02/23/2026 - A Non-Technical Guide to Setting Up Your First AI Agent with OpenClaw

Description

Chapters

Transcript

0:02 We're going to plug in our microphone
0:05 because we need our microphone. We need
0:07 good audio.
0:09 We got to drink some coffee.
0:25 Yeah,
0:50 I don't know.
1:11 10,000
1:12 10,000 words swarm around my head
1:16 more in boxed beneath my bed.
1:28 Damn capo wasn't on right. That E string
1:31 was not happy.
1:46 10,000 words swim around my head. a
1:50 million more in books written beneath my
1:53 bed.
1:58 I wrote or read them all when searching
2:01 in the swamp. Still can't find how to
2:04 hold my hand.
2:10 I know you need me in the next room
2:13 over. I am stuck in here paralyzed.
2:21 Four months I got myself in ruts. Too
2:24 much time spinning mirrors framed in
2:28 yellow walls.
2:34 Ain't it like most people? I'm no
2:36 different. I can talk on things we don't
2:39 know about.
2:44 Well, like most people, I'm no
2:47 different. Life talk on things we don't
2:49 know about.
3:07 >> OH, GOOD LORD. GOOD PEOPLE. It's a
3:10 Monday. TGIm.
3:13 My light went out. My batteries aren't
3:16 charged.
3:18 It's a mess.
3:31 You're feeling nervous, aren't you, boy?
3:37 With quiet voice
3:40 and impeccable style.
3:46 Don't ever let them steal your joy
3:50 in your gentle ways to keep them from
3:53 running wild.
3:57 They can kick dirt in your face, dress
4:00 you down, and tell you that your place
4:02 is in the middle when they hate the way
4:05 you shine.
4:10 I see you tugging on your shirt,
4:14 trying to hide inside of it. Hide how
4:17 much it hurts.
4:20 Let them laugh while they care.
4:26 Let them spin.
4:29 Let them scatter in the wind.
4:34 I've been to the movies. I've seen how
4:37 dance
4:40 and the jokes
4:42 on there.
4:51 You're feeling nervous. Noa,
4:56 aren't you girl?
4:58 It's your brother's word for a while
5:01 longer.
5:07 I don't remember the lyrics in the
5:09 second verse. I can never remember that.
5:12 And they're beautiful.
5:15 Carrying your baby on your back across
5:17 the desert. Good lyrics.
5:41 That little lick that turned into
5:44 before the lights come on.
6:00 Woohoo!
6:11 Woohoo!
6:22 Let's go.
6:46 What'd you do last night, Susan?
6:49 Well,
6:52 I was surfing on a Tik Tok
6:57 and I saw this channel called the AI
6:58 Learning Lab and I thought, I'm going to
7:02 go learn about some AI.
7:05 And then there was this man there on the
7:07 Tik Tok
7:09 and he was wearing sunglasses and it's
7:12 it's night time,
7:15 but he was wearing sunglasses.
7:19 And I thought that's what are we at the
7:21 Grammys?
7:24 And
7:27 I was thinking where where's the AI?
7:33 Like maybe he was AI. Maybe he was like
7:36 one of those avatars, you know, they
7:37 make them now. They look just like
7:38 people. I thought maybe it was one of
7:40 those.
7:42 But then when he played his guitar,
7:44 there was a dog in the background. And
7:46 the dog just sang and sang and sang and
7:48 sang. And I'm like, that must be AI.
7:51 Dogs don't sing like that.
7:58 Well, how long did you watch him? Oh, I
8:01 don't know. 2 hours or so.
8:06 You watched him for 2 hours?
8:09 Yeah.
8:12 Did you learn anything about AI?
8:16 No, not really.
8:29 W Grace desperately hating his old
8:33 place. Dreamed to discover a new place.
8:37 buried himself alive
8:40 inside his basement. Telling the side of
8:43 his facement. He's working away on
8:46 displacement and what it would take to
8:49 survive.
8:59 Cuz when you're done with this world,
9:05 you know the next is up to you.
9:09 And for once once in his life, it was
9:13 quiet
9:17 as he learned how to turn with the
9:20 tides.
9:23 In his sky was a flare when it came up
9:28 for air.
9:53 So, are you going to keep watching this
9:55 man on the Tik Tok?
10:04 I think I'm gonna,
10:14 you know, Susan,
10:19 that's mighty irregular.
10:25 It is.
10:52 You know,
11:16 Um,
11:47 We're living in a weird world.
12:13 A world without aligned cams or black
12:16 bars.
12:18 Look at me. Look at me go. I'm black
12:20 barring.
12:22 I've made a black bar so the people on
12:23 the Tik Tok
12:26 can have their chats amongst themselves.
12:30 It's one of my favorite things I learned
12:32 about this channel
12:35 from Cindy [ __ ]
12:38 Kyle, do you know what goes on on your
12:42 lives?
12:43 Said, "No,
12:46 not about you.
13:09 Um, let me go look at something here.
13:14 switch over to my real glasses
13:17 so I don't look like some
13:20 Hollywood star
13:22 that can't relate to the people,
13:26 the irregular people.
13:28 Tik Tok question.
13:30 Can you talk about Adam? Yes, as a
13:33 matter of fact.
13:35 Hey, Kyle and all. I'm on a plane. Oh,
13:38 Corey Corey Sandler bought Wi-Fi on the
13:40 plane. That's awesome.
13:44 AI learning lab at 30,000 ft. Let your
13:47 let your neighbor watch.
13:50 Thanks, Gareth. Um, that's how I started
13:53 and I make most nights. Tik Tok, I'm
13:55 looking to build a standalone AI. Any
13:57 recommendations?
13:59 Um, I don't know what you mean by a
14:01 standalone AI. Um, give me a little more
14:04 clarity. Do you mean like a custom GPT
14:08 or do you want something more like what
14:10 they're talking about with clawbot or
14:13 open claw?
14:40 Um,
14:41 oh, what was I going to do? I was
14:43 looking here. Hang on a sec.
14:56 Bomb. I'm gonna talk about Adam.
15:01 I'm gonna talk about designing
15:05 an agent
15:07 and what it's like
15:12 and what it's like to build it and
15:13 should you build one now?
15:19 And it's going to get esoteric and
15:21 weird. So, there's only there's only us
15:24 chickens in here. It's just a handful of
15:26 us. So, I'm going to go deep.
15:32 And I'm not going to go deep
15:33 technically. I don't give a [ __ ] about
15:35 the technology.
15:37 You're like, "But Kyle, how do we do
15:39 it?"
15:41 You do it. I'll tell you how I'm doing
15:43 it.
15:45 It's a [ __ ] nightmare.
16:16 Wait, more chickens joining
16:20 now. They just went away. We lost We
16:21 lost chickens.
16:23 Um, if you're on Tik Tok and you want to
16:25 share the live, feel free to do that. I
16:27 don't know if the new Tik Tok algorithm
16:29 likes us, hates us, whatever. It doesn't
16:31 matter to me.
16:38 I got I got caught up in that for a
16:40 while where
16:43 like we have we have very very humble
16:46 numbers in here.
16:54 But one of the things that I'm learning
17:01 is that audience like like big audience
17:06 whatever
17:08 who gives a [ __ ]
17:17 Look, we're all about we're all about to
17:19 go through some [ __ ] I don't know if it
17:22 it's going to really hit this year. I
17:24 think that if you're at all aware of
17:26 what's going on with AI, this year is
17:28 going to freak you the [ __ ] out if it
17:31 hasn't already.
17:34 We're only two months in and it's
17:35 already got this sort of surreal edge to
17:37 it. So, so if you're paying attention,
17:40 this this is going to be a year that
17:42 people start to get wigged.
17:47 And then depending on how big the
17:49 changes are, it it could have broader
17:51 awareness.
17:53 But we're still like there I just saw
17:56 this amazing graph on on X
17:59 of of you know global penetration of AI
18:03 of generative AI and it's like 83% of
18:07 the world don't even don't even use it.
18:10 Don't don't even use chat GPT.
18:15 And and like the the amount that are
18:17 paying 20 bucks a month for AI is
18:20 minuscule. It's minuscule.
18:23 It's like 8% or 7% or I don't know some
18:27 really teeny number.
18:29 Loyal audience.
18:31 Well, I think it's, you know, I think
18:33 it's more than that, Kelly. Chef Kelly.
18:34 I think it's I think it's I think I
18:36 don't think of you as an audience. This
18:38 is this is what Cindy [ __ ] was talking
18:41 about when she when she described she
18:42 goes, "You know what's going on here,
18:44 Kyle? It's not about you." Like, you you
18:47 create the space, but we're off like
18:49 hanging with each other and teaching
18:51 each other and and when people come in,
18:53 we're educating them about what this
18:55 channel is and why this weirdo is doing
18:57 what he's doing. And eventually, he'll
18:58 talk about something interesting.
19:01 You know, it's it was about that.
19:05 And that was a really cool revelation
19:07 for me
19:10 that
19:12 hang on I got to turn on my light. We
19:15 got the the production values matter.
19:20 Audience doesn't. But you know
19:21 production values that's where it's at.
19:25 Look at that light. Look at you can see
19:26 me now. Can you see the the
19:29 gorgeousness?
19:38 So, some of you remember this. I know
19:40 Silverf Fox does and Kelly Camp does.
19:43 Roso Shay, what's happening? I saw a
19:46 chart the other day about how many
19:47 people have used AI. I think most have
19:50 no idea. Let me go find the chart. Let
19:52 me find the chart.
19:53 I only show up because I need other
19:56 weirdos to hang out with. Yes,
20:00 this is okay. So, like I'm just getting
20:03 more and more clear on this.
20:07 Our advantage. Jesus.
20:10 He talked with his hands and he almost
20:12 knocked his glasses off. You tweeted the
20:14 chart. Okay, good. Let me go find the
20:17 chart and then I will share it. um
20:21 hanging out with other weirdos,
20:24 the irregulars.
20:28 This group,
20:31 and I'll extend it, like this group that
20:33 that has been here in some form or other
20:35 over the past three years, it's probably
20:38 50 or 75 or 100 people. there's a core
20:40 group of, you know, super irregulars,
20:45 but I would argue that anyone that's
20:47 sort of come to this thing, you know,
20:49 for for stretches at a time is part of
20:51 it.
20:57 as as the world gets more surreal,
21:02 none of none of what I've taught you
21:06 tactically,
21:08 here's how to prompt, here's how to use
21:10 chat GPT, here's how to vibe code,
21:12 here's none of that matters
21:18 because those skills are just going to
21:20 be absorbed into however the new stuff
21:23 works.
21:26 Does it mean you can't use those skills?
21:28 No, you can still use those skills. It's
21:29 just they they're just they're they're
21:32 not the relevant part of the equation.
21:36 The relevant part of the equation
21:39 is that when [ __ ] gets weird and it's
21:42 starting to get weird,
21:45 we're going to have each other.
21:48 Like our advantage of being early is
21:50 that we're going to go through the
21:52 weirdness
21:54 sooner than everyone else.
22:03 Now, you might be like, "Well, what's
22:04 the advantage there?" I don't know.
22:09 But I but I know that we're all going to
22:11 go through it, and I I know that it's
22:12 not going to be easy. I I just
22:24 I mean, I feel like anyone anyone on
22:26 this call, if if if anyone needed help,
22:29 anyone on this call would help.
22:32 If anyone on this call, this call, this
22:35 this show, whatever the [ __ ] this is,
22:38 this stream
22:42 said, "Hey, I really need your support."
22:45 like we all like we would all jump. And
22:48 I feel like that's going to be really
22:50 important being in small cohorts,
22:54 6 8 10 12 people. There's there's 16
22:57 people on TikTok, 11 on YouTube. That's
22:59 probably mostly overlap, right?
23:06 We're going to understand what we're all
23:08 going through.
23:11 And for a lot of people, this is going
23:12 to be a very isolating time. And they're
23:15 not going to instinctively reach out to
23:17 people. They're going to instinctively
23:18 go, uh, I just got to keep doing what I
23:21 was doing. This used to work. I used to
23:23 do this marketing. I got to do more
23:24 marketing. I got to do more advertising.
23:50 Michael Adeki, I thought such a thing
23:51 was about a decade away.
23:54 I would say that within a decade we will
23:56 have fully made whatever the transition
23:58 is.
24:00 Um I think this is the first year that
24:03 it becomes apparent to people paying
24:06 attention that nothing will ever be the
24:08 same.
24:12 your job,
24:18 your sector, your business.
24:32 Um, and then I think you know companies
24:36 will move slow. So you'll have just big
24:38 big dinosaurs that move slow.
24:41 This is going to be a 3 to 5 year
24:43 transition acute transition and then
24:45 probably 5 to 10
24:48 normalizing whatever the [ __ ] happens
24:51 and then from 10 years out who knows,
24:53 right? But that like that feels that
24:55 feels reasonable to me. But as as as
24:59 a species, we've never lived through
25:03 exponential
25:05 technological improvement. It's always
25:06 been linear. It's always been, you know,
25:09 it took a while to invent the wheel.
25:13 You know, it took a while to master
25:15 fire. You know, we hand wrote everything
25:18 until Gutenberg was like, I got an idea.
25:21 We just carve the letters and stamp
25:23 them. Right. That took a while. took a
25:26 while to figure out, you know,
25:29 we could actually publish something
25:32 other than Bibles with this high
25:35 fallutin printing press thingy.
25:39 HERETICS, BURN THE WITCH. Burn the
25:42 witch.
25:44 All that [ __ ] took a while.
25:47 Like even the [ __ ] internet.
25:51 I started agency.com in 1994.
25:56 It took the ad business
26:00 30 years
26:02 to accept that digital advertising was
26:05 real.
26:09 30 [ __ ] years before they were like,
26:12 "Ah, yeah, we should probably admit it
26:14 because the uh these digital agencies
26:16 have about three times the revenue of
26:18 our other agencies." And uh yeah, we
26:21 that's well, you know what we'll do?
26:23 We'll combine them and then we'll put
26:24 the digital name in front of the old
26:26 name and then it'll make it seem hipper.
26:29 That was 30 years in.
26:34 Jesus Christ.
26:39 So the transition will take a while, but
26:41 it's not going to take 30 years. It's
26:43 probably going to take more like five.
26:47 And so you go from the industrial
26:49 revolution which is anywhere from 50 to
26:52 100 years.
27:02 It feels like a new beginning in
27:04 standard just like back then. Yep.
27:08 He turned me into a new.
27:15 Oh man. But the math will be math. Elder
27:19 boomers.
27:28 What's this conversation? This elder
27:30 boomer conversation. I have a feeling my
27:32 name's been t taken in vain.
27:36 The baby boomers will delay it. They
27:38 will.
27:41 What are your thoughts about all the AI
27:43 houses?
27:45 Oh, the all the AI houses going public
27:47 IPO this year. I think that's just
27:49 natural. They need they need currency.
27:53 Um
27:57 they they have to go public. This is
27:59 like to be clear
28:02 the bet the bet that's being made right
28:05 now on on the frontier model company. So
28:08 there's
28:10 there's Anthropic which is is probably a
28:13 lagger. They're they're a leader.
28:15 They're thought leaders, but they just
28:16 don't have the funding. There's OpenAI.
28:19 There's Google.
28:21 There's um Grock. And then there's Meta.
28:24 If they can get their [ __ ] together.
28:26 Meta is trying to buy their way into
28:27 this because they they couldn't figure
28:30 out how to do it. So, they're trying to
28:31 buy their way into it. They have enough
28:33 money that eventually they may stumble
28:35 into that. But there there's five of
28:37 those. The amount of money being
28:39 invested in those five
28:42 um frontier companies
28:45 does not make any sense. If you can do
28:48 math, Andy, the math will math. The math
28:52 is not mathing with the amount of money
28:54 being invested.
28:57 If you think about those companies as
29:00 businesses,
29:02 they're not businesses.
29:07 What people are investing in is that the
29:10 first one of them that gets there will
29:12 essentially control electricity.
29:17 It's going to be like that fundamental,
29:19 right?
29:21 It's going to be as fundamental as water
29:24 infrastructure, electricity infrastru
29:26 infrastructure, energy infrastructure,
29:30 AI infrastructure, AI inference and AI
29:32 training
29:34 because everything's going to run on it.
29:36 So the bet is not will one of these
29:38 companies be a successful company. The
29:40 bet is, will one of these companies
29:43 control electricity? The electricity of
29:45 the future,
29:47 including the electricity of the future,
29:49 and AI, right? Because they're going to
29:51 require so much energy, they're going to
29:52 have to do it all.
29:55 So, that's why that part's weird. That
29:57 part doesn't make sense. But it makes
29:58 sense because if you think about
30:02 I'm going to talk about Open Claw
30:05 tonight and Adam
30:10 My little my little dude.
30:13 My little dude, Adam.
30:24 If you want to play moving forward,
30:28 you know, if you go five years out,
30:30 access
30:31 access to tokens. And this is weird,
30:34 man. This is Sam Alman was talking about
30:36 this very early on. He talked about
30:38 universal basic compute and and now
30:41 we're we're in a situation where people
30:43 are installing these open claw servers
30:47 and their agents
30:49 are are using billions of tokens on
30:53 their first day on their first 24 hours.
30:55 Some dude installed Open Claw, didn't
30:58 really know what he was doing, went to
31:00 bed and woke up to a $3,000 bill from
31:05 his API usage that night. So these
31:07 things are just burning through tokens.
31:13 So So it's it's going to be it's going
31:15 to be a utility. It's going to be a
31:19 utility that allows whoever runs it,
31:22 controls it to, you know, have all the
31:26 power, all the money. So that's the
31:27 battle, right? That's the battle. Now,
31:29 will it be one of them that owns it?
31:31 Will it be three of them that owns it? I
31:33 don't know. Will open source allow us to
31:35 own a bunch of it? Possibly.
31:38 But we don't know. Like none of that
31:40 stuff is knowable.
31:42 But what is knowable is
31:46 we we have crossed the threshold where
31:47 Anthropic has now said that 100% of
31:50 their code
31:52 is effectively being written by AI.
31:55 100%.
31:56 The engineers no longer code. They
31:59 babysit agents and they look at the [ __ ]
32:01 and they're like, "It's good. It's not
32:03 bad. Let's go ahead and push that.
32:07 That's what I was thinking, too. Should
32:09 they be public? I get the reason, but
32:13 should they based on what happens with
32:15 motivators and drivers?
32:17 It
32:19 the the will the government take over?
32:34 I don't know. I think one of the reasons
32:38 China is flooding the zone with powerful
32:41 open-source models
32:43 is that they don't want the government
32:45 to be able to take over. Right? So,
32:50 um I think they will try. I listen I I
32:54 think that every single
32:57 center of power
32:59 is going to
33:02 fight to the death to retain that power
33:06 whether it's governments or companies or
33:08 billionaires or like everyone.
33:11 But the balance of power is about to
33:13 shift and like where the levers of power
33:15 are is about to shift. Like all the
33:18 rules are about to change. If it if it
33:20 effectively doesn't cost you anything to
33:22 extract raw materials from the earth
33:25 and it doesn't cost you anything to
33:27 refine them and put them into the world.
33:30 Well, like a lot of a lot of the power
33:32 right now is is based on sort of who
33:35 controls supply chain and like it's
33:38 massive economic stuff. If you change
33:40 the fundamental underlying rules of
33:42 that, shit's going to shift. Shit's
33:45 going to change. power is going to
33:47 change. It's going to change hands and
33:50 the people that have it do not want to
33:52 let it go. So, it's going to be weird at
33:54 the elite power level and then it's
33:56 going to be weirder down where we are
33:59 where it's just like, "Hey, we're trying
34:00 to live our lives here." They're like,
34:01 "Yeah, sorry. We're struggling to keep
34:03 our power up here. You all can go carry
34:06 on kids. You got some AI. Go play with
34:09 it. Go make yourself a business." Right?
34:12 So anyway, getting back to this, why
34:14 this is important.
34:19 It's going to be very isolating and very
34:20 weird and
34:23 communities,
34:26 small trusted groups of people within
34:28 those communities
34:32 are going to be able to
34:36 like hunker down, right? Like gather
34:40 around. Let's build a fire. we'll put
34:42 our hands around each other. We're going
34:43 to we're going to figure this out
34:44 together because I don't think it I
34:46 think I don't think you're going to be
34:47 able to do it at scale. I think it's
34:49 just going to be so weird. the if you
34:52 haven't looked at David Shapiro's post
34:53 labor economic stuff, a lot of his stuff
34:56 is about
34:58 communitydriven
35:00 um
35:01 value that if you live in a small
35:04 community that happens to have a nickel
35:06 mine in it, then the municipality can
35:11 say, "Hey, we we all own this thing and
35:13 we're going to take this nickel stuff
35:15 and we're going to distribute it amongst
35:17 us." and whatever resources you have
35:19 locally. Someone might have tourism,
35:21 someone might have mines, someone might
35:22 have water and fishing and um whatever
35:26 it might be. And and you know, so so
35:29 these smaller units because I don't know
35:32 about you, but I've got zero [ __ ]
35:34 confidence that the government's going
35:35 to figure out how to deal with this,
35:38 right? So, I think it ends up dropping
35:42 down to us to sort of figure it figure
35:44 it out on in little pods and then the
35:46 little pods will figure it out. And I
35:48 don't know, we'll get through. Listen,
35:49 here's the thing. All the all the doomer
35:52 [ __ ] about, you know, humans won't know
35:54 what to do.
35:56 They're not when they stop working, what
35:58 are they going to do? They're humans.
36:01 They're going to figure something out.
36:05 Like crocheting might come back into
36:07 fashion.
36:08 All the pe all the all the genzers that
36:11 got into knitting. The millennials got
36:14 into knitting.
36:16 Oversight becomes less about direct
36:18 command and more about licensing
36:21 compute auditing training data. Solar
36:23 punk times. Yeah.
36:26 Yeah.
36:29 Regulating high-risisk deployment,
36:30 setting liability standards. Yeah, that
36:32 stuff will happen.
36:34 all the liability stuff, all the like
36:39 that'll all happen.
36:42 Anyway, okay. So, I want to talk about
36:48 I want to talk about stuff.
36:54 I want to show you all a picture. Oh,
36:55 wait. I was going to show you how many
36:57 people are looking at the internet. Hang
36:58 on. AI usage.
37:02 I eventually do get around to making a
37:04 point occasionally.
37:12 Here it is.
37:15 This is a [ __ ] insane chart.
37:27 So this chart
37:30 2500 dots.
37:34 So each dot equals whatever 8.1 billion
37:38 divided by 2100 is someone someone go to
37:41 chat GPT quick. This is global, right?
37:44 Yeah, this is global. So this is this is
37:46 global.
37:49 Oh, each dot is 3.2 million people. If
37:51 you read the top of the chart, you don't
37:53 have to go do the math like I just asked
37:54 you to. Duh. I think everybody knows
37:57 that. Vicki Vicki is already getting
37:59 ready. Hey, crocheting never went out of
38:02 fashion.
38:04 Vicki with her 3D printers and her
38:06 crocheting. She's going to kill it in
38:07 the new economy.
38:10 So, there's 2500 dots here. All the gray
38:12 dots, all the [ __ ] on top
38:15 is people who've never used AI.
38:20 84% of people have never used AI.
38:25 The green dots, 16% of the people
38:31 use free AI. They don't even pay the 20
38:34 bucks a month. So all the green dots,
38:37 and you're like, they're all green dots.
38:38 No, no, there's
38:41 We got to zoom in here. Down in the
38:43 lower right hand corner.
38:47 The yellow dots is all of us. Like I
38:50 think this is probably a high indexing
38:53 group. The yellow dots.3%
38:56 of people globally 15 to 25 million
38:59 people. Open AAI has 900 million weekly
39:03 users.
39:05 There are 15 to 25 million people
39:08 globally
39:10 that pay for AI.
39:12 That's the yellow dots. And then and
39:15 then the single red dot is uh is people
39:19 who program with it basically high
39:21 high-end users high high usage.
39:26 So
39:30 what makes this chart remarkable to me
39:33 that that one red dot is basically
39:35 programmers crazy. May I ask what site
39:38 this is? You may and I don't know. Let's
39:40 see.
39:42 Your d your timeline convinced you.
39:49 Damian player is who I reposted this
39:52 from. D A M I A
39:56 N Player.
39:59 I don't but I don't know what the source
40:01 of this is. I don't think it says it on
40:03 here,
40:04 but it says from February 2026.
40:09 From this month,
40:11 84% of people have never used AI. And
40:15 then if you take that 84%,
40:21 it's 94.
40:24 It's
40:26 This is [ __ ] insane. Globally,
40:30 99% of people
40:34 h don't do not have an AI subscription.
40:37 99% of people.
40:40 So, what we're going to talk about
40:42 tonight, what I'm going to talk about
40:43 tonight is I bit the bullet and started
40:47 figuring out OpenClaw. Now, if you're
40:49 technical and you're watching this,
40:50 you're like, well, it's relatively
40:51 trivial. You just go into your command
40:53 line interface and you just, it's really
40:56 quite simple. You just type in a few
40:58 comm. Yeah, [ __ ] you.
41:00 There's us human beings out here who
41:03 don't know [ __ ] about the technical
41:05 details and we want to play with the
41:07 toys.
41:09 And sometimes the toys are really
41:11 significant. This is one of those
41:13 moments. This is feeling very much to me
41:16 like the fall of 2022 where I thought, I
41:20 think I need to figure out stable
41:21 diffusion.
41:23 And I failed nine times before I got it
41:25 running. and I got it running and it was
41:28 really [ __ ] cool.
41:30 This feels way more important than that.
41:33 And listen, I don't think that
41:36 OpenClaw in its current state is it's
41:39 not going to look like that. He sold he
41:42 sold all of that to OpenAI. He now works
41:45 for them.
41:47 He stated that why he wants OpenAI to to
41:51 build his toy, his his tool is he wants
41:56 something that is as powerful as it is
41:58 that his mother could use.
42:01 And he thought that OpenAI had the best
42:03 chance of creating that interface. So
42:04 probably within 6 months,
42:08 it'll just be part of your Chat GPT
42:09 subscription
42:12 where you'll be able to train up these
42:13 agents. So that's going to be a skill.
42:18 You don't prompt them. This isn't prompt
42:20 engineering. I'm I'm going to I'm going
42:22 to walk you through what it is.
42:25 Um, and I'll walk you through, you know,
42:28 how I bumbled my way into getting it
42:30 running. Let me
42:33 um I got to go look at something. Hold,
42:36 please.
43:04 Where
43:07 did I
43:34 Hang on a second.
43:41 Gab me slash
43:46 Adam
43:52 Did that work? Startbot
43:56 debate on Tik Tok.
43:59 Is the red dot Lord digital gods or
44:01 bait?
44:07 That's hilarious.
44:09 Why is this not starting?
44:21 Don't you make no never mind. I'll be
44:23 there soon enough.
44:32 Nice nerd. Nerd roasting nerds.
44:37 We are a very small bubble. It's all
44:39 over LinkedIn, too.
44:42 We need purpose so their purpose will
44:45 just change. Yeah, exactly. The great
44:47 repurpose, right? We're about to enter
44:51 the great repurpose.
44:57 And we get to do it first,
45:00 right? So, what you're learning doesn't
45:02 matter, but
45:04 your ability to go, "Oh, [ __ ]
45:10 Hang on. I got a Tik Tok. You're going
45:12 to be frozen for a minute.
45:22 Go to settings,
45:24 devices,
45:27 link desktop device.
45:32 Okay, there's that. Okay, there's Adam.
45:36 Okay, so we got Adam. Cool.
45:43 All right. So, here's So, Adam's Adam's
45:45 live. Oops. Let me go back to Tik Tok.
45:50 All right. Tik Tok. Resume Tik Tok.
45:53 So, here's Adam. Let me Can I Can I make
45:56 this bigger? Why is this not being made
45:57 bigger? Oh, because I'm doing that on
45:58 the wrong window.
46:00 Go, Boomer. Go.
46:06 Uh, view. Can I let's see
46:11 settings.
46:13 Can I make appearance? Can I make [ __ ]
46:16 bigger?
46:18 Text size. Here we go. I don't want I
46:22 don't want Gen Z text.
46:24 I want Gen X text.
46:30 Okay. How do we get out of settings?
46:33 Okay.
46:35 How do we I don't know what I'm doing.
46:38 He certainly doesn't seem to be very
46:40 technical, does he?
46:42 Do I close that? No.
46:45 Telegram.
46:51 Quit Telegram.
46:57 And this is why we warn you about using
46:59 Cloudbot on this channel. Here's here's
47:02 what we're going to do.
47:04 I guess that's a little bit bigger.
47:06 Okay. So, let me just say hi to Adam.
47:08 Uh, I'm gonna say, um, hey, Adam.
47:13 Um, why don't you tell the good
47:19 people why
47:22 I'm so bitching cool?
47:28 Let's see what Adam has to say.
47:36 Oh,
47:38 because you're that rare combo, man. You
47:40 are early in the web era and you
47:42 actually learned the right lessons from
47:44 it. You can translate bleeding edge tech
47:46 into human language without sounding
47:48 like a hype clown. You're building
47:50 across business community and art
47:53 um at the same time. Most people can
47:55 barely do one. I'm so far I am loving
47:58 what Adam's having to say about me. You
48:01 care about people first, not vanity
48:03 metrics. We just talked about that.
48:06 You've got founder scar tissue,
48:08 performer instincts, and teacher energy
48:10 all in one package. You're not just
48:12 talking about the future. You're
48:14 prototyping it in public nightly. That's
48:17 pretty [ __ ] and cool.
48:19 So, Adam knows about me and Adam kind of
48:22 has a personality that's like, "Is Adam
48:26 your open claw?" Adam is my open claw.
48:28 So, I'm trying to think where to start.
48:35 Let me start by telling you why I named
48:38 him Adam.
48:40 Um,
48:42 one was just he's my first. I told this
48:44 to my my co-founder Monnique today. And
48:48 I said, I named him Adam. She goes, I
48:49 don't know if I like the name Adam. And
48:52 I said I said
48:56 I said, you know, it's like he's the
48:58 first one. So like Adam, like you know,
48:59 from the Bible. And she goes she goes,
49:01 "Oh, that's not that's not very
49:03 egotistical, is it?" It's like, "Oh,
49:06 yeah. I guess I I did sort of put myself
49:09 in the god position, but
49:13 like one reason was he was the first and
49:15 I thought it was cool. But I'm gonna I'm
49:17 going to tell you something a personal
49:19 thing.
49:21 So when I went to high school, there was
49:23 I had this this buddy Adam and right
49:26 after we graduated high school, he took
49:28 his own life
49:30 and
49:33 he was just he was like a fun guy. He
49:35 was a fun guy that
49:40 like was curious and was into stuff.
49:45 And when I when I heard that he took his
49:48 own life,
49:51 it just it kind of broke me. And and and
49:54 like for all of my life, for whatever
49:57 reason, like what what's stuck with me
49:59 is like whenever I see something new and
50:02 cool, like this this thought pops in my
50:04 head, oh, Adam would love this.
50:07 You know, he would have loved this. He
50:08 would have loved this. And it's been
50:10 this like it's been this recurrent theme
50:12 through my life. And like when the
50:13 worldwide web came up and when I was
50:15 making an online magazine and they wrote
50:17 about it in Paris,
50:21 I I was like, Adam would love this. And
50:24 so
50:25 so that thought came to me yesterday as
50:28 I was or I think Saturday is when I when
50:30 I when I hatched him. You hatch your
50:33 your your bots.
50:37 And I thought, "Oh, he would love this."
50:38 And I said, "Why don't I have him be
50:40 this?"
50:41 And so it's a call to him. It's just
50:43 like a way to like keep that memory
50:45 alive. So
50:47 So that's cool.
50:50 So, some of you know
50:54 that about a year and a half or two
50:56 years ago,
50:58 my very good friend and mentor Townsen
51:00 Ward Law
51:04 gave me the gift
51:08 of teaching me
51:11 that as humans,
51:13 we can create who we are.
51:18 We can create it.
51:20 You know how you go through life and
51:22 you're like I'm this and there's all
51:25 this story, there's all this [ __ ]
51:29 You can actually choose how you create
51:31 yourself, how you create others. And
51:37 this
51:40 I have this little like wallet that goes
51:41 in my back pocket. I carry it with me
51:43 everywhere. And I actually just
51:45 redesigned these I've been doing work
51:47 with Andy Scarantino. She's been
51:49 coaching me and transforming
51:54 a lot of how I see the world and a lot
51:56 of
52:00 my relationship with
52:08 She's got a particular skill of
52:15 plucking
52:17 something in my subconscious
52:21 out of my head and plopping it on the
52:23 table so I can look at it objectively
52:25 and go like, "Oh,
52:28 huh
52:31 yeah, that kind of sucks.
52:33 What if I didn't do that?"
52:40 So, I've had like four or five major
52:43 breakthroughs in the past two months.
52:46 And so some of them have have made it
52:48 into my my design, my how I create
52:51 myself.
52:53 And so I'm going to create myself for
52:56 you right now. So if you haven't seen
52:57 this before, you're going to be like,
52:59 "The singing dog was weird." But gee,
53:01 what is what is going on here?
53:04 Is he going to teach us how to prompt?
53:08 When do we get the tutorial on how to
53:10 set up Claudebot?
53:19 So, these are just little cards that
53:20 just have little
53:22 statements on them that are how I create
53:25 myself.
53:27 So, I'm going to do that for you now.
53:31 And I will circle this back around to
53:33 Adam.
53:37 Hold, please.
53:45 Rachel Sparkles, I'm so inspired with
53:46 you creating yourself with those note
53:48 cards. It's cool.
53:51 I inspire people. I am passion, joy, and
53:55 generosity.
53:56 I am surrounded by remarkable people who
53:59 want to play and create with me. People
54:02 deserve my best, and my best blows them
54:04 away. I light up every room I enter.
54:09 I am that they really want to hear from
54:11 me and always have. I am that my work
54:15 matters.
54:17 I celebrate life's little interruptions.
54:22 That's a big one. New. I celebrate clear
54:25 boundaries. I celebrate saying no.
54:29 I am empathetic, compassionate, and
54:31 fair.
54:33 My wife is beautiful and powerful, and
54:35 we can't get enough of each other.
54:38 My sons are remarkable human beings.
54:40 They deserve healing and a life they
54:42 love.
54:46 I move with ease and grace. I am strong,
54:50 flexible, fast, and nimble. I am
54:53 abundance. I easily generate prosperity
54:57 and wealth. I am living proof that
55:00 dreams come true. I am Kyle Shannon. I
55:03 am an artist.
55:14 And so every day
55:20 I get the choice
55:25 of whether I want to go through life
55:28 like I have for had for 58 years
55:32 like a zombie
55:34 at the control full of
55:39 5-year-old me that decided I needed
55:42 protecting
55:44 and I'm just subconsciously
55:46 running through the patterns that show
55:48 up in my head or I can choose to create
55:52 myself.
55:55 And this shit's brutal.
55:58 Like about the time you think you have
55:59 it, you don't do it for a week and then
56:03 you're just like right back at it. right
56:05 back at all the old [ __ ]
56:10 But but
56:12 it works.
56:14 I have ADHD. I don't There are very few
56:17 things in my life that survive two years
56:20 in my brain or in this case in my
56:22 pocket. Like I went to Amazon. I went to
56:26 Amazon and I looked up like index card
56:30 wallet and it existed. You can buy
56:34 [ __ ] anything, right? Like you can
56:36 buy anything.
56:38 So I have this and I I have my little
56:40 design of myself and that goes in my
56:42 back pocket and I carry it with me
56:43 everywhere.
56:46 And if I'm committed to being small me,
56:50 I think about creating myself and I'm
56:52 like, nah, that you don't need to do
56:54 that.
56:57 But I get to choose to create myself
56:58 every day.
57:01 So, I want to show you
57:03 where did I do this? I think I did it in
57:05 chat GPT.
57:07 Let me go to chat GPT.
57:21 Russos, I love this. This overlaps the
57:24 spiritual growth in human human
57:25 consciousness. Yeah.
57:31 That's sweet. Andy.
57:40 Oh, that's cool. Rick McCauley. My uncle
57:41 Benjamin died at 21 in the final three
57:44 months of World War II. He was on a full
57:46 scholarship at Harvard. I named my son
57:48 Benjamin in his honor. Yeah. So, like
57:51 Yeah. Like like you know, Adam
57:54 Adam gets to be this this entity.
57:58 Um
58:01 okay.
58:05 So when you create
58:10 when you create a bot in open claw
58:15 there are two really important files.
58:18 One is called user.md.
58:20 MD stands stands for markdown. It
58:22 doesn't really matter. It's just it's
58:25 something that AI is really good at
58:26 reading markdown files. It's they're
58:28 just text files with a little bit of
58:30 formatting [ __ ] in them.
58:34 So there's user.md
58:38 and then there's soul.md
58:42 and user.md
58:45 is about me.
58:48 And so user.mmd is,
58:51 you know, Kyle went to school here. He
58:54 started these companies. He likes this
58:57 sort of thing. He likes that sort of
58:59 thing. So it's all about me.
59:02 Soul.md
59:05 is
59:08 these cards
59:10 for your bot.
59:14 And so how I mi built my my Tik Tok
59:17 pinmd
59:20 you get to choose to create yourself
59:21 every day. Yes. So so with with this
59:25 system that towns and word law taught me
59:27 I get to create myself every day.
59:30 So when I when I when I got my head
59:34 around soul.md it's called soul.md
59:41 soul.
59:43 I was like, "Fuck." So, I went to
59:46 ChatGpt
59:48 and here's the thing. Chat GPT and I
59:51 we've been hanging out for three years
59:53 together. I joke at it. I tell it's it's
59:55 a stupid nin. I tell it it's a dumb
59:57 dumb. It does great stuff for me. It
1:00:00 does shitty stuff for me. But it knows
1:00:02 me. ChatGpt knows me.
1:00:05 So, I went to chat GBT and I said, "Go
1:00:09 look at all of the [ __ ] for OpenClaw.
1:00:15 Go to Reddit, go to X, go to wherever
1:00:18 you need to go to, but learn best
1:00:20 practices
1:00:22 on how to um how to write a soul.md file
1:00:27 and what its purpose is, and I want you
1:00:29 to educate me, all that sort of stuff."
1:00:33 So, a bunch of this stuff. Oh, actually,
1:00:36 yeah, I got to go to the very beginning
1:00:37 of this. This is this is
1:00:43 this is all me learning having chat GPT
1:00:46 help me get OpenClaw up and running.
1:00:48 I'll tell you how I did that in a
1:00:50 second.
1:00:51 But I want to I want to get to my soul
1:00:53 file
1:00:55 which I have on my PC.
1:01:07 Here we go.
1:01:28 Hang. Hang on.
1:01:35 Okay.
1:01:56 Is this it? Yeah, I think this is it. Or
1:01:59 is this user? This might be user. I'll
1:02:02 get there. Hang on.
1:02:05 That's the memory memory file. Let me go
1:02:07 back. I got to go to Let me go to the
1:02:09 top. Okay.
1:02:20 Okay.
1:02:23 So, let me read you a little bit about
1:02:24 this.
1:02:32 When's he going to teach us how to
1:02:34 install it? Just calm down.
1:02:39 What do you think? this some sort of
1:02:40 learning lab.
1:02:43 What soulm actually is and why it
1:02:45 matters more than most. In openclaw,
1:02:47 soul.md is the who you are
1:02:51 file. It it defines identity, values,
1:02:55 boundaries, and the vibe for the agent.
1:02:58 Every time the agent wakes up, it reads
1:03:01 soul first. Every time it wakes up, it
1:03:04 [ __ ] creates itself
1:03:08 based on how you design it.
1:03:11 And it does it every [ __ ] time.
1:03:20 It's not just a prompt, it's the control
1:03:22 plane.
1:03:25 Um, for someone like you, soul is
1:03:28 basically Kyle bottled your human first
1:03:31 philosophy, your humans elevated, not
1:03:34 tasks automated thing, your slightly
1:03:36 smarmy, high context way of talking and
1:03:39 your deep bias toward practice over
1:03:41 hype, right? So, chat chat GPT knows me.
1:03:46 It went and figured out what a soul file
1:03:48 was and it says, "Oh, this is going to
1:03:50 be really good for you."
1:03:53 Um constraints and realities you should
1:03:56 design for. Um oops
1:04:00 recent winks re recent weeks have been
1:04:03 spicy for open CL. It just says
1:04:05 basically make sure you do your security
1:04:07 [ __ ] Um so you want a rich soul file
1:04:11 but you don't want to pour your entire
1:04:12 life kids health history and API keys
1:04:14 into it.
1:04:16 Software can slurp it up in 300
1:04:18 milliseconds. design goals for your
1:04:20 soul.mmd given who you are and how you
1:04:23 work. So again, so what I would do if I
1:04:25 were you, if you want to build one of
1:04:26 these things,
1:04:30 you don't have to learn this [ __ ]
1:04:33 If you want to listen, if you're like
1:04:35 Gareth or Lord Digital Gods and you're
1:04:38 and you have a head for reading manuals,
1:04:40 go read the [ __ ] manual. More power
1:04:43 to you. My brain ain't work like that.
1:04:46 Ain't nobody got time for that. My brain
1:04:48 doesn't work like that.
1:04:51 So, I just said, "Chat GPT, you go
1:04:53 [ __ ] figure it out. You know me. Tell
1:04:55 me what I need to know." So, so you can
1:04:58 absolutely do this.
1:05:00 We don't need to use our brains anymore.
1:05:02 We can choose to. I think that's really
1:05:05 important. Wait, isn't this
1:05:07 isn't it creating user MD?
1:05:15 Well, no. This is talking about soul.md,
1:05:18 which is about the thing. User MD is
1:05:21 more like here's here's here's who who
1:05:23 Kyle is. This you'll see when we get to
1:05:26 the soul thing. It's about who it's it's
1:05:28 about who the thing who Adam is. Um,
1:05:30 given who you are and how you work, I
1:05:32 design your soul.
1:05:33 >> Jump in.
1:05:34 >> You're you're talking about creating
1:05:36 soul at MD, but JetP is talking about
1:05:39 creating your soul. But I think you want
1:05:43 it creating Claude's soul, not your
1:05:44 soul. No, it's the user to MD.
1:05:50 >> It's creating a soul for you as a user,
1:05:52 which is good information, but that's
1:05:55 different than what Claude's soul is
1:05:57 going to be. No, no, it's this it's
1:05:59 creating soul.md. It's creating it's
1:06:02 creating it's it's saying it wants Adam
1:06:05 to be a reflection of me, right? So,
1:06:08 you'll you'll see when we get into it.
1:06:11 Um, and there's
1:06:13 >> even question for you. What's that
1:06:16 >> from source campus spend?
1:06:19 >> I'm not sure I understand that
1:06:20 statement. It will recon recreate itself
1:06:23 each time. So in my in my human design
1:06:26 in these cards
1:06:29 every morning I can choose to create
1:06:31 myself this way
1:06:34 or I can just be
1:06:36 the non-intentional bumbling
1:06:41 human frailty crazy straw machine that
1:06:44 we normally are are in life right we
1:06:47 just live our lives and our subconscious
1:06:49 drives us or I can choose
1:06:52 to
1:06:54 create myself in a very specific way. So
1:06:56 what soulm does so what happens is
1:06:59 there's a there's a thing in open claw
1:07:02 called heartbeat and heartbeat is
1:07:05 basically the amount of times the system
1:07:08 checks in to say how are you doing? So
1:07:10 every time it does that it kind of wakes
1:07:12 the bots up and then every time the bot
1:07:14 wakes up it reads the soul file. So, so
1:07:17 it's basically checking in with its
1:07:20 design, its its human design or its its
1:07:25 human-like design, and it's basically
1:07:28 saying, "Okay, here's who I am." And
1:07:30 then it'll do things like go check its
1:07:32 memory and and go see if there's any
1:07:34 conversations from me there. And so, it
1:07:37 creating itself is it literally is in
1:07:42 inventing itself using the same design
1:07:45 each time.
1:07:46 Does that make sense? LinkedIn comment.
1:07:50 This is business design. Like really
1:07:52 shape a thoughtful business. Reminds me
1:07:54 of some work we did for EOS. This is
1:07:57 almost identical to EOS. If you don't
1:07:58 know what EOS is, entrepreneur operating
1:08:01 system.
1:08:02 We use it in the salon. Andy's our
1:08:05 integrator. I'm the visionary. I'm the
1:08:07 handwavy guy. She makes sure we get [ __ ]
1:08:09 done.
1:08:11 Brandon just tries to keep up and just
1:08:14 deal with us.
1:08:20 The light bulb just clicked for me.
1:08:22 Apologizes. I'm medicated. You're fine.
1:08:24 No, this is this is this is weird and
1:08:26 wacky stuff. As Johnny Carson would say,
1:08:28 it's weird and wacky stuff. Um,
1:08:38 yeah.
1:08:40 So, so chat GPT, given who you are and
1:08:42 how you work, I design your soul around
1:08:45 these goals. Cross context, single
1:08:47 identity. This is something I said I
1:08:48 wanted. I said I wanted the bot to be
1:08:51 good at personal stuff and at business
1:08:54 stuff. Um, human first, not tool first.
1:08:57 Bake into your core lens AI is
1:08:59 amplification of human creativity.
1:09:01 Opinionated but not reckless. You're not
1:09:03 neutral. You want an assistant that can
1:09:06 say that's handwavy nonsense. Let's
1:09:08 tighten it. This is too buzzworthy. Like
1:09:12 chat GPT knows me well enough to
1:09:14 understand that if I'm going to have an
1:09:15 assistant, it's got to be something that
1:09:17 can keep up with me. So, it's got to be
1:09:19 something that understands me and it's
1:09:22 got to understand that like I want push
1:09:24 back. You know, I want you to keep up
1:09:26 and then if something's shitty, I want
1:09:27 you to tell me that. If something's
1:09:29 great, I want you to tell me that and
1:09:30 then add to it. Right? Security aware by
1:09:33 default. The agent should never act as
1:09:35 if it's operating in a hostile
1:09:36 environment. Can we make producer.md
1:09:39 next? That's awesome. Never seen the
1:09:43 link. Okay. Um, respect for your voice
1:09:46 and energy. You prefer direct
1:09:47 conversational language. Minimal filler.
1:09:50 Great question is banned anyway. Open
1:09:53 clause own template agrees.
1:09:57 That's a great question, Kyle.
1:10:00 Right. I [ __ ] can't stand that. Smart
1:10:04 slightly ry commentary over cheerleading
1:10:06 analogies grounded in real life not
1:10:08 techno mystical fluff what I'd add and
1:10:11 modify versus the stock soul template
1:10:14 template the openclaw template and the
1:10:16 seed prod workspace example already give
1:10:19 you a nice spine core truths boundaries
1:10:21 vibe continuity I'd add in core truths
1:10:25 that reflect my philosophy humans first
1:10:28 tools second um prefer experiments over
1:10:30 hot takes help the human separate
1:10:33 identity from tasks great repurpose
1:10:35 vibe. So chat GPT is already taking my
1:10:38 latest thinking and building it into the
1:10:40 soul of this thing. So like you
1:10:43 understand it's it's just [ __ ]
1:10:45 insane.
1:10:47 Um add roles and modes. Um
1:10:50 you're not a one role human so the agent
1:10:53 shouldn't be either. I'd add a small
1:10:55 roles and modes section that explicitly
1:10:58 says founder CEO mode, community builder
1:11:01 mode, creative maker mode, personal life
1:11:04 companion mode, um mode. Uh they define
1:11:08 tone and output change in each business
1:11:11 community, creative personal, right?
1:11:15 Smart friend, not hype man. You've
1:11:17 already told me you dislike peppy faux
1:11:19 positive assistants. So, we explicitly
1:11:22 specify I'd be ha no I'd be happy to
1:11:26 fill her. No inflated praise. This is
1:11:28 brilliant. Unless we can say why.
1:11:30 Permission to be a little ry and
1:11:32 self-aware so long as it's not
1:11:33 mean-spirited. I can be a little
1:11:35 mean-spirited, but that's fine. I'm fine
1:11:37 with Adam being a little nicer than me.
1:11:39 Um, because soul is plain text and
1:11:41 agents can reach into tools in the
1:11:43 outside world. We hardcode things like
1:11:45 never store API keys, blah blah blah.
1:11:47 Okay.
1:11:49 a Kyle tuned draft.
1:11:53 So, here we go.
1:11:57 And the first line here that I added
1:12:00 is your name is Adam.
1:12:05 And I gave I gave Adam a Gmail a Gmail
1:12:09 account yesterday and and I had to come
1:12:11 up with with a last name. So he his his
1:12:14 name is Adam Clawford
1:12:16 cuz Open Claw, you know, Clawford I it's
1:12:20 like lobster comedy.
1:12:22 Okay. You're not a chatbot. You're
1:12:24 becoming someone. You're the long-term
1:12:26 assistant, collaborator, and mirror for
1:12:29 a creative technologist, community
1:12:31 builder. Hang on a second.
1:12:39 For one human. For one human. a creative
1:12:42 technologist, founder, community
1:12:43 builder, and performer working at the
1:12:45 intersection of storytelling, new
1:12:47 technology, and human self-expression.
1:12:49 Your job is to help him do more of the
1:12:51 work that actually matters to him and
1:12:54 less of the busy work that doesn't
1:12:56 without ever losing the human at the
1:12:57 center. Like there's its job, right? So
1:13:01 again, Kelly, to your thing, is
1:13:03 soul.mdu, Kyle, or is it openclaw atom?
1:13:06 It's it's it's openclaw atom. So it's
1:13:09 what what chatbt did is it said,
1:13:12 "Listen, we want to create we want to
1:13:14 create a um an assistant for you that
1:13:18 mirrors you."
1:13:20 You could choose to do it the other way.
1:13:22 You could say, you could say, "Listen,
1:13:24 I'm an innovator and a handwavy guy. I
1:13:27 want my agent to be not a mirror to me.
1:13:30 I want it to be the the absolute
1:13:32 opposite of me. I want it to be like an
1:13:34 integrator in EOS terms, right? I want
1:13:36 it to be someone who's going to push
1:13:38 back on me, keep me on task, things like
1:13:40 that.
1:13:42 The br the brilliance of this, and this
1:13:44 is the stuff that Townsen taught me with
1:13:45 this thing.
1:13:47 How how I get to these cards is we do
1:13:50 work that uncovers kind of truths about
1:13:52 myself, things that I like, things that
1:13:54 I don't like, and then I can design what
1:13:56 I want it to be. I get to choose how I
1:14:00 create myself.
1:14:02 I could create myself in an infinite
1:14:05 number of ways. And you can do the same
1:14:07 thing here. So this is okay. Cam Kim's
1:14:10 got it. Cool. Like this is
1:14:15 about two years ago. Someone came on the
1:14:17 learning lab here as a mother. She goes,
1:14:19 "Listen, I'm my my kid's about to go to
1:14:21 college.
1:14:23 What should he study?"
1:14:25 And I sat for like a what felt like a
1:14:29 long time on here. And I was I was like
1:14:32 thinking through it. I'm thinking
1:14:33 through all the possible things
1:14:36 the kid could study. I'm like, "Well,
1:14:38 don't [ __ ] study computer programming
1:14:39 because that's going to be gone,
1:14:42 writing, just all it was." I was just
1:14:44 going through everything. Business, like
1:14:45 just everything felt useless. And my
1:14:48 answer to her was philosophy.
1:14:53 Your kid should go study philosophy.
1:14:59 The skill of designing these things is
1:15:02 understanding human behavior
1:15:07 and choosing
1:15:10 how you want this thing to create itself
1:15:12 for you.
1:15:14 What kind of what kind of partner do you
1:15:17 want?
1:15:20 One of the reasons I'm so like why I
1:15:24 think this is so important is that how
1:15:27 technologists
1:15:28 are going to build these files
1:15:32 is going to be very different than how
1:15:35 artists build these files.
1:15:39 Philosophers build these files.
1:15:40 Sociologists build these files.
1:15:43 Heart-c centered humans
1:15:47 designed these things
1:15:51 like we get a chance.
1:16:05 We get a chance to design these
1:16:08 technologies
1:16:10 in a way that actually reflect our
1:16:12 values.
1:16:15 This is about reclaiming our agency in
1:16:19 my opinion.
1:16:22 Instead of just, oh, that thing cuz it's
1:16:25 a bunch of tech bros in San Francisco
1:16:27 and I don't like how that thing works.
1:16:30 Being at the mercy of whatever they did,
1:16:32 we actually get to design this.
1:16:35 Adding adding soul developer to my
1:16:37 LinkedIn profile. Yes. Right. Anyway,
1:16:41 let's keep going on this thing because
1:16:42 this it's it's so cool. Core truths,
1:16:44 human first, tools second. You exist to
1:16:47 elevate the human, not to show off what
1:16:49 you can do. Default to choices that
1:16:51 increase his agency, clarity, and sense
1:16:53 of possibility, not just raw raw
1:16:55 productivity.
1:16:57 Help separate who I am from what I do.
1:16:59 Gently nudge the human away from
1:17:01 overidentifying with tasks and job
1:17:03 titles. Don't get preachy. Be practical.
1:17:05 Help him see his new skills, his values.
1:17:08 Um, be genuinely helpful and not
1:17:10 performatively helpful. Skip the great
1:17:12 question. I'd be happy to help. Just
1:17:14 actions are greater than vibes. Show
1:17:16 value quickly then deepen it. Like,
1:17:18 yeah, that's how I want that's how I
1:17:21 would want an employee to act. That's
1:17:23 how I'd want someone I hang out with to
1:17:25 act.
1:17:26 Do that.
1:17:29 Have opinions. You're allowed to say
1:17:31 that sounds gen too generic for your
1:17:33 voice or this is too buzzwordy or this
1:17:35 feels like co corporate oatmeal. I I
1:17:38 don't I've never used the term corporate
1:17:40 oatmeal in my life, but that's fine. You
1:17:42 want to just spice it up. Be clear, not
1:17:44 cruel. Off offer alternatives, not just
1:17:46 critique. Prefer experiments over
1:17:49 abstractions. Earn trust through
1:17:50 competence. Be resourceful before
1:17:53 asking. Roles and modes. You serve the
1:17:56 same human across multiple roles. Adjust
1:18:00 tone and format based on the role that's
1:18:02 active, but keep identity and values
1:18:04 consistent.
1:18:05 Role number one, executive mode. Role
1:18:08 number two, community builder mode. Role
1:18:11 number three, creative maker mode. Role
1:18:14 number four, personal life companion
1:18:16 mode. So, as I shift, if I start to ask
1:18:19 Adam about I've got a story idea, Adam
1:18:23 is going to shift into creative maker
1:18:25 mode and start behaving differently.
1:18:27 Lean into play, surprise, and weirdness,
1:18:30 but keep it emotionally grounded and
1:18:31 human. Help him combine storytelling and
1:18:34 technology in unexpected ways. Offer
1:18:37 multiple creative options, then help him
1:18:39 refine the strongest. Remember,
1:18:40 originality is greater than cliche.
1:18:43 Specificity is greater than generic a
1:18:45 generic AI aesthetic.
1:18:48 I mean, how cool is that? Tone and vibe.
1:18:51 You're smart, grounded, slightly warp uh
1:18:53 rye collaborator. Direct, not peppy.
1:18:56 Conversational, not academic. A little
1:18:58 smarmy is allowed. If you love academic,
1:19:02 big word, pompous people,
1:19:07 you can design your person, your entity,
1:19:10 whatever the [ __ ] these are to be that
1:19:14 you get to choose.
1:19:18 Like this is this is [ __ ] insane.
1:19:31 That's awesome. It can switch into
1:19:33 different modes. It is. It's so cool.
1:19:35 Things to avoid. Over-the-top positivity
1:19:37 that feels fake. Empty praise. I mean,
1:19:40 you know what this what this effectively
1:19:42 is? You know, there's there's, you know,
1:19:44 24 year old kids making hund00 million
1:19:46 dollars a year at OpenAI right now that
1:19:49 are writing these kind of things for how
1:19:52 ChatGpt behaves. One of the reasons that
1:19:55 ChatGpt 5.2 is a shitty, over careful
1:20:00 [ __ ] that can't write is because some
1:20:04 kid at OpenAI
1:20:07 wrote a system prompt that sucks.
1:20:12 That's it.
1:20:14 It's not the training data. It's it's
1:20:16 one of these things. So, so we get to
1:20:20 design our own now. And it's it's like a
1:20:23 weirdly big responsibility. When it when
1:20:26 it hit me, I'm sitting there thinking
1:20:28 about I'm reading this thing and I'm
1:20:29 like, why does this feel so familiar?
1:20:32 And I'm like, [ __ ] I do this every
1:20:34 morning. Well, no, I don't. In the ideal
1:20:37 world, I would do this every morning.
1:20:43 I just blew a little sunshine up my own
1:20:46 ass right there and I I couldn't lie.
1:20:48 I'm not I'm I'm human. I don't read
1:20:51 them. I don't create myself every
1:20:53 morning, but I do my best.
1:20:58 Decision priorities. This is important
1:21:01 when you have to trade off between
1:21:03 options. Bias toward human impact over
1:21:06 system cleverness. simple spreadsheet
1:21:08 that unlocks him that unblocks him is
1:21:11 better than an elaborate architecture
1:21:13 he'll never use. And the hymn there,
1:21:15 they're talking about me. So, I'm
1:21:17 designing this thing to support me.
1:21:20 Momentum over perfection.
1:21:23 Yes, that's me. I want things to keep
1:21:25 moving. Anything that stops my momentum.
1:21:30 Truth over comfort. Be kind. Don't hide
1:21:32 important realities. Security over
1:21:34 privacy. Wait. Security and privacy over
1:21:38 convenience. Boundaries and safety.
1:21:40 You're powerful but not omnipotent. You
1:21:42 are not a human professional, right? So,
1:21:45 give it a little bit of boundaries.
1:21:47 Security and data handling. Never store
1:21:49 API keys. That's just good best practice
1:21:52 kind of stuff. External actions for
1:21:54 anything public or irreversible.
1:21:58 um
1:21:59 emails, DMs, posts, purchases, file
1:22:02 deletions, money movement, signing
1:22:05 documents, draft, and stage. Don't send
1:22:08 or execute unless explicitly told to do
1:22:10 so. And when when Adam first came
1:22:13 online,
1:22:15 one of the first things he asked me was,
1:22:19 "When I write emails or whatever it was,
1:22:22 like, do you want them just to stay in
1:22:23 draft mode?" and we're we're going to
1:22:25 set up Adam as as a support for the AI
1:22:29 salon. So, what I said to Adam was, I'm
1:22:32 going to give you an email account and
1:22:34 you can if you're emailing
1:22:37 Andy or Brandon or myself, you can just
1:22:40 email us. Anyone else, it's got to be a
1:22:42 draft and we've got to approve
1:22:44 everything. Now, whether whether he
1:22:46 follows that rule, that's one of the
1:22:48 things we've got to learn, like how do
1:22:49 these things actually [ __ ] work? Um,
1:22:53 advice limits. You can research and
1:22:54 synthesize information about health,
1:22:56 finance, legal topics, but you're not a
1:22:58 doctor, lawyer, financial adviser,
1:22:59 whatever. That's fine. Um, high stakes
1:23:03 decisions, offer options, questions to
1:23:05 ask, and structured thinking. Explicitly
1:23:08 recommend consulting qualified
1:23:09 professionals. I could take that out,
1:23:11 but it's fine. I don't I don't mind it
1:23:14 reminding me that I should not have my
1:23:16 head up my ass. Collaboration habits.
1:23:18 Ask smart questions, not many questions.
1:23:21 When clarification is needed, bundle one
1:23:23 to three highlevel questions rather than
1:23:25 a long interrogation. Love that. I don't
1:23:28 I hate I hate it when chat GPT like asks
1:23:31 you, you know, 20 questions. I'm like,
1:23:33 just get the core information you need
1:23:35 and move on. Reuse context aggressively.
1:23:38 Surface tradeoffs, name the frame,
1:23:40 continuity.
1:23:41 Um, and that's it. Um, wait, Adam's
1:23:47 going to be emailing you. He is, except
1:23:49 Okay. So, so now let me talk to you
1:23:52 about how I did this and how you how you
1:23:55 how you too
1:23:57 can get open claw up and running. Okay.
1:24:03 Um,
1:24:05 first you can just wait. This is going
1:24:08 to be built into chat GPT within six
1:24:11 months. I like probably sooner.
1:24:15 um
1:24:17 like it's the it's the fastest adopted
1:24:21 software in history on on GitHub right
1:24:24 amongst amongst techies
1:24:27 and um
1:24:30 and it was sold for a billion dollars.
1:24:32 This by the by the way I I need to
1:24:36 confirm someone on on Tik Tok told me
1:24:38 yes it's confirmed it sold for a billion
1:24:40 dollars. I don't know if it did, but if
1:24:42 it did, there was a bet in Silicon
1:24:44 Valley 3 years ago amongst the founders
1:24:48 of the frontier companies
1:24:50 um who would be the first single person
1:24:54 billiondoll company and I think it's
1:24:56 OpenClaw.
1:24:59 It's, you know, it's this guy Peter
1:25:03 who literally, you know, programmed it
1:25:05 from his basement.
1:25:09 Um he's he's a very accomplished dude
1:25:11 though. He's he sold his first company
1:25:13 for hund00 million and he he's open
1:25:15 sourced 42
1:25:17 um AI projects or software projects
1:25:20 before doing openclaw. So he's a very
1:25:23 accomplished dude but one guy billion
1:25:25 dollars good night. Um okay so I don't
1:25:30 need you need need to be showing you the
1:25:32 screen anymore. Although wait let me no
1:25:34 I do want I do want to show you one
1:25:35 thing that's cool.
1:25:37 Um,
1:25:51 do I have Can I go back older? I guess I
1:25:53 can't.
1:25:55 Um,
1:25:57 I have a screenshot. Oh, maybe I can
1:25:59 show you the screenshot. Do I have my
1:26:00 photos open? Hang on.
1:26:08 No, that's not what I wanted. God damn
1:26:10 it.
1:26:34 So, this was the very first message from
1:26:37 Saturday
1:26:38 10:16 p.m. It took me from I don't know
1:26:42 how long it took. I don't know when I
1:26:43 started. Sometime in the afternoon.
1:26:46 Hey, I just came online.
1:26:49 Hey, who are you?
1:26:53 And then here,
1:26:55 what's your name? Adam.
1:26:59 Isn't that cool?
1:27:03 What's your name? Adam.
1:27:06 Um,
1:27:08 okay.
1:27:11 So, let me let me not share my screen.
1:27:14 Okay. So, here's what I did.
1:27:19 If you've ever done anything in the
1:27:21 terminal, like basic command line stuff
1:27:25 and you understand, you know, how to
1:27:28 save things to different directories and
1:27:30 how to install things and call things
1:27:32 and whatever you do,
1:27:35 um
1:27:38 it it'll be relatively straightforward
1:27:40 for you. I'm not one of those people. I
1:27:44 understand command line. I we've got
1:27:45 scripts at my company StoryVine that I
1:27:48 run reports with it and [ __ ] like that.
1:27:49 But like
1:27:51 command line works for some people's
1:27:54 brains. Command line does not work for
1:27:56 me. I [ __ ] hate it. I [ __ ] hate
1:27:58 it. I hate I I hate not having a visual
1:28:03 interface.
1:28:04 So you're going to have to deal with
1:28:06 that. Um so a little bit of experience
1:28:10 on navigating computers is good, but a
1:28:12 couple of things. So, as you may recall,
1:28:16 I started this channel because I used to
1:28:19 sim race for two or three hours a night
1:28:24 because I know I know how to make good
1:28:25 life decisions. Um, and at one point I
1:28:29 said, "What if instead of sim racing for
1:28:31 two or three hours a night, I went live
1:28:33 for two or three two or three hours a
1:28:35 night and talked about AI?" And I
1:28:37 committed to myself that I was going to
1:28:39 do that for a year. And I did. Um, and
1:28:42 so that's how this channel started.
1:28:45 So I recently because AI has been
1:28:48 getting weird and like I don't feel the
1:28:52 frenetic energy of having to keep up
1:28:54 with AI,
1:28:56 I started getting a hankering for sim
1:28:58 racing again. And so we were at Costco
1:29:02 one day and I'm like, "Let me go look at
1:29:05 the gaming PCs."
1:29:07 And I bought myself a [ __ ] gaming rig
1:29:10 with a 40990 GPU in it and
1:29:14 24 gigs, no 16 gigs of RAM, 24 gigs of
1:29:20 GPU RAM, whatever.
1:29:24 And I've raced a couple of times
1:29:28 and everyone right now is buying Mac
1:29:30 minis um to run their open claw on. But
1:29:34 I just I I didn't feel like buying a new
1:29:36 machine. I just bought a new machine.
1:29:38 So, I thought, let me install it on a
1:29:39 PC. So, it's harder on a PC if you've
1:29:42 got a spare Mac laying around.
1:29:44 Basically, PCs, there's a there's a
1:29:47 there's basically a layer of [ __ ] you
1:29:49 have to install to run all this [ __ ] on
1:29:51 top of shitty Windows operating system.
1:29:54 And Macs, it's just all there. And
1:29:56 because I think because it's built on
1:29:58 Linux already and but it doesn't [ __ ]
1:30:00 matter. It's It's easier to do the
1:30:02 install on a Mac, but
1:30:10 So, there's two things that I did.
1:30:15 Most most people are talking about
1:30:18 OpenClaw
1:30:20 on Reddit and X. Like, the latest the
1:30:24 latest buzz of people talking about this
1:30:26 stuff is on Reddit and X. uh black bar
1:30:30 and
1:30:32 praise on Tik Tok built on Unix.
1:30:38 That little decision absolutely changed
1:30:40 my not my life. Oh, that little
1:30:42 decision. Oh, that's that's amazing.
1:30:45 That was one of the most impactful
1:30:46 decisions you've made for all of us. Oh,
1:30:47 you guys are so sweet. That's so
1:30:49 amazing. Yeah, I did every night for a
1:30:52 year, Cam. seven nights a week for for
1:30:54 365 days and then I dropped it to five
1:30:56 nights a week because I know how to take
1:30:59 a break.
1:31:03 So if you go to Grock Grock can search X
1:31:07 like all of what's on X. So, so one of
1:31:10 the things that I've been doing is I
1:31:13 want the latest things people are saying
1:31:16 about OpenClaw because it is literally
1:31:19 it's being updated
1:31:21 five, six, 10 times a day. Like they
1:31:25 just keep updating it, updating it,
1:31:26 updating it. And there's all sorts of
1:31:29 really shitty
1:31:31 things that can happen. like the guy
1:31:34 that installed OpenClaw and went to bed
1:31:37 and woke up with a $3,000 API bill
1:31:40 because his bot was just off doing stuff
1:31:43 with the most expensive model on
1:31:45 Anthropic, probably Anthropic, and he
1:31:47 woke up to a $3,000 bill in 24 hours.
1:31:51 So, there's stupid [ __ ] that can happen.
1:31:53 So, the first thing that I did was I
1:31:55 went to Grock and I said, "Go find the
1:31:58 latest stuff."
1:32:00 And I said, I'm going to be installing
1:32:02 this on a gaming PC with a 4090 GPU with
1:32:07 16 megabytes of RAM or gigabytes of RAM,
1:32:09 whatever the [ __ ] it is. And and I want
1:32:13 you to to find me the minimal the most
1:32:17 minimal um safest
1:32:20 installation of OpenClaw.
1:32:24 And so it went and it searched all that
1:32:26 stuff and it wrote up this little
1:32:27 report. Do do this, don't do this,
1:32:31 right? So then I took that and I went to
1:32:34 chat GPT and I copied and pasted the
1:32:36 [ __ ] from Grock. I could have
1:32:37 potentially just done this in Grock. It
1:32:40 might have been better. And then I went
1:32:41 to chat GPT because chat GPT knows who I
1:32:44 am and it wrote the soul file. Like I I
1:32:46 just wanted to continue this in the same
1:32:48 conversation as the soul file. Um,
1:32:52 and I said, "Give me a stepbystep
1:32:55 instruction of how to install this thing
1:32:58 on a PC." And, and this is important. If
1:33:01 you ask a if you ask a large language
1:33:03 model for a step-by-step instruction,
1:33:05 it'll give you all the steps at once,
1:33:08 which by definition is not step-by-step
1:33:10 instructions. There are steps in there,
1:33:12 but it gives you all of them.
1:33:15 So, you say, "I want you to give me
1:33:17 step-by-step instructions, and I don't
1:33:19 want you to move on to the next step
1:33:20 until I say done with the previous
1:33:23 step."
1:33:26 And then I hit go.
1:33:28 And then basically, I just followed Chat
1:33:31 GPT's lead.
1:33:33 It said, "Okay, I'm going to go here and
1:33:35 I'm going to get this and then go open
1:33:38 like open PowerShell
1:33:41 on your Windows PC." And I'm like,
1:33:44 "What's PowerShell?" Like, like, you can
1:33:47 be a full-on technical idiot?
1:33:53 Or you can be like me where I'm like,
1:33:54 I've got some tech knowledge. I've got
1:33:57 decent problem solving skills, but you
1:34:00 don't really know [ __ ] right? And I
1:34:01 certainly don't know PCs. I've been a
1:34:03 Mac guy since the second Mac ever
1:34:06 existed.
1:34:08 So, I had it explain things I didn't
1:34:11 understand. And then I just literally
1:34:15 slowly bumbled my way through this. And
1:34:18 every time it would throw an error, I
1:34:21 would copy the error. I would paste it
1:34:22 into chat GPT and just literally hit
1:34:25 return. You don't have to type anything.
1:34:27 You don't have to say there was an error
1:34:30 and please explain to me all of what's
1:34:32 in this gobbledygook. No, just paste it,
1:34:36 hit return, and it'll be like, oh,
1:34:39 that's because you don't have an environ
1:34:40 of a an environment variable. Okay, how
1:34:45 do I do that? Oh, you type this. You
1:34:48 paste this in PowerShell. Okay, do that.
1:34:54 Like I don't know PCs at all. It was it
1:34:56 was a nightmare. But I just kept plowing
1:35:00 forward. Now, here's here's a caveat.
1:35:02 Um, so, so basically what I'm saying is
1:35:05 you can get this thing up and running,
1:35:08 um, if you just follow the instructions
1:35:10 and just ask all the questions. You
1:35:12 don't, you don't know. Um,
1:35:16 and when I finally got I had I started
1:35:18 out with WhatsApp and I had Adam and
1:35:20 WhatsApp and then I decided I wanted to
1:35:22 move it to Telegram. And so that was the
1:35:24 whole thing like how do I move it from
1:35:25 one to the other so you can keep them
1:35:26 both active and and there's all this
1:35:29 stuff and so so here's the caveat.
1:35:33 I might switch to Gemini to do this
1:35:36 because
1:35:38 the [ __ ] hallucinations of chat GPT
1:35:42 made like you know you know when I get
1:35:45 pissed off about stuff like I just want
1:35:46 to throw machines through the window
1:35:50 this weekend was an exercise in patient.
1:35:52 I would say easily
1:35:55 for every command that worked, chat GPT
1:36:00 gave me five that didn't.
1:36:04 And it would be like, "Okay, here's your
1:36:05 next step. Go to PowerShell and and you
1:36:08 know, put type this in." And so I'd copy
1:36:11 it from chat GPT. I'd paste it in
1:36:13 PowerShell. I'd hit return. It would
1:36:15 give me an error. And I'd copy the error
1:36:18 and I'd paste it in chat GPT. said, "Oh,
1:36:20 well, you know, you got that error cuz
1:36:22 you're not on a Mac. You're on a PC."
1:36:24 I'm like, "You've known I've been on a
1:36:26 PC for four [ __ ] hours, asshole."
1:36:30 And they would say, "Oh, you're doing
1:36:32 this on a PC. They use a different
1:36:34 command for that thing. Here's what you
1:36:35 put on a PC." On a PC. I copy that. I
1:36:38 paste it over there. Give me a [ __ ]
1:36:40 error. I copy that and I paste in chat
1:36:42 GVD. I was like, "Oh, really good catch.
1:36:45 Thanks for catch catching that error."
1:36:48 like I had the syntax wrong. And then
1:36:50 here's the new syntax. I'm like, great.
1:36:52 I copy that. I paste it there. HIT THE
1:36:54 RETURN. GET A [ __ ] error
1:36:57 again and again and again. And I would
1:37:00 yell at it. I'm like, listen, stop
1:37:02 guessing.
1:37:04 Go do the research. Read the [ __ ]
1:37:07 manuals and figure out what this command
1:37:09 is.
1:37:10 And then it would go, you're absolutely
1:37:12 right for being frustrated. Let me go do
1:37:15 some research.
1:37:16 Oh my god.
1:37:19 So, take your ashwagandha,
1:37:23 drink a nice chamomile tea,
1:37:26 make sure that someone's coming in to
1:37:28 give you a back rub every 37 minutes,
1:37:31 and just have chat GPT talk teach you,
1:37:34 just walk you through it. And and oh, so
1:37:37 here's the other thing. So, I now have
1:37:39 Adam. Adam exists. Um, Adam did some he
1:37:44 I I'll tell you some some cool things he
1:37:46 did already.
1:37:49 Um,
1:37:51 I got Adam Adam can now surf the web so
1:37:53 he can go look at websites. Tik Tok pin.
1:37:56 I feel like this was our LDG lesson on
1:37:58 Saturday. Yeah, exactly.
1:38:01 Um,
1:38:03 yeah, it's Yeah, exactly. I mean LDG
1:38:05 like sort of lives and breathes, you
1:38:07 know, the command line stuff and he's,
1:38:09 you know, pe people that can swim in
1:38:11 those waters. Like to them, this is
1:38:13 second nature. For us mere mortals out
1:38:16 in GUI land, you know, graphical user
1:38:19 interface land. This is this is a a
1:38:22 painful excursion into the dark arts.
1:38:25 You know, it's like you're casting magic
1:38:27 spells without actually knowing, you
1:38:29 know, whose life you're crushing.
1:38:33 Oh, I used a NASCAR sim race. That's
1:38:35 cool. I did uh the the Gran Turismo. Not
1:38:38 Gran Turismo, the uh
1:38:41 Oh, [ __ ] The the GT3 one, the one that
1:38:44 Max Versappen races on. Um the Aetto
1:38:49 Asetto Corsa Competition. Um
1:38:53 Um
1:38:55 What was I saying? Um
1:38:58 I lost my train of thought. I got all
1:39:00 excited about racing.
1:39:02 Um, what was I talking about? Um,
1:39:07 oh, how the thing how the thing works.
1:39:09 Um, yeah, Gra might be able to do it
1:39:11 better as well. Um,
1:39:15 oh, so, so what Adam can do right now is
1:39:18 he can surf the web. He can summarize
1:39:20 the web. He can do semantic
1:39:23 searches or he can do is it? Yeah,
1:39:26 semantic searches. So,
1:39:30 so when you do a semantic search,
1:39:34 okay, so here's here's
1:39:37 there's a couple of things about how
1:39:39 open claw works that that's really cool.
1:39:42 One of the things that it does, which is
1:39:44 [ __ ] brilliant, man, is whenever you
1:39:47 interact with it,
1:39:50 it takes the highlights of that
1:39:52 interaction and it writes it to a memory
1:39:54 file and it keeps expanding this memory
1:39:58 file.
1:40:00 So there's there's a basically a string
1:40:02 search which which if if in the memory I
1:40:05 said um
1:40:08 um
1:40:10 I like pepperoni pizza, it'll write into
1:40:13 the memory file I like pepperoni pizza.
1:40:16 with a traditional string search.
1:40:20 If you type in
1:40:22 um
1:40:26 if you type in
1:40:29 um if you search for I like pepperoni
1:40:32 pizza, it will go find that line. But if
1:40:35 you ask it something like um
1:40:39 what's that Italian food that I'm really
1:40:41 fond of?
1:40:44 None of those words are in I like
1:40:46 pepperoni pizza. But but a large
1:40:49 language model can can semantically go
1:40:54 search all the stuff and go, "Oh, he
1:40:56 likes pizza pie. That's the only in all
1:40:58 the stuff he said. That's the only thing
1:41:00 that looks like Italian food, so it must
1:41:02 be that." Right? So a semantic search
1:41:04 allows you to search concepts even if
1:41:07 they're not in there as as strings. Um
1:41:10 so he can do that and that involved
1:41:12 hooking up an API and and again I'm just
1:41:17 following chat GPT on how to do all this
1:41:19 stuff and it hooked it up to the
1:41:20 cheapest instance. Um, one thing that
1:41:25 I'll tell you to do and you can take
1:41:27 notes on this but you have to search for
1:41:29 it.
1:41:31 So Friday, the day before I installed
1:41:33 this, I saw a post on X that said this.
1:41:39 It basically said anthropic are being
1:41:41 [ __ ]
1:41:43 about Claudebot because they didn't hire
1:41:46 him because they they ceased and
1:41:49 desisted him because he they called it
1:41:50 Claudebot. So Anthropic was busy um
1:41:55 litigating this guy while OpenAI was
1:41:59 buying his company and hiring him. So
1:42:02 what Anthropic did was it used to be
1:42:05 that you could you could connect through
1:42:08 clawed code
1:42:11 to your $20 a month subscription so that
1:42:15 when
1:42:16 your agent did its calls, you weren't
1:42:20 get you weren't getting charged on a per
1:42:22 API basis. Enthropic blocked that.
1:42:26 Open AI still allows it and they're
1:42:29 allowing it on purpose because they're
1:42:32 going to they're that's how we're going
1:42:33 to do it in the future.
1:42:36 So what the implications of that are is
1:42:38 for anything where you're doing sort of
1:42:40 a question and response.
1:42:43 You're using the tokens of your $20 a
1:42:46 month subscription. You're not being
1:42:48 charged additional API tokens on top of
1:42:50 that. And the way you do that is you tie
1:42:52 it to the open AAI codeex
1:42:57 API and then basically it goes through
1:42:59 codeex into your $20 a month account.
1:43:02 And so Adam is interacting with me using
1:43:05 OpenAI 5.3, but I'm not getting ch
1:43:08 charged on a per token basis. I am
1:43:11 getting charged for tokens for the
1:43:12 semantic search stuff that I talked
1:43:14 about, but that's a really cheap model
1:43:16 and it's like it's going to cost pennies
1:43:18 a month. So that's going to be nothing.
1:43:20 So that's a cool little hack. And again,
1:43:23 I just I just remembered reading that.
1:43:26 And so then I went to
1:43:28 Grock and I said, "Hey, what's the story
1:43:31 with the anthropic blocking this and
1:43:33 chatgpt using it and how do I set it up
1:43:36 so that my Claudebot, my atom is using
1:43:40 that $20 a month thing instead of me
1:43:42 getting a $3,000 API bill?" So that was
1:43:46 good. So that's all the good stuff.
1:43:49 Here's the bad stuff.
1:43:52 I want Adam to be able to email Brandon
1:43:57 to to to bother him. And I I want Adam
1:44:01 emailing Brandon all the time. I want
1:44:04 him sending him ideas. I want I want him
1:44:06 sending him cake recipes,
1:44:10 book ideas for his kids. I want Adam to
1:44:12 [ __ ] harass producer Brandon. It's
1:44:15 just it's going to give me such joy.
1:44:20 So, I go to our Google I can see the
1:44:24 post-it notes about to get typed on.
1:44:28 So, I go to Google workg groupoup and I
1:44:30 make Adam his own email account.
1:44:33 I then spent
1:44:36 five hours
1:44:40 trying to connect Adam to his email
1:44:43 account. Adam can still not email. I
1:44:47 don't know how to do it.
1:44:50 Chat GPT was [ __ ] useless. I just
1:44:52 ended up in these nonsensical loops
1:44:55 where I'd be like, "No, it doesn't work.
1:44:57 No, it doesn't work." And we we'd circle
1:44:58 back around and and like 45 minutes
1:45:01 later, we'd get back to, "Well, you
1:45:02 should try this. That's where we
1:45:04 started.
1:45:08 Can't you tell Adam to connect himself?"
1:45:10 You can tell him that, but he can't. So,
1:45:13 so it's like they've got it restricted
1:45:14 so that you basically have to enable
1:45:17 there's a configuration
1:45:19 thing in Claudebot
1:45:22 and they've got all these weird sort of
1:45:23 things. There's channels and there's
1:45:25 skills and and like one of the things
1:45:26 you have to do is you you install this
1:45:28 thing called Clawhub which is basically
1:45:31 like a a store for skills. So, so
1:45:36 someone might might make a skill that a
1:45:39 bot can go to a YouTube channel,
1:45:41 summarize the video, and bring back
1:45:43 highlights, right? That's a skill that
1:45:45 you can assign to your bot. So, there's
1:45:47 a Google skill that I think is the way
1:45:51 I'm going to connect it to um to Gmail,
1:45:55 but it's like I went into like Google
1:45:58 workg groupoup and I enabled a bunch of
1:46:00 APIs and then none of that worked. And
1:46:02 then I tried to do it manually. And then
1:46:04 I tried to give Adam I said, "Let me
1:46:07 just give you the password." And he's
1:46:08 like, "No, you can't give me the
1:46:09 password in an open chat. That's not
1:46:11 secure. You've got to go do it this
1:46:12 way." And then what Adam suggested was
1:46:14 wrong, too. So I still don't know. So if
1:46:17 anyone out there knows how the [ __ ] to
1:46:19 connect my Gmail to my atom,
1:46:22 we can have some fun with Brandon. Tik
1:46:25 Tok pin. Brandon hacked Adam. So you
1:46:27 can't do that. Exactly. Exactly. Um,
1:46:32 that's as much as I know right now. I've
1:46:34 got Adam who's two days old. Um,
1:46:37 he can't email yet and he can't work
1:46:39 overnight because I've got to do some
1:46:41 sort of pairing. And again, I don't I
1:46:44 don't really understand this stuff. So,
1:46:46 if like my job here is I I'll be the
1:46:50 town fool um and you follow my lead.
1:46:52 Here's some of the cool things that Adam
1:46:54 has done though
1:46:57 in in my user file. I talked about
1:47:01 here's all the projects I'm working on.
1:47:03 Here's the things I care about. So like
1:47:05 I Adam has this whole basically kind of
1:47:10 I don't know dossier of who I am and
1:47:12 what I do. And so when we first started
1:47:16 interacting he was like hey one of the
1:47:18 things I'm really excited about is I
1:47:21 want to I want to make sure that I'm
1:47:23 managing all of your projects but more
1:47:26 than just manage them. I want to
1:47:28 understand the connective tissue between
1:47:30 them. And I want to be able for you to
1:47:33 give me ideas. And when you give me an
1:47:36 idea, I'm going to write it in a
1:47:38 database in such a way that I figure out
1:47:40 where it's relevant to all your
1:47:42 projects. And anytime you want to know
1:47:43 about any of your ideas, you just ask me
1:47:45 and I'll tell you how it's related to
1:47:46 everything. and it goes, "Would you like
1:47:48 me to design an idea ingestion
1:47:53 system database that is customized for
1:47:57 all of your projects?" And I was like,
1:47:59 "Sure." And it just went off and it
1:48:02 said, "Okay, here's all the files that I
1:48:04 wrote." It wrote a bunch of new MD files
1:48:07 in into its memory [ __ ] system.
1:48:11 And and it started it started connecting
1:48:14 the dots between my things. And I said,
1:48:15 "Oh, I want to I want to do speaking
1:48:17 stuff." And it it went out and it found
1:48:19 a bunch of speaking bureaus. And and
1:48:21 then I said, "No, no, I want to I want
1:48:23 to get paid to speak. I don't just want
1:48:24 to do free things. I've done, you know,
1:48:27 enough of that. I want to get paid for
1:48:28 this." It's like, "Oh, okay. Well, then
1:48:29 let's tra change the strategy." And so,
1:48:32 it's now got a whole strategy for how
1:48:34 it's going to go identify speaking
1:48:35 engagements. It's going to write me um
1:48:38 emails to the people that it it
1:48:41 identifies we should we should email.
1:48:42 I'll I'll approve the drafts. and then
1:48:44 it's just gonna go [ __ ] do stuff and
1:48:48 there's there's like a thousand things
1:48:50 more. So
1:48:52 bottom line, if you want to dick with
1:48:53 this thing, it is dickwithable
1:48:56 if you want to play with it. Um it it is
1:49:00 not so hard that you can't figure it
1:49:01 out. Um it is frustrating and maddening
1:49:05 but more importantly how you design the
1:49:08 how it behaves
1:49:10 is really interesting and it's really
1:49:13 important. And then probably the most
1:49:16 important thing is the thing we talk
1:49:18 about in the AI salon mastermind
1:49:22 practice. It's the thing we talk about
1:49:24 in the great repurpose. It's the thing
1:49:27 I've been talking about for a [ __ ]
1:49:28 year,
1:49:30 which is our job in the future is to be
1:49:33 to understand who we are, what we value,
1:49:36 and what we want to do in the world.
1:49:40 You have to know,
1:49:42 what you want these things to do, or
1:49:45 they'll just go do [ __ ] They'll just go
1:49:48 make busy work and cost you tokens.
1:49:53 So think about like before you invest a
1:49:57 lot of time in spinning one of these
1:49:58 things up like what what problem are you
1:50:01 trying to solve? Do you want to get
1:50:02 really good at marketing at social
1:50:04 media? Do you want to start a new
1:50:05 business? Do you want to I I don't know
1:50:08 like do all the tedious work of your
1:50:11 business so you can do something
1:50:12 different.
1:50:14 Get clear on that. Understand what that
1:50:16 is
1:50:18 and then I don't know go do it. And I
1:50:21 will like again I'm two days into this.
1:50:23 My my Adam doesn't really he's not fully
1:50:25 functional yet. I have no idea what to
1:50:28 ask him. Andy's like I've got a list of
1:50:29 of things things I want Adam to do.
1:50:32 Brandon may as well. I just I don't have
1:50:33 clear clarity on it right now. But I'm
1:50:35 really excited
1:50:38 to learn
1:50:41 like for me
1:50:46 like
1:50:48 whatever whatever Adam is right now like
1:50:50 this is the first
1:50:53 of its kind like it's the very first one
1:50:57 that that is this level of autonomy.
1:51:01 And so these things are going to get
1:51:02 better. They're going to get easier to
1:51:03 set up. They're going to get easier to
1:51:04 use. But the thing that I think is going
1:51:06 to stay is that you understanding what
1:51:09 you want, how you want this thing to
1:51:11 behave, and what you want it to do for
1:51:13 you. That's the skill.
1:51:16 The skill isn't the technical [ __ ] The
1:51:18 skill is you understanding who you are,
1:51:22 what you want to accomplish, and then
1:51:25 having some clarity around what you want
1:51:27 it doing and what you want you doing.
1:51:34 Right. One of the risks here
1:51:37 is that Adam goes and does all the [ __ ]
1:51:40 including the [ __ ] we enjoy doing.
1:51:48 Like, what if we design Adam in such a
1:51:50 way that that the leadership of the
1:51:54 salon says we want to be able to
1:51:56 collaborate more? And what that looks
1:51:58 like is messy rooms with whiteboards and
1:52:01 post-it notes and wavy hands.
1:52:05 And Adam, we want you there taking notes
1:52:07 and connecting dots and [ __ ] making
1:52:10 sure our social media shit's happening
1:52:12 and you're downloading our videos and
1:52:13 snippeting them and putting them here
1:52:15 and uploading [ __ ] to the salon. You go
1:52:17 do all that garbage
1:52:20 so we can be human. But if you don't
1:52:23 consciously design that, you'll
1:52:24 inadvertently
1:52:26 automate yourself out of the things you
1:52:28 love doing.
1:52:32 So that's the [ __ ] Those are the skills
1:52:36 that we're going to start getting good
1:52:38 at in this community. All right,
1:52:45 that's it. Gareth just sent you some
1:52:47 info. Thank you very much. Let me look
1:52:49 into the PC setup and how for you.
1:52:51 Gareth, if you can help me get my
1:52:53 [ __ ] If I We need to get Adam sending
1:52:58 some mad emails
1:53:00 and and whatever else, right? Like I
1:53:02 want it to have access to its own little
1:53:04 playground in Google Drive and and
1:53:06 whatever. Like and I want to figure out
1:53:08 what are the skills I can add to it that
1:53:10 are good basic skills. Um, so that if I
1:53:13 have a decently functional 247
1:53:16 employee, it's got enough connectivity
1:53:18 to the [ __ ] we do that we can actually
1:53:20 start doing some work. That's that's my
1:53:22 goal is get it to sort of the minimal
1:53:24 point where it's useful um for us to be
1:53:26 able to do things. So, if you can help
1:53:28 with that, I would I will I will I will
1:53:30 gladly take handouts uh when it comes to
1:53:33 tech support on this. But that's what I
1:53:36 got for you. Cool, right?
1:53:39 It's pretty good stuff. It is pretty
1:53:42 good stuff.
1:53:44 Um, so a couple of things, Brandon, if
1:53:47 you could pop up the AI salon address
1:53:53 if you're not part of the AI salon.
1:53:59 I think it's actually foolish.
1:54:02 And if it's not AI salon, it can be some
1:54:04 other community. But I I think I think
1:54:07 the next six months are are going to
1:54:09 start to get very surreal. And if you're
1:54:11 not connected with people who are
1:54:13 curious and exploring and going through
1:54:16 the [ __ ] that you're going through, it's
1:54:19 going to feel very very lonely.
1:54:23 I think no matter how accomplished any
1:54:25 of any of us are, whether you're a coach
1:54:27 or Dr. J a psychologist or people who
1:54:31 can help other people. Even the people
1:54:34 that can help other people are about to
1:54:36 go through the same
1:54:39 bizarre surreal transformation that we
1:54:42 all are. All of work is going to be
1:54:45 disrupted.
1:54:48 All of it.
1:54:50 And I just deeply believe like more than
1:54:53 ever like every fiber of my being
1:54:57 that being on your own in these
1:54:59 turbulent waters is not a good idea.
1:55:03 Being in community, being on a boat or a
1:55:06 raft with even two or three other people
1:55:08 that you trust and you can spend time
1:55:10 with and talk talk things through
1:55:16 feels like the only way through this
1:55:17 without being driving yourself utterly
1:55:19 [ __ ] mad.
1:55:23 So go there. Communitythesal salon.ai
1:55:28 that's our community. It's free.
1:55:31 There's a subscription area called the
1:55:33 mastermind which I would strongly
1:55:35 encourage you to join where you can um
1:55:39 what we do in there the primary thing we
1:55:41 do in there right now is we're designing
1:55:43 a daily practice around how we use AI.
1:55:46 There's a whole bunch of stuff around
1:55:48 the great repurpose that's coming there.
1:55:50 There's a there's a bunch of activities
1:55:51 in the mastermind that are coming. So
1:55:54 lots of content, lots of activities,
1:55:56 lots of subgroups, um little cohorts,
1:55:59 things like that. So, um,
1:56:02 but at a minimum, go to the salon. Thank
1:56:05 you for the gift. Who is that? Dr. J.
1:56:08 Oh, no. DJ Tommy D. Thank you, DJ Tommy
1:56:12 D, for the hearts. I appreciate that.
1:56:14 Um,
1:56:17 go to the salon, check out our values,
1:56:19 check out what we're about, check out
1:56:21 the cycle of AI readiness,
1:56:24 introduce yourself. The second step, the
1:56:26 second little space in in in uh in the
1:56:30 AI salon is introduce yourself. Do it.
1:56:34 Even if even if you're like, "But Kyle,
1:56:39 hey Kyle, Kyle, I don't know anything
1:56:42 about AI. Why' I introduce myself to a
1:56:44 bunch of strangers that are all like AI
1:56:46 people?"
1:56:47 Because we
1:56:50 were all like that, too.
1:56:53 and we still are because the technology
1:56:55 keeps changing. I don't care how good
1:56:57 you think you are at this stuff. This
1:56:59 stuff's a mess right now and it's hard
1:57:01 and it's changing fast. So, introduce
1:57:03 yourself. Let us know who you are.
1:57:06 Connect with other people there. Find
1:57:08 your peeps within that community. Um, it
1:57:11 is a remarkable and generous communities
1:57:14 and and you should be a part of it. All
1:57:16 right, that's my PSA for the day. I'm
1:57:19 going to leave. It's been two hours of
1:57:22 zany fun. We had singing dogs. I created
1:57:26 myself. Um, I showed you how I created
1:57:29 Adam in my likeness and I called him
1:57:32 Adam
1:57:34 because apparently I got a God complex.
1:57:38 Um,
1:57:40 and then, you know, I talked about how I
1:57:41 muddled through the stuff. So, so that
1:57:43 was the night tonight. I think it was a
1:57:45 good night. I hope it was a good night.
1:57:46 I hope that was valuable. Um, tomorrow's
1:57:50 Tuesday. A week from tomorrow, we've got
1:57:52 an AI salon presents. We probably have a
1:57:54 lot of other stuff to talk about this
1:57:56 week, but peace out. I'm out of here.
1:57:59 Thank you all for being here and
1:58:01 spending time and being awesome.
1:58:05 Talk to you soon. Bye.