What's up, everyone? Welcome back to

another episode of Human in the Loop,

the most actionable AI show on the

internet. I'm one of your hosts, Alex

Lieberman, co-founder and managing

partner at 10x, and this is one of the

best conversations that I've had since

starting the show. We have an expert AI

builder who takes us through how to have

a 10x setup with Claude CoWork so it can

truly become your AI chief of staff, and

then three amazing use cases with

CoWork, building your own CEO coach,

building an advisory board, building a

thought leadership skill that helps you

create content online. You're going to

learn so much in this episode, and it is

truly going to turn Claude CoWork into

your AI work companion. Let's hop into

it.

JJ, thanks for joining the show. It is

my pleasure to be here, Alex. Thank you

for having me. We did it. So, if folks

watch until the end, what can they

expect to get out of this? They'll get a

CoWork setup that acts like a their

executive assistant rather than just

like a chatbot. You know, something that

they can use in many aspects of building

with AI. So, they can not only chat with

AI, but they can do with AI. Love it.

And

you've been creating a lot of content

recently about CoWork, and some people

that are watching may know what Claude

CoWork is, other people may not. Why do

you think CoWork is such a powerful

tool? Yeah, so I use Claude Code, I use

CoWork, and I use Open Claude, and

they're good in each of their own ways,

right? Claude Code is really meant for

like building stuff. Like, I want to

build real infrastructure. CoWork, its

bread and butter is in the everyday

productivity. I need to connect Slack,

Notion, uh Jira, whatever, bring all

that together, and then layer in an LLM

LLM on top of it to use my computer to

like help me do real stuff, but with my

own tools. I have like morning recaps of

like strategizing like me for the day,

looking at my calendar, looking at my

email, my Slack events, preparing me,

things that just come together so easily

that I can use now with AI to just

supercharge me from a productivity

standpoint. And then open claws like a

really good everyday like little

assistant, but it's more personal.

Co-work is like you need a co-pilot

that's AI powered for your everyday

work, that's co-pilot. Love it. Well, a

lot. I'm going to stop talking. Let's

fire this thing up. I know you want to

start with a setup because you've set up

your co-work in a very specific way that

you think makes it way more powerful

than just co-work out of the box. So,

let's go through it. So, for those that

are just getting started, this is Claude

and Anthropic. I definitely recommend

that you check it out on the desktop

app. There's also a mobile app for it.

Mobile app provides chat and code, but

no co-work yet. Maybe stay till the end,

I'll show you a tip on how you can get

co-work on mobile as well. But, we have

our chat. Chat is like chat GPT, right?

We all know what that is. You're talking

with AI, that was so 2022. Now, we're in

co-work. So, I can pull up co-work here

and I can see a lot of different things

here, but generally you can see what are

we going to knock off your list, okay?

Before I get into that, I just want to

quickly go over to code. Code is like

actually coding stuff, but for here

we're going to stay in co-work. Now,

what I always recommend is you get

started with co-work by going into your

settings and going into co-work here,

and you want to set your global

instructions here, all right? Your

global instructions are who you are,

uh how you're going to process this work

this workspace here, uh the people

related to me, all that kind of stuff.

Like everything that you want Claude to

know about you and every single prompt

goes here.

I also include things that Claude should

not do in my global instructions. Claude

should not delete anything without my

approval. Claude should not work in a

different directory without my approval.

All that kind of stuff because the

global instructions are like your Claude

MD, and that means that every single

prompt your global instructions are

included there. So, it's very, very

helpful. So, always start with global

instructions here.

Next, we can go over to connectors.

These are very simple connectors that

Anthropic has built for you, allowing

you to connect your key tools. So, as

you can see here, I've already connected

my Gmail, my Google Calendar, my Notion,

my Slack. All right? It's as easy as

just signing in with that platform. Once

I connect these connectors, that means I

can go back to Cohere Work here

and I can say, "Hey, you know, tell me,

you know, what's on my calendar for the

day?" And it'll find my calendar and it

will, you know, help me do it in that

way. And it's not just reading, but it's

also doing things for me. So, I can say,

"Hey, I I want to do a event a meeting

with Alex on Friday. Check our

calendars, find a time that works, send

him an invite, you know, an email him,

and let him know." Cohere Work can do

all of that. So, again, it's reading our

data and then it's helping to use AI in

a helpful way to like do things on our

behalf. All right? Any questions so far,

Alex? No, I think that all makes sense

and I think that clarification of like

understanding all of your context, but

then taking action based on it is a huge

unlock with Cohere Work because I feel

like even if we talk about chat,

chat, I believe, like if you have

connectors set up, you could

hypothetically chat to ask what like

your, you know, what does your calendar

look like for tomorrow, but if you

actually want to take actions on these

things, I feel like Cohere Work is like

I almost view it as like

chat is if I had like a

a um

executive assistant who I just had like

create reports for me or like feed

information to me, but if I wanted like

a truly like AI-powered chief of staff

who kind of acted as like this engine

behind me for all of the work I need to

do in a given day and actually take

action on the things I need done, that's

kind of where Cohere Work thrives. Yeah.

And I would say that Cohere Work is

still new and it's still not 100% there,

but like you could see the promise and

you could see that it's worth investing

in now and you can already see the

gains, and I'm very confident that

Anthropic team has

essentially said, "This is validated.

Now, we're putting a lot more resources

there, and we're pushing forward with

it." I guess someone at Anthropic said

that Cowork was built in 10 days by like

two engineers with co- you know, Claude

code, you know, to like validate it.

It's been validated. Now, hundreds of

millions of people are using it, and I

think we're going to see even more

improvements in the future. Got it. So,

in terms of setup, first thing you said

was global instructions. My

understanding is like global

instructions are what are all the

context and ways of work So, the context

about you so that Cowork knows who you

are, knows what you do, know what knows

what your expertise is, knows how you

spend your time. Then, kind of like the

ways in which you like being worked with

so it engages you in the right way, and

then things it shouldn't do. And so,

then anytime you're doing a net new task

in Cowork, global instructions are

getting pulled in. So, that was the

first thing you did. Yeah.

>> Second thing was connections and

or connectors, and the way my guess is

you think about connectors is like what

are the core applications that you're

using on a daily or weekly basis to do

your work, and make sure those are

connected

to like or those connections are set up,

and

we don't have to talk about this now,

but a question that comes to mind is

like what if there are applications that

Cowork doesn't have an official

connector for? Is there any way to work

around that?

Um

So, those are the two things. What else

from a setup perspective is important to

you? And it sounds like you're also work

at you're going through custom

connectors right now. Yeah, so to your

answer your question, if they don't have

an official integration yet, but there

is an MCP server that's already built,

then you can connect to that MCPC server

with this custom connector. MCP servers

are kind of like API connectors kind of

all rolled up, a little bit technical.

We don't need to get into it now, but

that is something that, you know, our

team at 10x does for our clients and

whatnot, and and and it's very helpful.

So, definitely a lot of ability to

expand onto the Cowork setup as well.

Um and so I'm going to get back into a

new task here. And so this is a very

important thing that I need to show you.

So we're choosing which folder we want

to work in. And so co-work it has access

to your computer, right? And so you

could literally tell co-work go organize

the files on my computer or go into this

folder and

um organize the receipts for a business

trip and label them etc. And then open

up Excel, you know, and it's going to do

all this stuff. Like it's actually doing

stuff for you on your computer. But all

that also means that you need to show it

where to work in your computer. And

that's where we're going to take a step

back and just talk about our

organization of like the workspace that

we're creating for co-work here and how

we think about doing that. So

I'm now going to pull up my finder

window here. And so within my project,

within my Mac

I have all of my projects are within my

project folder. Now this is arbitrary.

It could be whatever is best for you.

For me, I put projects and then within

projects I have all this these different

projects, okay? These are actual like

coded projects that I'm working on in

one way or the other. And within that

project we start to have a file

structure here. Now the file structure

is very important because it is how

Claude can remember how to work with you

in an ongoing way. And this is where

chat fails miserably. Because chat

doesn't have access to ongoing

uh documents on your preferences, your

brand voice, your content. It doesn't

have any of that. And so whenever you

start a new chat, you lose all that

context of the previous thread. And so

you see a lot of people in like ChatGPT

or whatever it says like this is my

thread for writing emails and this is my

thread for writing newsletters. And like

if you ever create a new new thread,

it's like oh I got to start all over.

Not with co-work, not with the proper

setup here that we're going to show you

in this video, all right? But it does

take a second to just get started here.

And I'm going to put some resources

below that you can just copy and paste

so you can get up to speed faster.

However I want to show you first is my

brain. So this is like the first part of

it where I'm going into detail about who

I am, the channels that I run,

uh what I care about, how I think about

my work, my credibility, my positioning,

etc. All right? And then I get into my

voice, my strategy, even my email style

guide, the team members that I work

with. So, if I pull in Slack

communications or whatever, it knows who

those people are, who their roles are,

maybe if I got to communicate

differently, you know, if I'm talking to

our CEO, I got to communicate one way,

you know, that kind of thing. So, uh

working preferences, all that kind of

stuff. And so, Claude knows this every

time I start a new thread with it

because it's now working in this

workspace. And I don't need to tell

Claude any of this ever again. So,

that's very important. It's like adding

memory to your Claude, right? When you

go to onboard a new employee, you got to

get them up to speed. This is how you

get Claude up to speed. But then, once

they're up to speed, it's like, all

right, Claude, we have projects that we

got to do together, right? Uh and we're

going to work together in a various

different ways. And so, some of the

examples that I've been doing a lot of

or showing a lot of is how I do my

content creation, because I do a lot of

content, uh and my content has been

working quite well recently. It's really

exciting, uh but this is helpful and

it's helped me kind of take that next

step. And so, within the content

creation folder, I go into each of the

mediums that I write for. So, I have

LinkedIn posts, I write for a

newsletter, I have my X, and my YouTube.

All right? And now, within each of

these, think of it as a separate

employee that I have trained on a very

specific process for that. So, when I go

into X here, for example, I have a

folder with analytics of how my X

account is performing, including my

recent replies, and the performance of

each of those replies or posts, etc. So,

it took the analytics, and X gives you

this really nice export output of like,

here's all your posts and here's the

analytics. And I started with saying,

"Hey, Claude, take this analytics and

use it to help build me an X skill that

writes for me with within my style, my

voice, all that kind of stuff that could

help me write for X using, you know,

what has worked for me in the past." And

so, that was kind of the baseline of

that. And just quickly clarify again for

folks what a skill is exactly. Yeah. So,

within each of these folders, you'll see

that I have a skill outlines here. It's

a

We're so lucky because it really has

been so simplified today. Pre-

Previously, you literally had code, all

this kind of stuff, but today, the way

LLMs are working is a skill is

essentially an empty a document file

that goes into depth about a specific

process that you want to do in a very

consistent way. So, right now I have my

YouTube skill popped up, and you can see

this is all text. And anytime I can go

in here and I can actually edit this.

But, what it's doing is it's telling the

LLM how I want my YouTube outlines

built, the format, the hook, the

structure, um everything that it should

do. And so, that way when I say, "Hey,

write a YouTube outline," it goes, it

finds the skill in this folder, it gives

me an output that's exactly what I'm

asking for here. That allows me to get

consistent results in every kind of

folder that I'm working in, whether it's

YouTube, X, LinkedIn, whatever I have

set up there.

Because I've told Claude where all this

stuff is and how to navigate this

workshop workspace, it knows how to get

in there and access all the skills that

it needs to do to in do better work.

Awesome. So, what what I understand so

far about your setup is basically you

have this folder structure. In this

folder structure, you have different

things that are specific to your work

that you just want Co-work to reference

each time you're doing work in Co-work.

So, it's like the way in which you write

LinkedIn posts, the way in which you

write newsletter, the way in which you

write YouTube scripts, etc. Is there

anything not content-related that you

provide in your setup that will help you

with some of the things we're talking

about later, like you're going to be

going through how to have Co-work

operate like your executive coach or how

to build a board of advisors within

Co-work. What other important parts of

this setup structure do you need to

provide context in so that it works well

for some of these later use cases, if

anything? Yeah, I think the big thing

here is every time you want to do like a

new thing, a new task, you should create

almost like a new folder and a skill

within that folder specifying how you

want that task to be done. And I think

the biggest unlock here that a lot of

people are not doing yet is when they go

into their folders, they will maybe

create like a skill, but they don't

actually show Claude what success looks

like, right? And so in this case, I have

some tweets here from Alex and some

other people of like, this is what a

successful post looks like to me. And by

showing those real examples to Claude,

it then knows what is defined as success

to you and it can get very close to

that, right? But if you don't provide

that successful post within that folder,

then it's going to be guessing at what

success is and you're going to be kind

of like chasing it until you finally

figure it out. So, it's definitely

suggested to add your skill, add any

kind of related analytics, content,

brand guidelines, and then what success

looks like in this way.

Awesome. So, Co-work setup includes

global instructions, it includes

connectors, and then it includes this

kind of like folder structure that you

come back to when you're working on

tasks. Anything else before you want to

go into specific use cases?

Um I think that's it and I think we're

ready to start spinning up our first

thing here. Uh so let me just pull it up

on my screen. Cool. And as you spin it

up, what I'll say is I know JJ covered a

lot there. Like he said, he created a

guide for this and he has a guide. I

don't know if the guide you're referring

to, JJ, also includes the whole setup,

but um

there's a prompt that he has that you

can just end up feeding to Cogram that

will take you through the process of

doing the instructions that he just

shared with connectors, global

instructions, and the folder system. So,

don't worry about having to memorize

kind of what this whole setup is. Yeah,

I'm going to be sharing both of those

guides. So, I have it on screen here.

This is my 101 guide to getting started.

It includes everything and all my best

tips. And then I also prepared a guide

that we're going to be working on today.

It's called Cogram for CEOs. And what

we're going to be doing in the next 20

minutes or so is setting up a new folder

with new instructions that will allow us

to act as our CEO coach, our advisory

panel, and then like a thought leader

pipeline that we can use for content

creation. All right, so each of those

taking the persona of, "Hey, I'm a CEO.

I need some specific skills within

Cogram to help me do my job better."

That's what we're going to set up in the

next, uh, I don't know, 20 minutes or

so. So, I'm good.

>> Love it. And yeah, and one thing I'll

just say is like obviously, uh, the use

case here is like CEO coach, advisory

board, etc. But, while this is super

valuable for CEOs and founders, and

honestly I'm planning on using this

setup, uh, after we get off of this

call, um, to me this is valuable for any

knowledge worker who's looking for kind

of exact level accountability on the

work they're doing, a panel of advisors

who can provide perspective different

given different subject matter

expertise, um, and then thought

leadership. It's like every, from my

perspective, every knowledge worker

should be creating content and building

distribution online because it's an

accelerant in their career. So, I think

this is amazing for CEOs and executives,

but I think any knowledge worker would

benefit from this setup as well. Yeah,

100%. So, the first thing I'm going to

do is I like I want to add these skills

to my repo. And so, I got to start

laying the infrastructure so we can

build a skill that allows Cogram to

reference it, uh, in in a consistent

way. And so, I'm going to stay in my

Cogram workspace here, and I'm going to

go over to, um, right here. I'm going to

create a new folder. I'm going to say

CEO skills. All right. And then within

that folder, I'm going to go in here and

I'm going to create one folder for CEO

coach. And I'm going to create another

one for uh CEO advisory board. And then

I'm going to create another one for CEO

content creation. Now, what I'm able to

do here by foldering this is that it

allows me to go into each of these

folders and create a skill for each

specific task that we want to do, right?

And so now that I have this and it's in

this Co-work workspace here, I'm going

to go back to Co-work and I'm going to

tell it which folder I want to work in.

So, I could choose the top-level

workspace, which is appropriate. I could

also go into my folders here and go back

into my projects. And then I could pick

specifically the folder here. Um and I

think there's good ways to do it, pros

and cons. I think if you go directly

into this folder, it's going to be

probably less tokens usage and it knows

exactly what to do. But if you stay

here, it might be able to look at your

full repo, see how you've done it

previously, and apply that learning to

helping you build out this. So, I'm

going to stay in the workspace here. So,

that allows me to then connect to that

workspace on the folder. So, Co-work is

literally working in that folder now and

it's going to help me get set up here,

okay? So, the first thing that we want

to do is we want to build an executive

coach, right? So, I'm going to just talk

to Claude here and I'm going to help

have it help me get set up here. So,

Hey Claude, I want to help uh have your

help get set up with a CEO coach. Uh

this is someone that I'm going to be

able to uh

have within my directory. So, CEO coach,

I want you to build a skill within that

that almost acts like a therapist but

offering CEO level advice uh where I

could go to and ask questions about and

it's going to help me think about my

business at large, what these uh key

decisions are, uh maybe some, you know,

decisions that, you know, might come

from all these other decisions being

made and just think at a high level. So,

help me build out this skill. Ask

questions if you have any questions

about it. and then put the put this

skill in the CEO coach folder.

So, I'm going to do that and it's it's

going to start working on it. So, Cool.

>> what Claude is going to do now is it's

going to find that folder for me. It's

going to look at my repo, all that kind

of stuff, and now it's going to start

figuring out, all right, I got to create

this skill. It's probably going to come

back with questions to ask me as well.

And just so as an example for you, but

like I think this goes for anyone, say

you have your goals. Like if if it's a

CEO coach, it's going to want to know

your goals. Um say like your goals, I

know JJ, like live in Notion right now.

Would it Would it know your goals live

in Notion? Would it end up asking you

that? Do you have to point it to where

your goals live?

Yeah, no, that's a great question. You

definitely would want to point it there

because if I had that connector for for

Notion, it has access to it, but there's

so much that it has to look through that

it will be able to find it much faster

if you're like, "Hey, I have this Notion

doc here. I want you to review it,

analyze it, and then help me expand or

whatever." and you point it to there,

that will just give you a better,

quicker result without then having it

just kind of search for everything cuz

it literally will search for everything

until it finds that needle in the

haystack.

>> Yeah, you end up just wasting a bunch of

tokens the less that you're actually

prescriptive about things. Yeah,

exactly. So, it's asking, "What's your

primary role when using this coach?" So,

I think we could do it in a couple of

different ways. I'm going to say both

equally.

How should we the coach set up in style

style in terms of style?

I'm going to say

I want it to be a direct advisor.

Should the coach have contacts about

your specific business?

Uh

bake in my contacts there. And then

should the coach proactively surface

surface or consequences, etc.

Always proactive there. And then I have

that, and then I also have another

prompt that I want to put in

that I have provided you with in the

take home as well. So,

I'm going to say act as my CEO coach,

here's some more guidance

as well. And now on the right-hand side,

you can see Cohere is starting to work

here. So, it is act it's looking at my

instructions. It's looking at me, my

brand voice. It's learning more about my

company, which was a big benefit about

putting this in the general workspace

rather than just going to the specific

one. Remember we had that discussion of

like where do we go? By going top level,

it's able to access that top level about

me and get all that information about

the company I work for and all that kind

of stuff to now bake it into this skill

that we're now creating.

So, we have our progress on the

right-hand side here. We see that we're

now writing the CEO skill. Once we have

that, we're going to be able to go back

and forth with it. That's the beautiful

thing about writing a skill. Once you

have a skill, it's a very like linear

format of like here's what we need to do

X, Y, and Z. And as it runs through that

skill and asks you and you need to

modify it or like workshop it, it will

go go back, modify it until you get to

that consistent result that you want. Uh

and and that'll be really cool. While

that While we're waiting for

>> for that, one thing one thing I was just

going to say is going back to the

definition of a skill, like JJ said,

it's a markdown file. Another way I

think about it is it's just a long

prompt. Like like if you are in

uh Claude chat or Cohere or if you're

using ChatGPT or Gemini and you go into

chat and you tell the LLM to do

something, that is a prompt. You're

prompting it to do something. A skill is

nothing more than a longer kind of

standardized prompt that you think

you're going to want to use on a

recurring basis, so you want it saved

somewhere so you don't have to end up

typing out this long specific prompt

every time to get a similar type of

output. Yeah. Well well said. Well said.

I also want to show you while we're

waiting for this. If I do a backslash

here, I will have access to certain

things, and so I have different commands

here. I have different connectors here.

I have different plugins here. And while

we're waiting for this to finish, I can

go into manage plugins, and it's a new

thing that Claude has introduced in the

co-work. And so you can see plugins that

are pre-built for you. I personally use

a productivity plugin. I think it's

really good. And it helps you just put

together your to-do list and and think

through your day. It also comes with a

memory plugin built in that I think is

helpful for remembering additional

things. But you can build other things

like, hey, a customer support one, an

engineering one, etc. And the really

cool thing that they've added is you can

build your own custom plugins that you

add here, and then allow to share with

your team members in your own workspace.

So, for example, say I'm on the GTM

engineering team, I put together a

really cool skill for GTM workflows. I

can upload the plugin here, share it

with my GTM pod, and they'd all be able

to access it, but no one else on the

internet would be able to access it. So,

it's a safe way to create custom plugins

and to share custom plugins in this

manner here.

And just one question here is

go back to

chat for a sec where you did the slash

commands. I just want people to quickly

understand, like do a slash command.

Yeah. Okay, so what are the for people

who are looking at this and there's like

a lot going on. There's plugins, there's

commands, there's connectors, and then

we were just talking about skills. Can

you just give quick like quick hit

definitions of how these things are all

different from each other? Yeah, of

course. So, I'm going to start with

skills.

Skills are what we're creating now, and

it's like there's one thing that it

needs to be really good at. Right now

we're building the best CEO coach in the

world. However, at 10X we also have the

best AI newsletter in the world as well.

And I have a plugin that I use to help

me create that newsletter. And so, the

difference is with a skill you're doing

one thing, but with a plugin, I'm able

to have it do a series of things in the

order that I want to kind of essentially

build a workflow of the best way to

process this. So, for example, with our

Archer Think newsletter, I say, you

know, {slash} plugin, I do Ultra Think,

and then it says, "Okay, great. Let me

go do research on the internet." While

it's doing research on the internet for

latest AI news, it's also interviewing

me and asking me questions about what I

want to talk about, things I want to

point out, all that kind of stuff. And

then once it's done series one and two,

then it will write the the newsletter as

a third step. And that's a plugin. It's

taking three steps, combining it

together in a workflow, uh and you're

able to publish it in that way. Whereas

a skill, it's just very much a CEO

coach, and it's going to be a really

good CEO coach, but it doesn't have

multi-steps in that action yet.

Got it. Super helpful.

Um so, we are in uh Co-work is now doing

this live preview on the right-hand

side, new changes every day, love to see

it. So, we're seeing our CEO coach with

our skill here, uh and some and the way

that it has like kind of done it so far

for us, right? Um and so, now what we'd

like to do is kind of go back and forth

of like

seeing how it works, right? Um and I

don't know if there's any kind of

uh thing that we should feed it, Alex,

to see like if the feedback it gives us,

but generally there's always like a

workshop cycle of like this skill is now

live, now we want to feed it a real

scenario, see how it responds, make sure

it's helpful, if not, we tweak it until

it is. I I I have one idea, which is

something you and I have been talking

about is should we increase the

frequency of our newsletter Ultra Think

to two days a week?

I'd be very curious for it to help you

think through not only does it make

sense for us to do that, but what are

the trade-offs that you would make if

you had to spend the time because you're

currently writing our newsletter, if you

had to spend the time doing this, what

are the trade-offs you'd be making of

other things you couldn't do?

Yep.

That's great.

All right. So, uh my typing is now

terrible because I use voice to text for

everything, but there it is. We've added

in there. uh uh What do you What do you

use for voice to text? Uh Whisper, of

course.

>> Cool. Yep. Um so, let me read the skill

first, get the context I need before I

give you my full take. I need one piece

of context. When you say Ultra Think, is

it a newsletter? Um so, it's asking for

more information before it, which is

good. It's It needs that before it can

make a proper decision.

Uh so, it says that says I have enough

information. The question isn't really

should you go to exit The question is

what is a new newsletter job in the

flywheel and the frequency the

bottleneck or

and is frequency the bottleneck or is it

something else? My direct take, probably

not yet. Uh probably not for a while,

and here's why. The tradeoffs, the

sequence order cons consequences, the

risks. So,

listen, we did one prompt. We built this

skill. I think we're already seeing like

actionable like responses that are very

helpful and better than what we would

see in chat, and this is again after

just one skill. We could then go back

and forth with this and really workshop

it. And then it would at any time we

could just say, "Hey, a team member

comes to me." or say there's a Slack

conversation that's happening. You don't

really know how to do it. You could

literally direct Claude to that Slack

conversation,

um you know, with your connectors here

and say, "Hey, review this conversation.

How should I respond to it? What is the

leadership position here?" All that kind

of stuff. AI can start helping you with

that and think about it at a high level

with everything in mind rather than a

quick emotional response. So, it it can

help you in every aspect, and it we just

are 10 minutes into this. Yeah, yeah.

Just a question there is like very

tactically, if you wanted to ask it for

its take on a Slack conversation, how

would you functionally do that in

Cohere?

Yeah, I think again, going back to the

Notion example, I would actually share

the the Slack conversation with them

like the URL right there, and because it

is connected via the connector, it would

have access to do that. It'd be able to

read it all, and then it would be able

to uh you know, give you advice on

what's going on there.

>> Cool.

Awesome. Um, and then

I It's interesting. I don't know if

you've seen this. This is the first time

I've seen this like eval review on the

right side. Um, so

how is this supposed to work here? Like

how Do you Do you have any sense of like

basically I guess it's showing you a

scenario of how of how this skill could

be used and then you get feedback and

I'm assuming it would then just

automatically change the skill based on

the feedback you give. Exactly. They

created more of an interface to help you

process this because before you kind of

needed to know what the prompt and now

it's a little bit more clear of how we

can improve it together. That's awesome.

Cool.

Love it. Anything else on CEO coach

before we move to advisory panel?

>> No, no. I think that's it for now. I

think you now know the general gist of

how we get there and and you could

workshop this. It's very helpful. Let me

show you repeatable motion again and

this is going to be a sub agent advisory

board and honestly it's like low-key one

of my favorite things. Love it. So I'm

going to go back here. I'm going to say

all right. So we have our CEO coach here

and now look at that. Within that we

have our skill that we've already added

there. Fantastic. So I'm I'm really

excited about that. One thing though is

look at it didn't put it in the right

folder here. So I could just literally

take this, copy this, bring it in here

and paste it and that would be

sufficient. So like we're not This is

not rocket science. So like we know how

to organize folders and whatnot. It's

just we got to take the time and learn

how

LLMs are using the context of our

systems to make better outputs. Once we

got that we're like we're ready to

cruise here. So the next thing that we

want to do is build a skill for an

advisory board, okay?

So we already got that folder in place.

I'm going to go back to Claude CoWork

here. Now I'm actually going to create a

new thread. I'm going to get out of here

and I'm going to create a new task. And

the reason being is this this thread

right here is very context heavy in

creating a skill for an CEO coach. I

don't want that yet. I want a new

context thread started of like a new

skill for an advisory board. There's no

reason to mix that. So, if think about

threads as like for each thread it's

like your to-do list. It's one item on

your task. And I think the better that

you can manage that, there the less

tokens you'll spend and the quicker

you'll be able to get to the point

you're trying to do because it's it's

feeding through less content and it's

straight to the point. Make sense?

Make sense. Cool. All right, so I have a

another prompt that I'm going to copy

and paste here. Again, I'll give you

this in the guide, but

next up we want to create a new skill

for a CEO advisory board. And this

advisory board is going to be using a

series of sub agents that we have to

define a persona for to help us evaluate

decisions based on our ICP and our

customer persona. Here's a generic

prompt that I'm going to help for you

structure this skill, but go off, create

the skill, ask me any questions you

have, and then place it within the

folder for advisory board. Thank you.

You are just so damn polite to your

LLMs.

They will come back one day and

hopefully they will love me back.

And and we can do this when you finish

the advisory board, but one thing I'm

thinking about one thing about two

things. One is I think like as it is

these specific skills are super valuable

for viewers and they should set them up

themselves, but I think how you are

going about setting up skills like that

just the general flow is as important

because everyone has their specific ways

of working and just knowing okay, in

order to create a skill, I need the

right folder structure on my computer. I

need to tell

Co-work that I want to create the skill.

I need to give it some context and I

want it to ask me questions. Like that

flow and then once the skill is done,

let me actually test it out and let me

provide feedback. By providing feedback

the skill is going to get updated by

Co-work which is going to make it better

each time. Like understanding that flow

is so important. That's how you get

leverage of just like not just having a

single skill, but just having this army

of skills. The other thing that I would

love for you to share, whether it's with

the advisory board or the CEO coaches,

let's just say you you make this. You

come back a week later, and you're like,

"I want to invoke the CEO skill now.

Like, I'm making a specific decision,

and I want to use that skill." Like, are

skills always on, or do you have to call

them at like when you need them? How

does that work? Yeah, that's a that's a

really good question. Thanks for that

callout. So, you can build the the way

that the skill invoked into the skill

itself, and depending on how sensitive

it is, you can also reference that at a

high level. So, for an example, it's

going to know out of the box how to

invoke it, but there are situations

where it might not be invoked if you

don't tell it to. So, you can have uh in

the skill saying, "Invoke this skill

when I ask you about X, or I ask you

about Y." Those are clear indicators to

the LM that it needs to be invoked.

Cool. But, sometimes the skill are like

subfolders in it doesn't see it. So,

that's when you go back to your main uh

folder here, and you look at the

workspace map. This is very important

for Claude to understand what's going on

in your workspace, and it and the map

and how it should process everything.

And so, every workspace should have that

top-level map there that maps out

everything, and it needs to stay

updated, otherwise it won't be able to

find it as easy. But, this right now

will then show, "Okay, in X, here's my

clear skills, analytics, posts, etc." A

very clear structure of how Claude is

going to navigate this, so we can go

straight to what it needs um as well.

And then in this top nav as well, like

the workspace map, you could also say,

"Hey, how should you use this?" And then

you should can also add in there, you

know, specific skills that you want to

invoke, how you want to invoke it, kind

of like global instructions. This

top-level map is like a second layer of

like adding sensitivity to something

that you really want to be included. Two

questions here. One, why does workspace

map not get included in global

instructions? What's the reason you keep

it at this level? Um and then the second

is, did you have to build the workspace

map? Cuz I'm sure someone watching this

right now is like, "Holy like this

is a lot. Um

uh do I need to build the map? And also,

as I'm adding new folders and new

skills, do I have to update the map?

Like, what what is my responsibility

here?" Yeah, I know, that's a great

question. Global instructions, as you

saw from the input, it's like very

simple text, right? You can't do a ton

in there. And so, that is like top

level, I need you to know this, but I'm

not going to go like super deep. Though

then the workspace workspace map is

where I go super deep on the hierarchy

of everything. And yes, I do tell Claude

to make it, but it's as simple as

something along the lines of going into

a new thread and saying, "Update my

workspace map for the folder with

latest, right?" And then it will do

that, and then it will update that file

for me. And so, this could be um like a

scheduled task, for example, like I

could go in here and create a new

scheduled task to run every single day

to have that do that for me. Uh we

haven't talked about scheduled tasks

yet, but it's a really cool way that you

can set a frequency of co-work to

actually do stuff for you. Uh and one

example of this is every morning at 7:00

a.m. I have it run a check on my Slack,

my email, my calendar, uh my Notion to

see if there's anything that urgent that

needs my attention. And it kind of gives

me a daily uh uh

meeting a daily prep for like when I

walk into the office that is already,

you know, ready for me to see, and all

the high priority things are already

there. So, that's a daily briefing, and

there's a lot of different ways that you

can schedule these different tasks.

Love it.

So, we are here and uh working on the

CEO advisory board. It's building this

skill. Uh it found uh sub agents. Uh

well, in actual in the uh

the prompt that we gave it, we asked it

to do

uh five different sub agents here. So,

the operator,

the skeptic, the customer advocate, the

finance partner, the risk legal. So,

these are sub agents. Now, for those

that don't know what a sub agent is,

essentially, right now we're talking to

an agent. And normally, it has one agent

doing stuff for us. But, the whole

concept of this advisory board, which

again, I love and you can repurpose any

way that's best for you and your org, is

Claude is able to spin up sub agents.

So, those are agents that are working

under your main agent, and they kind of

have like a fresh context window, and

they're able to take on your persona

that you define it. So, for this

example, we would spin up a sub agent

that has a persona that's defined as the

operator. Now, we can go like niche on

this and really define what each of

those personas are. This prompt was a

little high-level, and it just said,

"Hey, take on a normal operator

persona." But, we could go in there, and

I do for my content, where I have, you

know, my ICP personas, and I use an

advisory board for my ICP personas to

analyze the content before I even

release it, so I can get their feedback

and proactively make those updates

before I even send it there with having

these sub agents. So, it's a ability

it's a way for you to send sub agents

with whatever persona you want to work

domains, uh work spaces, notion docs, or

whatever, to analyze something from

different various perspectives to bring

all the feedback back to your main

agent, and then your main agent will

sort through it, figure out what's worth

keeping, showing you all the feedback,

and then you can kind of workshop it.

So, it's a very clear feedback loop that

you can create using co-work and Claude

code and as well and as matter of fact.

Love it.

Um I also want to point out while this

is this is working here, uh the runs are

done, it's doing any val here, but this

folder structure that we have put

together, not there, it is Let me see.

Let me get back to it. This folder

structure that we put together here,

this is very universal. You can bring

this with you wherever you go. So, if I

wanted to go into Claude code, I'd say,

"Hey, you know what, Claude code? Go

into that folder structure and use that

same folder structure and work out of

it, right? If I wanted to go into Codex,

I could do the same thing and go into

Codex and find that folder structure and

work out of it. So, there's no matter

what you're building here, these AI

coding tools are kind of reading from

the same thing and you could take a lot

of what you build here and take it with

you as you go between it all. It's

really just a folder of all your

information, your memory, your contacts,

the operating instructions that you give

to that AI leader that it that is then

able to process it better for you. So,

this isn't just Cohort, it's everything.

Love it. One thing I'll just add, if say

you end up implementing the CEO coach,

the advisory board, the thought

leadership, um,

generator, which is the third, uh, skill

we're going to go through. But say

you're like, I don't know what other

skills I want to create in my role. One

thing I just want to quickly share while

this is still running is, so we actually

put this in, uh,

the most recent, uh, 10X newsletter

called Ultra Think. Basically, uh, what

it is is someone posted, "Run this

prompt frequently, you're welcome." And

what the prompt is is it says, "Scrape

all of my Claude sessions on this

computer. Give me a breakdown of all the

things I do, things that are worth

making it to skills versus plugins

versus agents versus claude.md." And so,

this is a great way where

you can just have an LLM identify

patterns of work that you do frequently

that can likely be turned into a skill

to support you in doing that work. So,

we'll we'll uh, share this in the show

notes, but I think this is a great

prompt. If you don't have a skill top of

mind that you want to build, use this

prompt to start getting the ideas going.

Yeah, uh, that's a great call out. Um,

and I think there's a lot of

self-learning, uh, to be had there. And

and you can also, um, use Cohort for

that. You can have it analyze your

workspace. You can have it optimize,

organize your workspace, like Totally.

Cohort can do a lot of this. In in guide

that I'm going to share with you

afterwards, um, right here. The main

prompt right here is this copy and paste

prompt. It's literally going to Once you

copy and paste it, it's going to

interview you and build out this folder

structure for you on your behalf. So,

you don't even need to build out the

folder structure, you just need to know

you need to do it and you need to be

spend the time to really kind of like

fine-tune it. But, once you have it

again, you can use it in so many

different ways. Wow.

This is still running, but let me show

you something that not a lot of people

kind of think about right away. You can,

of course, run multiple agents at a

time. So, while this is working, I'm

going to get a head start on our next

one. So, we're here working on the

advisory board skill, right? Now, we're

getting some strong results. It's

self-evaluating

itself and it's starting to do a lot

more self-testing. It looks like it's

almost there. But, something that a lot

of people don't really think about or

realize is you can run multiple agents

at a time. So, I have this one working

for me creating an advisory board right

now. But, I want to do something else as

well. I want to start getting to work on

our final one, which is a CEO content

pipeline, right? And so, again, I would

go back into my folder structure here.

I'd go back into my CEO skills here and

I'd see that my CEO content creation is

empty and that's where I want to work

out of. And quickly, I just want to

check my advisory board. I see that

there are stuff being done here. Cool.

So, I'm going to stay in this folder

here.

And the next thing I'm going to do is

I'm going to copy this next prompt in

and I'm going to say,

"Hey Claude, help me create this thought

leader pipeline and this is going to

help me create content that I want to

post on various platforms uh for my

thought leader style piece. This needs

to reflect my business, my learnings, my

history as a leader. Uh Think of it as a

executive thought leadership and I

really need you to help me fine-tune

this tone and create a style that and

skill that helps me write in this

manner." And now, I'm also pasting the

prompt that I am putting in here as

well. Again, I'm going to share this

with you, but in in more detail, it just

goes through the areas of content

pipeline that we want to think of

expertise audience platforms create like

an idea capture system and I topic

scoring

workflow

distribution cadence etc. It's really

going to help you put together a content

pipeline from that executive thought

leadership angle.

So that's going to get the work there.

This agent is now running and you see

this blue icon over here. That means it

is done and it's ready for us to review.

And so now back over here on our other

one, we see that this has done some cool

stuff for us. So the skill is built and

saved to CEO coach CEO advisory board.

The skill file is ready to install the

valve review HTML has side by side

outputs. So what the skill does? You

paste in a launch plan assumptions and

constraints. Five board members run

independent evals each with defined

missions. They ask questions. They ask

and then they score it and then they

give you a verdict. And so this is all

happening. Claude went in through and

did some self-evaluation

and now I'm able to invoke this skill

whenever I need to. So I could I think

could I go in here? Let's see. I might

not going to be able to invoke a skill

that way, but what I could do then is to

test it out is just some simply ask it

to

maybe a similar prompt to what we had

before of

Hey Claude, I want you to test this

skill out. We're thinking about turning

Ultra Think newsletter which is

currently just running once a week into

a multi-week newsletter post. Help me

evaluate the decisions of this use my

ICP personas to help me figure out if

this is something that they'd want. Do

some market research here and just come

back with my advisory board letting me

know if this is the right decision or

not. So I put it in there. Maybe the

scenario isn't best for an advisory

board. Let me give you another example

of how I use this scenario.

I recently had a created a

self-published a Slack case study. And

so when I was self-publishing that class

case study, I had the written output and

I wanted to have an advisory board

review it from a Slack perspective, a

copywriter perspective, and just an

audience perspective. And so, I fed it

the final output, the advisory board

went through and helped me workshop it

until I felt it was ready to publish.

Uh and then I published it and and

actually Slack went on the reposted and

supported and whatnot. So, that's a

really cool real world scenario of I

wrote something, I wanted someone to

check it uh before I put it out. Maybe

my employees, my team members were busy.

This is a way that I could use Super AI

to help me with that and make sure that

that first draft that I put out is the

best draft.

Love it.

All right. So, we're over here, back

over here talking about content pipeline

that we want to do. It's asking me

questions to further build this skill.

And I think after now, you've seen that

we've gone through this a series of

time. We built three different things

and the flow is very similar for each.

We're telling it to uh build skills for

us. It's asking us questions. Uh it's

figuring out what our preferences are.

And then once the skills are built, uh

then we can go ahead and test and

workshop them. And then we'll always

have those skills that we can use uh

going forward. And so, uh we've done a

lot in this workshop. Uh we've put

together multi-agents. We set up our

global instructions. We've set up our

repo here.

Um set up our repo

uh with our project folder. It's going

to be right in here. We talked about the

importance of adding the context here.

We talked about the importance of adding

a workspace map. We talked about plugins

versus skills, adding connectors, and

just generally the best way to work with

Co-work.

I would like to say at the end that this

isn't an advertisement for Co-work. This

isn't even an advertisement for

Anthropic. This isn't an advertisement

for you to do more with AI. AI is long

beyond AI slop. AI is so helpful in

every aspect of your day. GTM, customer

support, sales, engineering, whatever

you want it, you can get really

high-quality results. I'm using even

using Claude code to help me edit videos

these days. I just got off an event with

someone that's using Claude code to do

programmatic meta ads. Like, what you

can do with AI is mind-boggling. And so,

this is an advertisement for you to stay

curious, to keep experimenting, to keep

pushing the limits.

Uh and if you ever want to bring

something like this to your org, make

sure you call us at 10X because we can

help you do it at an enterprise level.

Love it. One last question for you cuz I

know you promised this at the beginning

and I want to make sure we deliver on

our promises. You mentioned about using

co-work in mobile. How do you do that?

Oh, this is a fun one. Uh all right. So,

if I were to pull up mobile,

um which I can't do just yet, but

co-work is not available on mobile. And

this actually, thank you for doing this

cuz we needed to really cover this.

Co- So, co-work is not available on

mobile right now, right? And co-work is

currently only on my device. Say I

wanted to bring all of this to Alex. I

couldn't do it right now because

everything is just on my device here,

right? So, that's where we need to

introduce a new system to allow you to

bring everything that you do here to be

accessible via the cloud. Because right

now, this is all on my device. But, I

can create stuff for my team, I can

create stuff for my department, my pod,

etc. And they can then all access it and

download it. And there's a key flow that

you need to know. And it's a scary word

called GitHub. It doesn't need to be so

scary though. Let me just show you a

high-level way of the way that this

would work. So, first thing that we want

to do is we want to create a GitHub

account. A GitHub account is a free

account and you can use it in a variety

of different ways. Now, why do we want

to create a

a GitHub account? Because you're going

to need to create a new repo. A new repo

is essentially a folder for your files.

Think of it like this. GitHub is the

cloud. It allows you to take the files

from your device, upload it to the

cloud. Then, I could share this folder

with Alex and you could download all of

those folders from the cloud into his

device and access it, right? So, that's

very important. So, right now what we're

going to do is we're going to say JJ

co-work setup. And I'm just going to say

I want this private. If you don't make

this private, it will be public on the

internet. So, make it private. We're

going to create this repo here.

Right? It's going to create it and it's

going to give us this URL. Now, we can

copy this URL,

go back into Claude here, and then we

would say something along the lines of

publish this folder to this repo, and

then paste that in. And then, what will

happen is it will look at the folder

that you're in, which is essentially our

co-work folder, and it will take all of

the folder the content in this co-work

folder and upload it to that GitHub repo

right here. Then, I could share this

with Alex, and he can download it. Or, I

can share it anyone on my team, they can

download and get access to it. But, the

real unlock here is that once you do

that, then you can go into Claude Code

on your mobile phone, and you can access

that GitHub repo, and you'd be able to

access all of the files that were on

your device now on your phone. And this

is very helpful because, like, I'm doing

a lot of content creation on my phone,

and so I have my co-work setup all on my

phone. So, I open up Claude Code on my

phone, it accesses my co-work setup with

all my X writing, my LinkedIn writing,

all that kind of stuff. I ask it to help

me write something, it will do it, and

it will leverage all of the skills in

the workspace that I've set up for

co-work.

I will say one thing here is that that

is a cloud

um

structure. So, for example, if I were

then to make updates to information on

my local here, I would then need to say,

"Hey, push those updates to the GitHub."

Or, if I wanted to take this and put it

on another computer that I have, then

I'd say, "All right, I go on the other

computer, I'd access this GitHub, I'd

download it to that computer." And then,

if I made updates to that other

computer, I would then need to push it

back to the cloud and then download it

again here. So, it is that local to

cloud download. We're very used to this

with, like, uploading files to Google

Drive and then downloading those files.

It's the same thing but in a code format

here,

um and it works really well. So, you can

keep it locally, you can put it in the

cloud, you can use it on your desktop,

you can use it on your mobile. The most

important thing is that you are using it

because this is an incredible time to be

doing anything with AI. You're actually

getting good results. Doesn't matter if

you're an engineering, customer support,

or whatever, you need to be doing more

with AI. It's a truly amazing time for

it, and hopefully this video just helps

you get started as a non-developer. This

was awesome. JJ covered so much ground

here. You're the man, 10X's resident AI

builder. Uh we're definitely going to

have to get you back on uh the show over

the next uh month or two, but I think

this was such a good breakdown of how to

get CoWork set up in a way that is

custom and really high-powered and

high-performing for kind of any

knowledge worker, especially executives,

and then the three use cases with the

CEO coach, the uh advisory board, and

then the thought leadership uh pipeline

for helping you create content online,

and then finally how you can do uh

CoWork skills but on your phone. Um you

just covered so much ground here in the

last 52 minutes, so I appreciate the

time, uh and yeah, we got to get you

back on soon. Hey, it's a pleasure to be

here. Um it's really exciting. Thank you

for that. I hope this all helps you a

lot. Uh put comments if you have any

questions, uh DM us. Me and Alex are

always available doing our best to help

ever- everyone out as we can. Uh but

this is a really exciting. I hope you

find it just as exciting, and thanks for

watching.

>> Thanks, everyone.