0:02 AI agents on cloud code, co-work, and
0:04 codecs can now automate workflows, write
0:06 code, and run entire business processes
0:08 autonomously. But no matter how good
0:09 they get, they'll only get as good as
0:11 the context you provide them. And the
0:12 people and the businesses that will get
0:14 the most out of these tools are the ones
0:16 with the best context infrastructures.
0:18 So in this video, I'll explain all seven
0:20 levels of using context in cloud, from
0:22 chat to projects to a second brain to a
0:25 full business OS. Show you what changes
0:27 at each level, why it matters, and how
0:28 to transition. and I'll show you how I
0:31 set up the final level, a full business
0:33 agentic OS with shared context and
0:35 skills across my team with permission
0:37 settings. Now, before explaining the
0:38 levels, let me quickly explain why
0:40 understanding this context
0:42 infrastructure is key. As said, we're
0:44 all seeing these AI agents becoming more
0:45 capable by the day and with fast
0:48 developments in MCPs, connectors,
0:50 browser use, they're starting to become
0:51 really good at using our softwares and
0:53 the internet. and with skills, plugins,
0:56 sub agents, uh, schedule tasks, and
0:57 other developments, they're getting the
0:59 capabilities to actually start executing
1:01 work for us more and more autonomously.
1:03 But in order for all of this to actually
1:06 be useful for us or our business, agents
1:08 need lots of context um, to actually
1:10 know how to do work for us. And that's
1:13 why context is the fundamental layer to
1:14 get right in the upcoming months. I can
1:16 tell you from my experience that having
1:18 this context layer set up well has made
1:20 a huge impact on how I and my business
1:22 work together with AI and is allowing us
1:24 to make co-work and cloud code become
1:26 more and more of our primary operating
1:28 system. But it is a topic that is not
1:30 straightforward and requires some effort
1:31 uh and takes some time to develop. But
1:33 the earlier you start with this the
1:35 better because the context compounds. So
1:37 if you start the day your agents will be
1:39 far more useful for you and your
1:40 business in a couple of weeks or months.
1:42 Now, I'll explain all the levels of
1:45 using context from simple chat context
1:47 all the way to an entire shared context
1:49 infrastructure across a business and
1:50 show you exactly how to level up no
1:52 matter what level you are uh at right
1:54 now. You can also jump to the level
1:55 that's most relevant to you. But even if
1:57 you're a bit more advanced, I think this
1:58 breakdown will hopefully help you wrap
2:00 your head around it a bit better. Now
2:02 level one is where unfortunately most
2:03 people in the world are still stuck
2:05 which is by providing context to a
2:08 language model in each chat manually or
2:10 even worse not providing models with any
2:12 context at all. Again no matter how good
2:14 these models get if you don't give AI
2:16 context using AI usually becomes a
2:17 frustrating experience. If I give a
2:19 prompt like write me a LinkedIn post on
2:20 why setting up a context infrastructure
2:23 is the key to making AI agents useful. I
2:25 get a very generic piece of content with
2:27 m dashes that screams AI. Now, of
2:29 course, most people know this and have
2:30 figured out that you need to provide
2:31 some context to actually get good
2:33 outputs. But what context do you give
2:35 it? This of course depends on the
2:37 specific task. But an easy way you can
2:38 think about what context you need to
2:40 provide to get better outputs on most
2:43 types of tasks are first giving context
2:44 around who you are and what your
2:47 business does. Second, context around
2:49 who this is for like your ICP, your
2:51 customer or the recipient. Third, which
2:53 is probably most important, which is
2:56 showing AI what good looks like for this
2:57 task. for example, through examples,
3:00 references or descriptions. And lastly,
3:02 defining what the roles and guardrails
3:03 are for this task. So, if I give that
3:06 same prompt, but with that context, we
3:08 instantly get a far better response. No
3:11 am dashes, formatted like a LinkedIn
3:13 post, more my tone of voice, and more
3:15 relevant to my ICP and an aligned call
3:17 to action. But of course, the limitation
3:19 here is that it's extremely inefficient,
3:20 timeconuming to copy and paste or
3:23 rewrite context in each new chat. Which
3:25 brings us to level two, which is cloud
3:27 chat projects. Again, most people have
3:29 figured this out, but projects in cloud
3:31 chat were of course developed to solve
3:33 the copying and pasting and rewriting of
3:36 context problem. In a project in cloud
3:38 chat, we can just add the context files
3:40 once together with a system prompt here.
3:42 We can also add a broader scope of
3:44 context. And each time you want to write
3:46 a new LinkedIn post or do a repetitive
3:48 task, we can just give a simple prompt
3:50 here while still getting good output.
3:52 But the downsides of these projects are
3:54 that first they live in isolated chat
3:56 windows. So we have to hop between
3:57 different projects for all our different
4:00 tasks. Second, clot can't actually
4:02 create, update or edit any of these
4:04 context files or the instructions
4:06 itself. So anytime you want to adjust
4:08 the process, update a context doc or
4:09 anything else. We need to manually
4:11 update it inside of the cloud project.
4:13 Thirdly, these chat projects are usually
4:15 limited to a specific task and work best
4:17 when I have separated projects for each
4:19 task. I'd need to set up a separate
4:21 project for YouTube packaging, YouTube
4:23 intro writing and ideation because I'll
4:25 follow a different process and have some
4:27 different context files. And lastly,
4:29 it's hard to actually test and improve
4:31 these projects without you extensively
4:33 using it and updating it manually. Which
4:35 brings us to the next level, skills,
4:37 which instantly resolve all of these
4:39 limitations and why I highly encourage
4:41 you to start using co-work or cloud code
4:43 if you're still in level two and working
4:45 in a cloud chat. Because for skills, we
4:47 need to make this transition. skills are
4:49 very comparable to these projects. We
4:51 have the skill MD this and this is
4:52 essentially the instruction or the
4:55 system prompt just like in projects
4:57 laying out the process and when to use
4:58 the different context files. For
5:00 example, here I have my LinkedIn writer
5:03 skill that includes a skill MD that lays
5:05 out the process it should follow and
5:08 when to read the context files and
5:10 inside the skill we have a references
5:12 folder with all the different context
5:14 files. But skills in contrast to
5:16 projects can be used in any chat at any
5:18 moment you want. This means for example
5:20 that mid conversation here in a YouTube
5:23 ideation chat I had I can turn a good
5:25 insight into a LinkedIn post instantly
5:27 by just telling claude write a LinkedIn
5:29 post based on the topic we discussed
5:31 using the LinkedIn writer skill and he
5:33 went ahead followed the SOP and wrote me
5:35 a LinkedIn post. Skills are also far
5:37 easier to build than projects. With
5:40 Enthropics built-in skill creator skill,
5:41 we can build them by simply telling
5:44 Claude to build us a skill. For example,
5:46 help me build an infographic skill. We
5:47 can also build them out of any
5:49 conversation you had with Claude by just
5:52 clicking here and select turn into skill
5:54 or by just telling Claude in the chat,
5:56 make me a skill based on the process we
5:57 followed in this chat. We can also
5:59 easily share these skills with our team
6:01 through zip files by just asking clot,
6:02 can you create a zip file out of the
6:04 LinkedIn writer skill which anyone in
6:06 your team can upload by going to
6:08 customize skills and clicking on the
6:11 plus icon here and create skills upload
6:13 skill. Or if you're on a team plan in
6:15 co-work specifically, you can just add
6:17 them to your organizational skills. We
6:19 can also easily adapt Enthropics
6:21 built-in skills, for example, these by
6:22 just telling CLA you want to customize
6:24 them or clicking here on edit. And we
6:26 can easily import skills from other
6:28 people and businesses by just going here
6:29 to browse plugins and then going to
6:31 entropics and partner. There are also
6:33 dozens of skill marketplaces around the
6:36 internet. And with built-in evals, we
6:37 can immediately test our skills and
6:39 improve them fast to make sure they
6:41 actually work and give us good outputs
6:42 consistently. For example, here I
6:44 created a newsletter writer skill and
6:46 then just told Claude, "Please test this
6:48 skill." The criteria for the test are,
6:50 "Is this skill functional? Is the word
6:51 count similar to my newsletter examples?
6:53 Is the sentence structure similar? and
6:55 is my tone of voice similar? It then ran
6:57 pre-ests in parallel and gives me an
6:59 eval report with a summary of the
7:01 results and suggested fixes which you
7:03 can apply immediately. You can even
7:05 autonomously let them improve themselves
7:07 through an auto research loop which I
7:09 recently did a video on which I'll make
7:10 sure to link in the description below
7:12 too if you haven't seen it yet. So, if
7:13 you haven't yet, you really want to
7:14 start building out these skills around
7:16 your repetitive tasks and processes.
7:18 We've been building out more than 60
7:19 skills across all our business
7:21 processes, which if you're interested,
7:23 you can also download and customize for
7:25 yourself if you check out my AI
7:26 accelerator in the link in the
7:28 description. And lastly, skills can also
7:29 be scheduled, which means we can now
7:31 trigger them autonomously through
7:32 Claude, which I'll show you some
7:33 examples of later in this video. Now,
7:35 skills are amazing, but skills are best
7:37 for pre-established processes of work,
7:39 and much of our day-to-day work isn't
7:41 actually a pre-established workflow. So,
7:43 this brings me to the next level, which
7:45 is using file access together with
7:47 skills. Because for many, if not most
7:49 tasks that AI can help us with, it
7:50 doesn't actually follow a
7:52 pre-established process or workflow. We
7:54 have one-off tasks. We have tasks like
7:57 ideiation, planning, strategy, or using
7:59 AI for decision-m. And for many of
8:01 these, we don't necessarily need or want
8:04 skills. But we do want AI to have more
8:06 context around you, your business, and
8:08 your goals. And this is where file
8:10 access in cloud code or cloud co-work
8:12 becomes powerful. Because with every new
8:13 chat we open in cloth coowork or cloth
8:15 code, I can now give cloth access to a
8:17 folder on my computer. For example, here
8:19 I selected a file with relevant YouTube
8:21 documents. And in that folder, I have
8:22 some documents about my YouTube
8:24 strategy, some old transcripts, a
8:27 hookbank, etc. And if I now just want to
8:29 ideulate or plan a new video together
8:31 with Claude, I'll get instantly better
8:32 outputs because it has context on my
8:34 strategy, me, my channel, and what's
8:36 important to me. For example, here I
8:38 wanted to ideulate on the video I'm
8:39 recording. And you can see it pulled
8:42 some data and context like my brand, my
8:44 ICP, my YouTube voice, and my YouTube
8:46 strategy in order to give me more
8:48 relevant ideas to plan out my video. And
8:49 because it has more context, it even
8:51 pushes back on some things. Because, for
8:53 example, now it knows most of my
8:55 audience is non-technical and I explain
8:57 something that's too technical. When we
8:58 start working with file access, we also
9:00 start working with the cloud MD, which
9:02 is basically an instruction on how to
9:03 navigate the folder, which becomes more
9:06 relevant if the context grows. But I'll
9:07 cover the cloud MD in more detail later
9:09 in this video. And if you haven't yet,
9:11 you really want to start using this file
9:12 access consistently because you'll be
9:14 surprised how much more relevant your
9:17 answers get. And it allows AI to become
9:18 much more of a strategic sparring
9:20 partner. And secondly, because it now
9:21 has access to a file on your computer,
9:23 it can't just read those files. It can
9:25 also instantly update the files. You can
9:27 save new files or assets like
9:29 presentations, Excel sheets, Google Docs
9:32 directly into the folder. So any uh
9:34 update you want to make in a context
9:36 docu, good outputs you want to save or
9:38 assets you want to save, cloud can
9:39 instantly do it. And essentially what
9:41 this means is the more you start using
9:43 file access, the more this folder will
9:45 grow with context naturally. Now, and if
9:46 you're just starting at this level, I'd
9:48 highly recommend putting in some effort
9:49 and setting up some of these important
9:51 context documents. I've added a free
9:52 resource also in the link in the
9:54 description below that's basically a
9:55 questionnaire where you can go through
9:58 and I highly recommend taking 30 minutes
9:59 with a tool like Whisper Flow where you
10:01 can talk to your computer and you just
10:03 do a brain dump and by answering all of
10:04 these questions you can then feed that
10:07 brain dump into Claude and he'll create
10:09 these structured context documents that
10:10 are important to have as an initial
10:13 start. I've also added in some example
10:14 reference files so you get an idea of
10:15 what these look like. in my AI
10:17 accelerator. We also have a full
10:18 step-by-step walkthrough on how to set
10:20 this up efficiently together with best
10:22 practices and uh unlimited one-on-one
10:24 life tech help. So, if you want some
10:25 help, you can also check out um the link
10:26 in the description below. Now, when
10:28 you're starting to use this more and
10:29 more and the context in your folder is
10:31 growing, it's natural to go into the
10:33 next level, which is using cloud co-work
10:35 projects. This can also be done through
10:37 cloud code. The same principle applies
10:39 if you use cloud code. and projects on
10:41 cloud co-work is essentially just a
10:43 better way to organize your context
10:45 across different areas of work. Now this
10:47 is different than a chat projects
10:48 because chat projects are very task
10:50 based. Co-work projects can be used on a
10:53 higher level for areas of work. For
10:55 example, I have projects here set up for
10:57 sales analytics, operations, agency
10:58 clients, community management and
11:00 YouTube. And projects are essentially
11:03 the same as file access but in this case
11:05 we just predefined the file here with
11:07 the relevant context for this area of
11:09 work. So now when I want to idate on a
11:10 new YouTube video, I can just directly
11:12 go into that YouTube project and the f
11:14 folder will already be selected. We'll
11:16 also have all our chats around this area
11:18 of work organized here below. We still
11:20 use skills for the repetitive task of
11:22 course. For example, in this chat when I
11:23 was planning the video and I got the
11:25 concept clearer, I used the YouTube
11:27 intro writer skill to give me some intro
11:29 ideas and variations according to my
11:31 framework. We can also see our scheduled
11:33 task that are relevant for this project.
11:35 For example, my YouTube ideation skill
11:38 runs every morning to give me new ideas.
11:39 But besides this better organization,
11:41 there's one more added feature to these
11:43 projects, which are instructions and
11:45 memory on the project level. And these
11:48 allow us to add in specific rules and
11:50 guardrails and specific memory for
11:52 specific areas of work. For example, my
11:54 YouTube project, I have a specific
11:56 memory that it needs to push back during
11:58 ideiation because I want to have
12:00 alternative framing and factchecking.
12:01 And you can make these memories or rules
12:03 by just telling clot in a chat that it
12:05 has to memorize this. Now when you're at
12:06 this level and you really start to use
12:09 projects skills, schedule tasks more and
12:11 more and consistently and really start
12:12 connecting it more and more with your
12:14 softwares, you'll start using AI more
12:16 and more as your operating system. And
12:18 honestly, if you use this infrastructure
12:19 well, you can already get a lot out of
12:21 AI for yourself and your business. But
12:23 when you start to use this more and more
12:25 and your context and your projects are
12:26 growing, you'll notice that even with
12:28 this project infrastructure, the growing
12:30 context will become harder to manage.
12:32 You'll have shared context files across
12:34 multiple projects, across multiple
12:37 skills, for example, common docs like an
12:39 ICP doc. And when something needs to be
12:41 updated, it needs to be updated across
12:42 all of these different projects and
12:44 folders and skills separately. So that's
12:46 where we want to start looking at the
12:48 next level, which is setting up a second
12:50 brain or a personal operating system.
12:51 Now, even when you're planning to roll
12:52 this out across a business on a
12:54 companywide level, which will be level
12:56 seven, I still highly recommend you
12:58 start with level six. Once you've set it
13:00 up and it works for yourself, then think
13:01 about level seven, where you actually
13:03 start syncing this across your team with
13:05 permission settings, etc. Now, in this
13:07 second brain setup, all we do is we just
13:09 add all of the context and centralize it
13:11 into one folder. And this becomes very
13:13 powerful when we have a lot of context
13:15 because we'll now have persistent
13:17 up-to-date context around an entire
13:20 business or life across any chat or AI
13:22 provider. We can do that by opening that
13:24 file through cloud code or just doing it
13:26 with file select in co-work or by
13:28 setting up one project connected to the
13:31 personal OS folder. But I can also do
13:33 this in codeex or any other AI provider
13:35 that allows for file access. And as I
13:36 said, we're still doing the same. We're
13:38 just adding all of that context into one
13:40 big folder. And this becomes an
13:41 advantage when you start using AI and
13:43 context around more and more areas of
13:45 work in your business or around more
13:47 departments because when you have a lot
13:49 of contacts, it's better because we now
13:50 just have one folder to structure and
13:52 organize. And this also means that
13:55 context docs don't have to be updated
13:57 across multiple projects or skills. I
13:59 can have all my projects or business
14:01 departments inside of the same folder.
14:02 And this is also the level where we can
14:04 start to add real-time context by
14:05 automatically adding your meeting
14:08 transcripts uh your daily task updates
14:10 or analytics through scheduled tasks.
14:12 This schedule task for example
14:14 automatically updates my second brain
14:16 with all of the uh meeting transcripts
14:18 across my team every day by using
14:20 Firefly connector. I also have a
14:23 schedule task here for team task roll up
14:25 which checks every day what my team has
14:27 been working on and updates that to the
14:29 second brain. You can also do this for
14:31 analytics. And then I can also do things
14:33 like a morning brief where it pulls
14:36 context from my second brain, knows my
14:38 priorities, knows our to-do list across
14:39 the business and gives me an overview of
14:41 what's important today. With this setup,
14:43 it also allows us to build better skills
14:45 and build them faster because through
14:47 this setup, we already have in-depth
14:48 context around our business, which we
14:50 can refer the skill to. So all we need
14:53 to do is lay out SOPs or workflows and
14:55 link them to which files in the second
14:56 brain it needs to read to get more
14:58 context. So in this setup, I highly
15:00 recommend starting to build your skills
15:02 a little bit differently. So instead of
15:04 adding context docks into the reference
15:06 files inside of the skill, you actually
15:08 want to make the skill reference where
15:10 it can find the reference files in your
15:12 second brain. For example, as you can
15:13 see, I did in this one, it only has a
15:16 skill MD with references to where it can
15:18 find the different reference files to do
15:19 its job better. And this means when I
15:21 make an update on my ICP document, all
15:23 my skills that refer to that file are
15:26 instantly updated to. You can do this by
15:27 just telling cloud I want to adapt the
15:29 intro scale. I want you to add the
15:30 reference files to the ben iOS and make
15:32 the scale reference the files instead of
15:33 having them in the reference files
15:36 inside the scale. Now a couple of things
15:38 become important at this level. Firstly
15:40 the setup and the file structure are
15:42 important to get right because of course
15:43 you're managing a large amount of
15:45 context. Now that's why I highly
15:47 recommend you use Obsidian which is
15:49 basically a free tool that helps you
15:51 visualize a folder on your computer with
15:53 some extra benefits. You can download
15:55 Obsidian for free by just going to their
15:56 website. But it's important to
15:57 understand that Obsidian is not a
15:59 cloud-based software. It's just a tool
16:01 that helps you visualize, organize, and
16:03 structure a folder on your computer in a
16:05 better way. As you can see here, because
16:07 of course trying to do that inside of
16:09 the actual folder with a growing context
16:11 like this, it becomes hard to do. We
16:13 also get a nice graph view here to see
16:15 all the relations and connections
16:17 between all of our context files. And
16:19 then for this file structure here, it is
16:20 a nuance topic. There are some best
16:22 practices, but it will depend on your
16:24 unique situation, your business, and
16:26 your way of doing work. Now, I recently
16:28 did a full tutorial where I show initial
16:30 file structure that I've seen work well
16:32 for most businesses or solopreneurs,
16:34 which I'll add in the link in the
16:35 description below, too, together with a
16:37 plug-in that we've develop developed
16:38 that you can install in cloud code or
16:40 cloud co-work that walks you through
16:42 setting up this initial file structure
16:44 with the context for yourself. Now, that
16:46 plugin you can download and use together
16:48 with all our other plugins and skills
16:49 we're building out internally in my AI
16:50 accelerator in the first link in the
16:52 description below. You also have more
16:54 in-depth step-by-step guides on helping
16:56 you set up this OS and one-on-one uh
16:59 live help and multiple Q&As every week.
17:00 So, if that's interesting, you can check
17:01 it out in the first link in the
17:02 description. But you can definitely set
17:04 this up yourself. I think my last video
17:05 will help you a lot wrap your head
17:07 around the file structure. And it's also
17:08 important to understand that the file
17:10 structure and the context will grow
17:12 naturally and fall into place more and
17:15 more uh the more you use this. So the
17:16 important thing is to just get started.
17:19 Now secondly uh your cloud MD becomes a
17:21 much more important to optimize at this
17:23 level together with potential index
17:25 files. Now what is the cloud MD? The
17:28 cloud MD is essentially an instruction
17:30 layer between your agent and the
17:32 Obsidian vault or your OS folder. And it
17:34 basically makes sure cloud knows where
17:37 to pull context from in a situation and
17:39 where to update it. So it's just routing
17:41 it to the right place which you can
17:42 imagine becomes a lot more important
17:45 when the context grows. So you can see
17:46 here in this chat in coowwork where I
17:48 give it access to my OS folder. It has
17:50 an instructions document or the clock MD
17:53 here. And this basically lays out how to
17:56 use and navigate the folder.
17:58 instructions on what to do at the start
18:00 of every conversation, how to route
18:02 between knowledge with information on
18:04 how the folder is structured,
18:06 information on Obsidian syntax, how to
18:09 add wiki links, and rules on how to use
18:11 context inside of this folder. Now,
18:13 again, even the Cloud MD will naturally
18:15 evolve and get better the more you use
18:16 it. And Clot can create the initial
18:18 version itself. Our plug-in will also
18:20 help you with a cloud MD instruction
18:22 that worked well for us. And then Andre
18:23 Karpathy, one of the leading AI
18:25 researchers, recently added a new layer
18:27 to this too where if your context grows
18:30 even more, you can start using index
18:32 files in each of the subfolders. So your
18:34 agent understands better how to navigate
18:35 each of the separate subfolders. For
18:37 example, you can see on the shared
18:39 context file, I have another cloud MD,
18:40 which in this case we just called cloud
18:42 MD, but it could also be called an index
18:44 file with more information on how this
18:46 specific subfolder is structured, which
18:48 cloth can read to know how to navigate
18:49 this folder. And we have another one
18:51 here for each of the subfolders. Again,
18:52 this is something you want to start
18:54 thinking about when context is growing.
18:56 And Cloud can help you out with building
18:58 uh these documents, of course. Now,
18:59 thirdly, what's going to be important is
19:00 you probably need some of these
19:02 scheduled tasks to make sure your second
19:04 brain is up to date, just like I showed
19:06 you with the meeting transcripts, but we
19:07 can also do this for your daily
19:10 analytics, task lists, your CRM
19:12 pipeline, whatever is relevant to you.
19:13 And fourthly, uh which is an important
19:16 one, uh is there is a maintenance aspect
19:17 to this. I'd highly suggest going
19:20 through your files on a weekly basis to
19:22 make sure things are going right. Are
19:23 there no duplicates? Are the documents
19:25 put in the right place? Are there any
19:28 conflicts in the context? And you do
19:29 want to dedicate some time to this,
19:31 especially at the beginning because it
19:33 will take time to get this right. And
19:35 lastly, you will need to start using
19:36 this cons consistently. The only way
19:38 this is going to work for you is when
19:40 you use it a lot because the more you
19:42 use it, the better it will get. And
19:44 there is some learning curve attached to
19:45 this. I can tell you I'm definitely not
19:47 there yet, but it is getting better and
19:49 better and it's making a big impact on
19:52 the relevancy of my AI outputs across me
19:54 and my team's chats. But it does take
19:56 time to figure out what file structure
19:57 makes sense for you and to test its
20:00 capabilities. Generally, I try to
20:01 approach this with a mindset of trying
20:03 to let AI clock code or clockwork or
20:05 wherever you use AI become your main
20:07 operating system for work because if you
20:09 do that, you'll slowly but surely fill
20:11 in the gaps to actually make it become
20:13 your main operating system. And this is
20:14 where we're heading anyway. So you might
20:16 as well be early. Now if you want to
20:17 take it to the last level, which is
20:19 going to be a game changer for anyone
20:21 who runs a business, is to actually roll
20:23 this out and sync this entire context
20:25 data set and the skills across the
20:27 entire team. So all of your team
20:29 members, AI agents instantly become far
20:32 more powerful for your business. Now,
20:34 when rolling this out for teams, a few
20:35 things of course become important.
20:37 First, the file structure will probably
20:39 have to change a bit, and you'll need
20:41 some more files for using this across a
20:43 team. For example, in my business OS,
20:44 you can see I have a few more folders
20:47 like my departments, my team and their
20:49 roles and plugins and skills so they can
20:51 be easily shared across the team. Again,
20:53 if you want to learn more about the file
20:54 structure, I covered it in full in that
20:56 last video which will be in the link in
20:58 the description below. Now, for your
20:59 team members to get to this initial
21:01 setup, you can of course just share a
21:03 zip file with them uh of the entire
21:05 context doc so they can install it. And
21:07 then second of course when you want to
21:09 share this across the team updates need
21:10 to actually be synced across the team
21:12 and ideally in real time. Now we've
21:14 explored multiple options of doing this
21:16 and because of course they are local
21:17 files it's not extremely straightforward
21:20 to do but for syncing across a team uh
21:21 you have multiple options. First you can
21:24 use GitHub. Second you can use Obsidian
21:26 sync which is a feature of Obsidian.
21:28 Third you can even uh launch a
21:30 self-hosted solution to do this. But the
21:32 best option we found which we're
21:34 currently using is a plugin inside of
21:36 Obsidian called Relay. Now this is a
21:38 community plugin inside of Obsidian
21:39 which you can find here by going to
21:41 settings, clicking on community plugins,
21:44 go to browse, type in relay and from
21:46 there you can install it. Once you've
21:48 installed it, it'll be listed under your
21:51 community plugins. And through relay now
21:52 I can decide for each of the folder to
21:54 which of my team members these updates
21:57 need to be synced to. And through this
21:59 any change anyone in my team makes in
22:01 any of these contact stocks will it will
22:03 automatically be synced and updated
22:05 across anyone in the team in real time.
22:07 And you can use relay uh for up to three
22:09 people uh for free. But with this setup
22:11 of course uh permission settings become
22:13 important too. Not every team member
22:16 should be able to update every file or
22:17 not even uh every team member should be
22:20 able to see or access any file. So you
22:21 as a business owner of course need to
22:23 control some of these permission
22:25 settings. Now, unfortunately, this is
22:27 not very straightforward to do on Relay
22:28 yet. We actually talked to the founder
22:30 of Relay and this feature is in their
22:32 pipeline. But in the meantime, we've
22:34 built our own version or custom setup on
22:36 top of this relay plugin that actually
22:37 gives us these permission settings. And
22:39 see that we have installed here our Beni
22:41 relay plugin, which is just our version
22:43 of this app with permission settings.
22:45 And now, for example, I can still make
22:48 updates in this general context folder
22:50 uh with the important documents, route,
22:52 strategy, etc. But my uh team members
22:54 only get read access. So they can still
22:56 use these documents but they can't
22:58 actually update them. This is what one
23:00 of my team members would see a little
23:02 lock with this is a readonly file. Now
23:03 this setup is a little bit more
23:05 technical. Um so if you want to set this
23:07 up we have full guides together with all
23:08 the other guides on how to sync across
23:10 team members with GitHub and other ways.
23:12 So if that's interesting to you and you
23:14 want access to our customized plug-in
23:16 again you could check out the AI
23:18 accelerator. And then lastly, of course,
23:20 once you have set this up, it is key to
23:21 have one person in the business really
23:24 be the operator and the manager of this
23:27 context layer because it does require
23:29 maintenance. It requires effort to
23:31 actually keep this updated and well
23:33 functioning across the business and this
23:34 is going to take some time. So someone
23:36 needs to be responsible. Now again, I
23:38 think syncing and permission settings
23:39 become a lot easier very soon because
23:41 there are a lot of businesses trying to
23:43 figure this out and there will be more
23:44 of an infrastructure around this very
23:46 soon. Um, but this is the way you can do
23:48 it right now. Now, that's it for this
23:49 video. Thank you so much for watching.
23:51 Again, if you want more step-by-step
23:52 guidance on setting this up for
23:55 yourself, uh, multiple weekly Q&As and
23:56 unlimited one-on-one live tech help, you
23:58 can check out my AI accelerator in the
24:00 link in the description below. Thank you
24:01 so much for watching. If you got any
24:02 value out of it, I highly appreciate a
24:04 like and a subscribe. It really does
24:06 help me. And if you want to learn more
24:08 about cloth co-work and obsidian setup,
