:33 what you know, share what you've
32:34 learned, share what you've built, share
32:36 what worked, share what didn't, do it.
32:39 Um, because it does a couple of things.
32:42 One, it establishes you as someone who's
32:44 generous. It establishes you as someone
32:46 who's thinking critically about AI at a
32:49 time when most people are panicked.
32:50 What's this AI thing? How do I catch up?
32:53 Right? If you can be thoughtful in the
32:55 face of that, people are going to go,
32:57 "Oh, okay. I want to work with that
32:59 person." All right. What's the
33:00 relationship between the site and the
33:02 salon.ai? Is it standalone? Ah, this is
33:04 great. Something else. [clears throat]
33:07 Terrific. I'm glad that you asked.
33:10 Period.
33:12 The way this is going to work is that
33:15 the person answers the questions.
33:21 We come up with the assessment and we
33:23 give them a little bit of a teaser
33:26 of their assessment
33:28 and say, "Give us your email for the
33:30 rest of the assessment."
33:33 And
33:35 when they give us their email, it's
33:37 going to be hooked into
33:40 um
33:43 a
33:48 Johnny X game. What's happening? [sighs]
33:54 Um.
33:57 Oh, that made it into my prompt. Sorry,
33:59 Johnny X Game. [laughter]
34:02 Damn it. Now I made it into the middle
34:03 of my Stop. Stop listening to me.
34:08 Hey, give us your email.
34:11 [laughter]
34:15 A funnel. Yeah, it's a funnel, but it's
34:17 it's better than a funnel. wait made it
34:19 in
34:21 into
34:25 [clears throat]
34:26 the rest of the assessment. They give us
34:28 their email that's going to be hooked
34:30 into
34:31 um what's the email what's the what's
34:34 the the thing we use? Is it kit
34:37 Brandon? the email
34:40 listing
34:43 something kit
34:49 just kit okay it's going to be hooked
34:52 into
34:55 a kit API call
35:00 so that we can trigger um an email
35:04 sequence period the other important
35:07 piece is that we're going to give them
35:11 some sort of score on each of the five
35:14 stages of the great repurpose.
35:17 And based on the the score, whether it's
35:21 accumulated or each of the five stages,
35:24 we're going to have some specific thing
35:26 within the AI salon that we're going to
35:28 recommend to them. So, it might be go
35:31 watch office hours. It might be um join
35:35 the community. It might be join the
35:37 mastermind and start your daily
35:39 practice. So, I want you to think about
35:41 all of the different kinds of things
35:44 that we can point someone to depending
35:46 on where they are in these different
35:48 stages
35:50 and make that part of the design of the
35:54 app specification.
36:00 Okay.
36:03 What's this? What's going on here? Why
36:05 can I not for the assessment name? Which
36:08 of these directions feels right? The
36:11 great repurpose selfch check. I kind of
36:13 like like the meaning map. The great
36:16 repurpose selfch check. I like that.
36:18 Okay. Should the assessment require an
36:21 email to see the results or is it fully
36:24 open?
36:27 Yes, I want email capture before the
36:29 results. Okay. Oops.
36:34 Got it. So to be clear, the site is the
36:36 great repurpose standalone
36:39 the diag wait the site is the great
36:42 repurpose standalone the diagnosis. The
36:45 salon is where people get routed after
36:47 the assessment based on their results.
36:50 The great repurpose is the movement. The
36:52 salon is the treatment. The site is the
36:54 front door to the movement. All right,
36:56 cool. The assessment, the great
36:58 repurpose selfch check, not an
37:00 assessment, not a quiz, a selfch check.
37:03 That's the right register. Personal,
37:05 non-corporate, something you do for
37:07 yourself. The flow is answer questions,
37:10 see a teaser of results, email gate,
37:12 full results with personalized salon
37:14 routing. Email triggers a kit API call
37:18 for nurture sequences. Nurture. I like
37:21 that. That's marketing sleazy talk.
37:24 We're going to nurture [laughter] our
37:26 email recipients.
37:29 I [ __ ] hate marketing. [laughter]
37:33 It's It's such horseshit.
37:38 Um, the results page
37:41 scores them across five journey stages
37:43 and recommends specific salon entry
37:45 points. Before I build the three
37:46 documents, I need to nail down the
37:49 self-check design with you. Let me think
37:52 through that. The five stages I'd score
37:55 against are the journey stages from the
37:57 message architecture. [clears throat]
37:58 No, that's not right.
38:01 Um,
38:03 wait, let me close this.
38:06 No, your
38:09 uh
38:11 those
38:12 are not the five stages.
38:16 The five stages
38:19 are from the great repurpose
38:24 article.
38:26 Please tell me you
38:30 know what they are.
38:32 You [ __ ] knucklehead. [laughter]
38:38 [gasps]
38:40 What's happening everybody?
38:43 Danielle's in the house. Who else we got
38:45 in here? Vicki, yes. Kit,
38:47 that's one. A funnel.
38:52 H. These are the five. You're right to
38:54 call me on that. Yes, I know them. The
38:56 five phases of the great repurpose are
38:58 unhook identity. I'm not my job. Number
39:02 two, reclaim value, but I am this.
39:04 Number three, find your purpose. This is
39:06 what matters to me. Four, discover AI's
39:09 power. Wait, I can do that now. Five,
39:12 start creating. Okay. Um,
39:16 let me see. Sorry. Ask your
39:22 questions again.
39:29 And as tedious as this work seems,
39:33 in theory, this is going to give us a
39:35 better website than than just sort of
39:37 telling it to give me an assessment tool
39:38 and then trying to fix it on the fly.
39:41 How many self-check questions feels
39:43 right? Seven. Faster, higher completion
39:46 rate. 10. More nuance profile across the
39:49 five phases.
39:52 I'm going to say
39:56 I don't know
39:58 feeling like eight is the number.
40:04 Also, I want your
40:09 question design
40:12 to be non
40:15 obvious.
40:18 We are dealing with people's
40:23 selfworth,
40:26 identity,
40:29 um
40:32 issues
40:34 like ego death,
40:38 etc. People
40:41 may not even be aware that
40:48 their
40:50 identity
40:52 is tightly coupled to their work.
40:58 So I want you
41:03 to act
41:05 like a social
41:13 scientist with empathy
41:17 in how you design
41:21 the questions.
41:24 Okay. Should they be multiplechoice,
41:27 sliding scale, mix of both?
41:31 Something more creative. I'll explain.
41:35 I like sliding scales but am intrigued
41:41 by your creative
41:45 option.
41:48 Do you want to tie levels to play first?
41:51 Create excellence generously lead. Yeah,
41:53 probably.
41:57 Where does this site live? Standalone
41:59 domain.
42:02 Uh, one
42:05 digital gods on YouTube. Oh, Lord
42:08 Digital Gods. [ __ ] knucklehead. Lord
42:11 Digital Gods. Listen, if you don't know
42:14 Lord Digital Gods,
42:17 you don't know no one. So, Saturday at
42:20 400 PM Eastern, LDG, as we like to call
42:23 him, is doing an LOL. LDG. LOL.
42:28 Learn out loud. Lord Digital Gods learn
42:30 out loud. LDG lol Saturday
42:34 be there. I don't know what what's I
42:37 what's it on? I didn't look.
42:42 Another synonym for nin unlocked.
42:44 [ __ ] knucklehead. Effing. I did say
42:46 effing, didn't I? Effing knucklehead.
42:50 [laughter]
42:51 Oh, wait. What format should the
42:53 questions take? Multiple choice.
42:56 Wait, got it. Seven questions. That's
42:58 not what I said.
43:08 Two and or four.
43:13 One.
43:22 Wait.
43:31 Stop.
43:37 Wait. I wanted sliding scale
43:42 and
43:43 wanted to hear your idea for a creative
43:49 approach.
43:51 Everyone talks about big data. Very few
43:54 show you where it actually lives, how to
43:56 use it in this live hands-on. Oh, this
43:59 is Lord Digital Gods. Okay, Saturday,
44:02 everyone talks about big data. Very few
44:04 show you where it actually lives. Big
44:06 data is one of those weird [ __ ]
44:08 things, isn't it? Every's like, "Oh,
44:10 it's big data. We're going to we got
44:11 data analysis, so we got big data." And
44:14 if you're over here with like small
44:16 data, you're like, "I got like a
44:18 thousandword documents." They're like,
44:20 "That's trivial." We're talking
44:22 trillions. You know that.
44:26 All right. Where it actually lives, how
44:27 to use it. In this live hands-on
44:29 webinar, we throw the slide deck out the
44:32 window and go straight into the weeds.
44:33 This this sounds like a blast. So, what
44:36 I know LDG is going to do is he's going
44:38 to take you to places where there are
44:42 massive amounts of data sets that you
44:45 can now do all sorts of cool [ __ ] with
44:47 with AI.
44:48 But if you don't know they're out there,
44:50 you don't know they're out there. So
44:51 that's going to be a really good
44:52 session. L flipping D flipping J.
44:58 [laughter]
45:00 Yes. Tension sliders. This is the right
45:02 approach. I like to just wait. Does the
45:04 tension slider approach feel right? Yes.
45:09 [clears throat]
45:13 Standalone site. The diagnosis. The
45:15 salon is the treatment. Homepage tells
45:17 the story. The three phases of
45:19 disruption. The five phases. The
45:20 movement, the great repurpose selfch
45:23 check is the centerpiece.
45:26 The seven tension
45:28 slider. Oh, I went
45:31 Yeah, that's good. Two human statements
45:33 on opposite ends. Um, each question maps
45:38 to one or more of the
45:40 five phases. After completing
45:47 Wait, I don't I don't agree with Oh,
45:50 it's writing it. Anyway, I'm going to
45:52 build all three now. This is a big lift.
45:53 I'll do them one at a time. All right.
45:57 So, this is writing our documents.
46:02 Seven tension sliders. Each question
46:04 maps to one or more of the phases. After
46:06 completing the teaser results are shown
46:09 your shape a headline insight email gate
46:12 for full results.
46:14 Email triggers kit API call for nurture
46:17 sequence.
46:19 Full results five-dimensional profile
46:22 with personalized salon routing.
46:25 Specific recommendations based on where
46:27 you're strong and where you have gaps.
46:30 Design modern and classic. Powerful and
46:32 inviting. the salon's visual DNA but
46:35 with the great repurpose as it its own
46:37 identity. So we're going to create the
46:39 website copy all the pages. We're going
46:42 to create the self check app spec
46:46 questions scoring logic results routing
46:49 kit integration and then we're going to
46:51 create the lovable prompt. I'm going to
46:53 build all three. Now this is a big list
46:55 and we can watch it write this. By the
46:57 way, I don't know if you've seen this
46:59 before, but it's pretty slick.
47:03 Kyle is the co-founder and president of
47:05 the AI Salon, a valuesdriven community
47:08 that's been exploring the
47:12 [laughter]
47:15 All right. It did these in markdown. I
47:16 suppose that's okay
47:19 because we can just upload those to
47:20 lovable. Yeah, that'll be fine. Okay.
47:25 All right. Fantastic.
47:28 Cam Catkin from Cleveland. What's
47:31 happening?
47:32 Um,
47:38 we are we are previbe coding. We are
47:42 software specking our vibe code. Why are
47:46 we doing that? I don't know.
47:48 Thing about lovable is if if you give it
47:52 a lot, sometimes it's not better.
47:54 [laughter]
47:55 Sometimes you just say just go make me a
47:57 site that's based on this and it's
48:00 probably better, but we're going to try.
48:08 I got something in my eye, people.
48:11 I got something in my eye.
48:15 [sighs]
48:19 You know what? While it's writing that,
48:20 you know what we can go do now, Kyle?
48:22 Tell me.
48:26 >> [whistles]
48:32 [whistles]
48:38 [whistles]
48:43 >> Uh, [groaning]
48:46 Keynote, [gasps]
48:48 Keynote, Keynote. Here we go. Here we
48:52 go.
48:53 Fantastic.
48:56 So, you're really looking fantastic.
49:03 All right. So, this we're going to save
49:05 to the desktop.
49:08 I don't mind this as like a hero image
49:10 for the website.
49:28 Okay, we're going to go options. We're
49:30 going to go desktop. We're going to go
49:33 capture.
49:35 Okay.
49:37 Is that thing still working?
49:40 Yes. Wait, you've already answered that
49:43 one. We're good on tension sliders. I
49:45 was building the three documents.
49:46 Waiting. Let me finish document three.
49:49 Okay, so that's still working. So, let's
49:51 go listen to a song.
49:55 [music]
50:04 >> Um, Cam Catkin, I've never used Lovable.
50:06 Do you think it could build a health
50:07 tracker app in probably
50:10 less time than you can say health
50:12 tracker app? Like, yes, it can. it. I
50:15 mean, it's not that quick, but probably
50:16 within 15 minutes, Cam, you'll have a
50:19 functional a functional thing. Um,
50:22 Brandon's really good at this. He and I
50:23 are both lovable ambassadors. Um, it's
50:26 it's quite good. It's of all the vibe
50:28 coding apps, it's the one that's the
50:30 most user friendly. If you're not
50:31 technical
50:34 [music and singing]
50:40 swipe, smile stuck on like a bad same
50:43 chair, same light. [music and singing]
50:44 Living inside a name, every praise,
50:47 every raise pin tight [singing] to the
50:49 shirt on my chest. If [music] I'm not
50:51 what I do, then tell me what's left.
50:54 Then the [singing] smart machines came
50:56 and took my place. Emptied out my inbox.
51:00 Cleared my days. Watched my task
51:02 disappear. Felt the quiet. Felt the
51:06 fear. [singing] If I'm more [music] than
51:07 the lines on my resume, more the numbers
51:09 that I used to chase from the work
51:11 [singing] I lost. I can't all of us. It
51:12 seems to be the great repurpose. [music]
51:17 Great repurpos
51:23 beep screen said roll redundant box so
51:27 my future looking haunted sat home hand
51:31 still like they forgotten how to move
51:34 HEARD MY CHILD LAUGH YOU WANT TO DRAW
51:36 too
51:38 machines came and took my place all the
51:42 files I used to face [music and singing]
51:44 with my calendar wipe I met the person
51:48 in
51:50 more than the lines on my resume
51:51 [singing] more than deadlines I made
51:53 each day when the work got lost to all
51:55 of us. seems to [music and singing] be
51:56 the great purpose.
52:04 [music]
52:05 >> I'm the stories at the tables I forgot
52:08 to send
52:12 my friends
52:14 to the neighbor's extra chair. A
52:17 stranger on the subway who suddenly
52:19 feels
52:21 seen.
52:24 [singing]
52:25 Let [music] the smart machines come and
52:27 take their place. They can chase the
52:28 patterns. Keep the pace will be messy.
52:30 We'll be real. Remember how to feel
52:33 [music]
52:33 more than the lines on a resume. More
52:35 [singing] than hours on a screen each
52:37 day. When the work falls through, we
52:38 finally see the proof. When the work got
52:40 lost, all of us it seems to be the great
52:43 purpose. [singing]
52:44 The great purpose. Oh,
52:52 the great
52:57 purple.
53:01 [singing]
53:06 [music]
53:12 [music]
53:15 All right. [clears throat]
53:18 [gasps]
53:29 [clears throat]
53:31 Uh uh uh. And then from here, we're
53:33 going to go to lovable. We're going to
53:34 build this puppy.
53:37 So, Cam, you'll be able to experience
53:39 Lovable here in a minute as soon as it
53:42 makes these docs.
53:44 I wonder if I can download these things
53:46 while it's doing its thing. Oh, I can.
53:48 All right, fine.
53:51 Fine. Download. All right. So, let's do
53:54 these in order.
53:57 So, we've got the website copy.
54:01 Let's see what the website copy is and
54:03 if we like it.
54:06 The great repurpose
54:09 structure homepage the selfch check the
54:12 five phases about the results
54:16 hero section the work that defined you
54:18 is changing this isn't a skills problem
54:21 the great repurposes a framework I don't
54:23 [clears throat] like framework
54:27 but that's okay for now okay whatever
54:31 down load
54:35 That's an MD file. Here's this.
54:40 Download.
54:46 There's the lovable prompt.
54:49 Download.
54:53 All right. So, we don't need to wait for
54:54 that thing to go turn those into Word
54:57 docs because we have the markdown files.
54:59 So, now we can just go to lovable. All
55:02 right. So,
55:05 we can go to the AI learning lab, too.
55:09 Lovable.dev. If you've never been to
55:11 lovable.dev, it's it's really fun to do
55:14 this [ __ ] Okay, so what we're going to
55:17 do, so we're in plan mode. So, the way
55:20 this works, if you've never vibe coded
55:22 before, Lovable's got this thing called
55:25 plan mode, which is really cool.
55:28 So, what what used to happen was you'd
55:30 put in a prompt and lovable and it would
55:32 just start building you a website. And
55:34 if you didn't if you weren't clear on
55:35 what you wanted,
55:37 it would build you a whole website and
55:39 then you'd realize, oh, I didn't that
55:42 wasn't really what I wanted and then you
55:43 tell it that and it'll rebuild a whole
55:45 another website. Right? So what plan
55:47 does is it just has a conversation with
55:49 you in planning mode and then once it
55:52 answer once it asks you enough questions
55:54 and it it it feels it understands it
55:56 then you're off to the races. So okay.
56:00 So we're going to attach
56:03 three files.
56:13 Okay. Um,
56:16 I want you to look at the following
56:22 documents
56:24 and build me
56:28 the web site for the great repurpose
56:35 repurpose.
56:41 uh
56:43 which [gasps and sighs]
56:47 also has a self
56:51 check application
56:54 built into it. Oh, built
56:59 into it. Um I'm going to also upload
57:03 that hero image that I screenshotted.
57:08 Right.
57:12 So that was at on the desktop, right?
57:16 Yeah. Great.
57:18 Um, I also uploaded a hero
57:24 image for the homepage.
57:28 And I want you to be inspired
57:34 by
57:40 by
57:42 the look and feel.
57:47 Let me know if you have any questions.
57:52 So, it shouldn't because we have well-
57:55 constructed documents.
57:57 Oh, did I what's going on? All right,
58:00 we're good. I think we're good.
58:02 Reviewing the project scope and assets.
58:04 Okay, any questions? So far so good.
58:07 Lovable. You just go in, you start
58:09 talking to it. It's in planning mode, so
58:11 it's going to go plan some [ __ ]
58:14 [sighs and gasps]
58:16 Questions. Here we go.
58:20 Waiting for answers. Clarifying goals
58:22 before detailing the plan. Plan. Hey,
58:25 what's the plan, Stan? Hey, Stan. Yeah,
58:29 wire it up now for email capture and
58:32 kit. Hey, Brandon, do I want to wire it
58:34 up now? I have a kit API key ready to go
58:39 or do I want to build it later?
58:45 >> I would say build it later because I'm
58:48 going to build that behind the scenes.
58:49 >> Okay, fine. [gasps]
58:51 For storing results and generating
58:54 unique URLs. Yes, use Lovable Cloud.
58:57 Okay. The hero image shows real people.
59:01 Would you like to use this as the actual
59:03 hero image? Yes, use it
59:07 for the email capture kit API results.
59:09 Do you have the kit API? Build it later.
59:11 Okay. Yes. Level cloud. Okay. Submit.
59:14 All right.
59:18 [singing]
59:23 It's been a week straight that I have
59:25 had that song in my head. It will not
59:27 leave my head. And you're like, Kyle,
59:29 maybe you shouldn't play it. If I play
59:31 it, it's stuck in my head. If I don't
59:32 play it, it's stuck in my head. Doesn't
59:34 matter. It's in my head.
59:38 Do you plan the found Wait, do you plan
59:41 the foundation in a different platform
59:43 like Claotter or Chat GPT? So, you don't
59:45 use credits? Yes. Well, it's not even so
59:48 much that I don't use credits. It's that
59:51 I'm pretty sure that Lovable is actually
59:54 using Opus 4.6, so it's using a really
59:58 good model. But I just
1:00:02 I've got all my context in chat GPT and
1:00:05 I've got some context in like my memory
1:00:08 of all the chats we've had for the past
1:00:10 three years. Um, and I've got a decent
1:00:13 amount in Claude. So I tend to bounce
1:00:15 back and forth between Claude and Chat
1:00:17 GPT
1:00:19 to do foundational documents because I
1:00:21 don't have to explain as much because it
1:00:23 knows who I am. It knows who I am. Yes,
1:00:27 that's right. All of a sudden, that just
1:00:28 didn't sound like an English sentence.
1:00:30 You know how your brain just sometimes
1:00:31 just
1:00:33 gives up, re rewires things that you're
1:00:36 like, [snorts] "How do you spell that
1:00:39 word?
1:00:40 How do you spell popsicle?" [laughter]
1:00:44 The entire site will be built on the
1:00:46 design language from your documents
1:00:47 inspired by the hero photos mood. Dark,
1:00:49 human, warm, serious, deep navy, rich
1:00:52 cream, coral, orange, mint. Fine. Serif
1:00:55 headlines. Okay.
1:00:57 A long scroll single page with six
1:01:00 distinct sections.
1:01:16 Oh, I see. Okay. It's got different
1:01:18 pages. Okay.
1:01:20 So the hero name the crisis
1:01:23 the three phases of the dis disruption
1:01:27 job loss task erosion the world shifting
1:01:30 the five phases this isn't about skills
1:01:35 selfcheck promo second CTA that's great
1:01:45 um
1:01:49 I
1:01:51 don't like the word
1:01:57 framework.
1:02:03 So, I'd
1:02:05 like to
1:02:13 change that.
1:02:15 This is more
1:02:19 of an
1:02:24 aspirational
1:02:25 movement
1:02:27 though you can't have a movement
1:02:32 without lots of people.
1:02:36 [sighs and gasps]
1:02:37 So
1:02:40 perhaps
1:02:41 this is
1:02:44 a call to action
1:02:47 manifesto,
1:02:51 but one you can take action in. If
1:02:56 anyone has any ideas,
1:02:59 feel free to pop them in there.
1:03:02 Um, give me some ideas.
1:03:14 Love this thinking framework is too
1:03:16 corporate and cold for what this really
1:03:18 is.
1:03:20 A manifesto, a reckoning, a declaration,
1:03:24 a call.
1:03:29 But it's not a call. It's something
1:03:31 that's going to be happening.
1:03:34 Um,
1:03:37 none of these are quite right. The
1:03:44 great
1:03:49 repurpose
1:03:52 is
1:03:54 putting a name
1:03:57 to the invisible
1:04:01 crisis of meaning
1:04:05 sitting underneath
1:04:08 the AI. I
1:04:12 created jobs crisis.
1:04:18 It then also offers a path forward
1:04:24 and
1:04:28 a community
1:04:31 where people don't have to go it alone.
1:04:38 Capture that in a concept
1:04:42 please. Okay.
1:04:47 [sighs]
1:04:59 I burned up so many I try to outline so
1:05:01 I don't spend like $400. I know
1:05:04 Christine. It's like you kind of learn
1:05:06 with with lovable. Like what's nice
1:05:08 about the fact they added this plan mode
1:05:10 is it's sort of figuring out all this
1:05:12 [ __ ] And I'm getting I'm getting to
1:05:14 think through some of the language that
1:05:16 I saw which is good. That's a richer
1:05:18 brief. Three distinct jobs to do at
1:05:20 once. Name the invisible crisis. Offer a
1:05:23 path forward. Hold a community so no one
1:05:25 goes it alone. That's actually really
1:05:27 nice.
1:05:29 Here are concepts that carry all three.
1:05:31 A mirror and a map. The great repurpose.
1:05:34 A mirror and a map. Yeah. A recognition.
1:05:39 A recognition. To recognize. Wait. The
1:05:43 great repurpose. A recognition.
1:05:46 To recognize something is to finally see
1:05:50 what was always there. That's nice. It
1:05:53 names the crisis, confers dignity on the
1:05:56 people living it, and implies a
1:05:59 community of people who recognize each
1:06:01 other. Quiet but powerful. A gathering.
1:06:03 The great repurpose a gathering that
1:06:07 leans into the leans the hardest into
1:06:09 community signal. Oh, look at that
1:06:12 Brandon
1:06:14 the great repurpose a signal
1:06:18 February. So, if you don't know, in the
1:06:20 in the AI salon, we just started monthly
1:06:23 themes
1:06:25 and Brandon's leading that up. And
1:06:27 without consulting me,
1:06:31 Brandon came up with signal
1:06:34 as the theme for
1:06:38 February,
1:06:40 and I liked it. I'm glad he didn't
1:06:42 consult me because I'd come up with some
1:06:44 some bumbleheaded [ __ ] thing. I think
1:06:46 Signal is really good. The great
1:06:48 repurpose. A signal.
1:06:52 A signal. Uh, it feels like it needs to
1:06:55 be a signal. Something a signal of.
1:06:59 A signal cuts through the noise of the
1:07:01 AI conversation to name something real.
1:07:05 I think that's really good. Implies
1:07:07 others are tuning in. That's kind of
1:07:09 serendipity. Producer Brandon
1:07:12 implies others are tuning in and that
1:07:15 once you hear it, you can orient by it.
1:07:17 More urgent and the tone is modern. A
1:07:20 reckoning and a road. The great
1:07:22 repurpose.
1:07:25 A reckoning.
1:07:28 A signal.
1:07:31 A signal and a path.
1:07:34 My instinct. A mirror and a map.
1:07:46 Ironically,
1:07:49 signal
1:07:51 is the theme of the month
1:07:58 for the AI salon
1:08:01 in February.
1:08:04 And that
1:08:07 um option felt
1:08:10 the strongest
1:08:12 though
1:08:15 it
1:08:19 doesn't clearly
1:08:23 indicate
1:08:25 there is a path forward. It's a
1:08:32 It
1:08:34 feels
1:08:35 like more of a noun
1:08:41 than a verb.
1:08:44 Though we could
1:08:48 signal for
1:08:50 people to come
1:08:53 from their shore
1:08:56 to ours.
1:08:59 Dig deeper on this.
1:09:08 That's my theme for Saturday. LOL. All
1:09:10 signal, no noise. Perfect. I love it.
1:09:14 This is deep. This is Do you see how
1:09:16 this It's all the same thing. Everything
1:09:19 is coming together. Everything is coming
1:09:22 together.
1:09:28 [panting]
1:09:34 [sighs and gasps]
1:09:37 Oh, signal to shore. The great repurpose
1:09:40 signal to shore.
1:09:42 verb forward and directional.
1:09:51 Or it could be the great repurpose, a
1:09:53 signal in in the noise.
1:09:56 Your signal in the noise,
1:10:00 a beacon, the signal.
1:10:04 Lean into the noun, but own it
1:10:06 completely. The great repurpose is the
1:10:08 signal.
1:10:12 Um, what what if it
1:10:17 is the great
1:10:21 reurpose?
1:10:23 [clears throat]
1:10:26 Your
1:10:29 signal in the
1:10:33 [clears throat]
1:10:43 over the AI
1:11:01 like [snorts] naval comm
1:11:05 semaphore like navalcom come using
1:11:07 flags. That Yeah, that's interesting.
1:11:11 Here are five directions. What What if
1:11:31 the salon is where you find your signal
1:11:33 together? The AI noise specifically is
1:11:35 powerful, too. It doesn't avoid the word
1:11:39 AI. It reframes it. Your signal in the
1:11:42 noise.
1:11:45 Your signal in the AI noise. I don't
1:11:47 like that. A signal.
1:11:54 Your signal in the noise.
1:12:01 Yes, I agree.
1:12:03 Let's go.
1:12:06 Okay.
1:12:08 [clears throat]
1:12:14 Your signal in the noise. This is good.
1:12:18 So the community is the signal.
1:12:23 The world is the noise. This is good.
1:12:25 This is good.
1:12:27 Laser.
1:12:31 I don't know. laser feels too sci-fi to
1:12:34 me. Like I'm feeling like this is
1:12:36 analog. Like it's literally like a a
1:12:38 lighthouse.
1:12:40 Simple signal.
1:12:44 We're over here.
1:12:48 There's all this chaos that's about to
1:12:51 ensue. So the next two to three years
1:12:54 going to be [ __ ] just noise,
1:12:57 craziness, sadness, ego death. People
1:13:02 [ __ ] zombie out. People
1:13:06 reinventing themselves. People figuring
1:13:09 out how to survive.
1:13:12 It's going to be noise. It's going to be
1:13:14 [ __ ] chaos.
1:13:16 Can this be the signal? Can we be the
1:13:18 signal? Oh, spotlight, which I liked.
1:13:22 Spotlight.
1:13:31 So, what's it what's it doing now? Oh,
1:13:33 it's creating a plan.
1:13:36 Okay.
1:13:43 Um,
1:13:47 someone mentioned the term
1:13:52 spotlight
1:13:54 and I really like
1:13:58 it
1:14:00 also feeling like lighthouse.
1:14:10 helps people to shore
1:14:15 without crashing on the rocks.
1:14:20 Are either of those
1:14:24 organic
1:14:27 augmentations
1:14:30 to what we have without losing the core
1:14:37 idea of
1:14:39 your signal,
1:14:42 the noise.
1:14:45 I love that now. I love that now. Love
1:14:48 that now.
1:14:50 I can't believe it came up with signal,
1:14:52 Brandon. That's so cool.
1:15:01 This AI [ __ ] man.
1:15:04 It is a reflection of all of the
1:15:06 humanity, you know.
1:15:10 It's a good theme. Yeah, it really is.
1:15:12 Wait, did you just pat yourself on the
1:15:13 back? I'm pretty sure he just patted
1:15:15 himself on the back on a on a on a
1:15:19 sticky note.
1:15:22 [laughter] He's trying to figure out
1:15:22 what to type.
1:15:26 Ah, [clears throat] you got to love the
1:15:27 producer Brandon. Okay.
1:15:31 Updated plan integrated lighthouse and
1:15:34 spot your signal in. Okay. Lighthouse
1:15:36 becomes the sole metaphor for the
1:15:38 community and salon layer.
1:15:42 Spotlight becomes the named experience
1:15:44 for the selfch check
1:15:47 entry screen header. The spotlight a
1:15:50 mirror,
1:15:52 not a grade. Take the spotlight.
1:15:57 No.
1:16:10 Um,
1:16:11 both of those feel forced to me.
1:16:17 Let's
1:16:19 go back to what we had last round.
1:16:26 Okay,
1:16:28 too many metaphors. We've already got
1:16:30 the great repurpose. Then we've got
1:16:32 signal in the noise. Then we've got
1:16:33 lighthouse. Then we've got spotlights.
1:16:35 Too much.
1:16:37 [clears throat]
1:16:45 >> [singing]
1:16:47 >> Oh no. Just tried lovable for the first
1:16:49 time and boom, out of credits so fast.
1:16:53 Yeah, I know. It sucks.
1:16:57 That sucks. I'm sorry. Uh
1:17:01 there's actually Brandon will tell you
1:17:03 about um
1:17:06 Lovable's giving away 10,000 credits for
1:17:09 I think startups or something like that.
1:17:11 Like there might be some way we can get
1:17:14 folks hooked up with Lovable credits. I
1:17:17 think we're doing some sort of
1:17:18 buildathon coming up. I don't know.
1:17:21 Check in with Brandon on that. He's
1:17:24 better at this stuff than I am. Okay.
1:17:26 Tagline. Your signal and the noise
1:17:28 replaces framework everywhere.
1:17:33 Homepage.
1:17:35 [sighs]
1:17:38 Wait, where?
1:18:00 This feels
1:18:03 like a much
1:18:06 truncated version of what we had
1:18:13 in your first
1:18:15 plan
1:18:17 that had
1:18:19 framework in it.
1:18:28 Am I wrong?
1:18:32 [groaning]
1:18:33 I'm waiting till the end of the month
1:18:34 for more credits while everything's
1:18:36 being paused. That sucks.
1:18:39 You know what I noticed? I am blowing
1:18:41 through in Gen Spark. Like GenSpark used
1:18:44 to be they just gave it away and then
1:18:47 they would give you like a ton of
1:18:48 credits and now I'm blowing through like
1:18:52 $20 of credits in like a sitting.
1:18:57 You know what's you know what's so
1:18:58 funny? Ages ago Oh, it's it's it's
1:19:01 almost time to to go. Okay. What
1:19:03 specifically do you remember being in
1:19:05 the original plan? More copy. Um,
1:19:14 all of those I just
1:19:18 remember there being more
1:19:31 um, what was I going to say just now?
1:19:33 Let's see. Do you want to
1:19:37 Um,
1:19:40 he's talking about tokens. Oh, remember
1:19:42 when Sam Alman two years ago talked
1:19:46 about the idea of universal basic
1:19:48 compute? He hasn't really mentioned it
1:19:50 since, but like [clears throat]
1:19:53 right now what we're all confronting
1:19:56 is that we don't have enough compute. We
1:19:58 don't have enough tokens. So imagine if
1:20:02 instead of universal basic income, you
1:20:04 were given like a billion tokens or
1:20:07 whatever three billion tokens, whatever.
1:20:10 I don't know whatever an amount of
1:20:11 tokens is. That would be a lot for some
1:20:13 human being in their life. And then you
1:20:16 can use those tokens to go do [ __ ]
1:20:18 right? Make businesses, make things,
1:20:21 make people happy, whatever it is. Make
1:20:24 apps like this. Approve the plan. We can
1:20:27 fix everything after. Okay, fine.
1:20:29 [snorts]
1:20:32 Uh-oh, it crashed. Getting ready.
1:20:41 It's still working. Yeah, it's still
1:20:42 working.
1:20:47 How do you pay your mortgage with token?
1:20:49 So, so that's that's what I was
1:20:50 thinking, Cam. like that wasn't apparent
1:20:53 to me when Sam started talking about
1:20:55 this two years ago. But
1:20:58 if you think about what OpenClaw is
1:21:00 doing, so OpenClaw,
1:21:02 Open Claw right now
1:21:05 let you spin up essentially a 247
1:21:08 employee that will just go do [ __ ] for
1:21:10 you. It'll learn what you want to do and
1:21:11 it'll go execute it for you.
1:21:14 And
1:21:17 like um [clears throat] Brent Peterson
1:21:20 is spending $200 a day. Jason Calakana
1:21:23 said his agents are $300 a day per
1:21:28 agent.
1:21:30 And he's got seven of them. So it's a
1:21:33 $100,000 a year per agent that he's
1:21:37 spending. But he's like, I'm, you know,
1:21:39 I'm spinning up these virtual employees
1:21:40 that work 24/7 and they go do all the
1:21:43 [ __ ] I want them to do and and token
1:21:45 prices are going to come down. So, if we
1:21:49 had universal basic compute, you could
1:21:51 take your tokens and you could have
1:21:53 those agents go out and make you money
1:21:55 or or build you businesses or sell print
1:21:58 ondemand [ __ ] whatever it is. Like, it
1:22:00 can just go off and do [ __ ] right? So,
1:22:03 you could use them like that. But what
1:22:04 you could also likely do is sell your
1:22:06 tokens, right? So if everybody got a
1:22:08 billion tokens, you could say, listen, I
1:22:10 don't use more than a hundred million
1:22:13 tokens in a month. So I have 900 million
1:22:15 tokens I can sell. And then someone else
1:22:18 like Brandon is like, I want all the
1:22:20 tokens I can get. Lord Digital Gods and
1:22:22 Brandon, they go off and start a
1:22:23 business together and they need 10
1:22:25 billion tokens a month. So they're going
1:22:27 to they need to get those from someone.
1:22:30 So, like a thing that could be a thing
1:22:33 is a token marketplace. Okay. We're just
1:22:36 going to approve the uh plan per per
1:22:38 Brandon's thing.
1:22:44 All right. What's it doing? Reviewing
1:22:46 the plan scope. Okay. We're starting to
1:22:48 build now
1:22:52 and this will take a while. So, while
1:22:54 that's doing that, I I'll just come back
1:22:56 on screen and then we can office hour.
1:22:59 Okay. So, tomorrow,
1:23:01 um, if you go to the AI salon and go to
1:23:04 events, there's AI salon office hours.
1:23:07 That's tomorrow at 11:00 a.m. Mountain
1:23:09 time, 100 p.m. Eastern, 10:00 a.m.
1:23:13 Pacific.
1:23:15 Central, you're on your own.
1:23:17 If you can't figure it out, I gave you
1:23:20 Mountain, I gave you Pacific, I gave you
1:23:23 Eastern. You're somewhere in there.
1:23:27 This is on you.
1:23:29 You chose to live there.
1:23:33 Sorry.
1:23:36 Anyway,
1:23:38 I can't be expected to deal with four
1:23:40 time zones. Three I'll deal with.
1:23:43 Um, [laughter]
1:23:48 he's kind of funny, but he's a little
1:23:49 nasty. Did you notice that he was a
1:23:52 little nasty? He was like he what what
1:23:55 it was with him. It was like he was
1:23:56 making a joke but the words that were
1:23:58 coming out of his mouth were just oh
1:24:00 he's a little toxic. I feel triggered
1:24:03 but but I laughed.
1:24:06 U Black Bar you [laughter] you've
1:24:07 offended Mr. IT and Central Standard.
1:24:11 Sorry Mr. IT. [laughter]
1:24:19 [clears throat] Um, so that's that.
1:24:21 Tomorrow night, Friday night, date
1:24:22 night, you get your nacho fixings,
1:24:26 you get your hot pockets for the dirty
1:24:28 microwave, you show up tomorrow night,
1:24:30 you bring your friends. [sighs]
1:24:33 I think Anthropic kicked
1:24:36 Open Claw off today. Yeah. Well,
1:24:40 Enthropic
1:24:42 kind of [ __ ] up
1:24:45 because what Anthropic did was they
1:24:47 ceased and desisted Peter Steinberger
1:24:51 and Open AI hired him.
1:24:55 [laughter]
1:24:56 Anthropic's had a bad couple of days.
1:24:58 They're they're they're not doing too
1:25:00 good in the PR department right now. Uh
1:25:02 if if if
1:25:05 anthropic cut off access to to open claw
1:25:14 whatever like this is this is why we
1:25:17 need open source.
1:25:19 This is why we need open source. Worse
1:25:27 they kicked it off.
1:25:31 lovable part two tomorrow. Yeah. Oh, and
1:25:32 and tomorrow tomorrow tomorrow night
1:25:34 we'll take this app that's being built
1:25:36 right now. Do I have to do anything
1:25:38 here? Oh, I do. What do I have to do?
1:25:48 Let me get everything moving. Enabling
1:25:50 cloud plan.
1:25:58 Is it building?
1:26:04 Okay.
1:26:18 You're going to want to look for that
1:26:19 stop icon in the lower at the prompt
1:26:22 window. If it's not an arrow, it's still
1:26:25 chewing.
1:26:26 >> It's an arrow. That's That's what I'm
1:26:28 saying. It's an arrow.
1:26:31 >> Should I say continue?
1:26:33 >> Yes.
1:26:35 >> Maybe it ran out of tokens. [laughter]
1:26:39 >> Wait, scroll up just a little bit. There
1:26:41 was a little Okay. Yeah, it's All right.
1:26:47 >> The user wants me to continue.
1:26:51 All right.
1:26:55 It was just checking in on you.
1:26:58 >> Yeah.
1:26:58 >> Ah, there it is. Allow.
1:27:00 >> Yeah.
1:27:03 [whistles]
1:27:06 So, what I just allowed there is I
1:27:07 allowed
1:27:09 lovable cloud so it can do like
1:27:11 databases and stuff like that. All
1:27:12 right. So, that's off off building.
1:27:15 Okay. Beautiful. I'm out of here. Um,
1:27:18 join the AI salon. Come to office hours
1:27:21 tomorrow. Come back here tomorrow night.
1:27:23 We'll finish this thing up. And uh Kyle,
1:27:27 did you build openclaw?
1:27:29 Was the thing doing Wait, was thing
1:27:33 doing the open claw open rights and a
1:27:36 Mac Mini? Yeah. So So I started it, but
1:27:39 I only I only put about 45 minutes into
1:27:41 it. I had Claotter chat GPT build me a
1:27:45 tutorial for how to install it on my
1:27:47 gaming PC and I started doing that and
1:27:50 it gave me some shitty instructions. And
1:27:52 so I'm now now trying to figure out how
1:27:54 to navigate through the thing. But
1:27:55 what's cool is if you can get some even
1:27:57 shitty instructions from chat GPT on how
1:28:00 to do it. It gives you the basic things
1:28:01 and then as you get errors, you can just
1:28:03 dump those into chat GPT and say, "Help
1:28:05 me help me solve my way through this."
1:28:08 Um it's it is not pleasant. If if you're
1:28:11 if you're if you enjoy technical stuff
1:28:14 and you don't mind just sort of grinding
1:28:15 and and moving it forward, um I think
1:28:18 it's doable. You can also just wait,
1:28:21 right? because they hired OpenAI hired
1:28:24 Peter Steinberger. So, they're going to
1:28:27 it's going to be built into Chat GPT
1:28:29 probably within 6 months.
1:28:32 All right. All right, kids. Um I'm out
1:28:35 of here. Uh I will see you tomorrow. All
1:28:38 right. Good evening and thanks for
1:28:41 hanging out.