How to get started with OpenClaw | Alfred Simon, AI Architect @ Adwise

00:00:00: The number one thing you need to accept is that, You can't keep up.

00:00:04: It's impossible!

00:00:05: So if you expect to keep up with all the tools it's my job and I cant' keep up With

00:00:10: them.

00:00:10: There always a new kid on the block.

00:00:12: Now its like this open-clothed thing And i mean...you said..the first thing you have to accept Is That YOU CAN'T KEEP UP But you're also someone who immediately jumps into it.

00:00:23: When he came out My Twitter feed instantly got flooded With multiple people talking about him in like the first two or three hours.

00:00:31: Then I knew instantly, okay this is something.

00:00:33: tomorrow i need to book Two Hours In The Morning!

00:00:36: I Need To Test It Immediately.

00:00:37: Why Is This Substantial?

00:00:39: why should people care?

00:00:40: Open Claw is one of their most capable AI assistants at the moment.

00:00:45: that's on the market.

00:00:46: you have to think about it as a regular chat GPT Or Claude...it can do mostly things That Those Platforms Can Do For You But..It Has A Lot Of Extras.

00:00:59: The biggest difference between Openclaw and the chat GPT-like platforms is that...

00:01:05: Before we dive in, you're listening to the Masters of Search podcast with me your host Niklas Buschner.

00:01:11: Each week I sit down with some of these smartest people around the world on SEO and AI search.

00:01:22: If you enjoy this podcast, don't forget to like and subscribe.

00:01:25: And follow it in your favorite podcasting app or YouTube – It helps us get top-notch guests and create the best possible content for you!

00:01:33: Let's dive into todays episode.

00:01:36: Today's episode will be about AI But not the chat GPT kind of AI but building agents' kind.

00:01:44: We'll talk about what Alfred has been cooking over the last month, what FAFO really means.

00:01:49: Not just what the acronym stands for and about this new kit on the AI block that's called Open Claw.

00:01:57: if you want to understand where AI powered marketing is actually heading not The LinkedIn hype but the messy exciting reality of building stuff every single day.

00:02:07: This Is THE conversation.

00:02:09: That being said welcome to the podcast Alfrid.

00:02:13: Thank you very much, Nicholas for the invitation and happy to be here.

00:02:17: And talk about yeah one of the topics that nobody talks about AI.

00:02:22: Yeah I also feel like we have Uncovered a gap year in especially LinkedIn content with AI but Yeah jokes aside thanks so much for taking the time For people too better understand why?

00:02:37: You are a perfect person to listen to when it's this topic, can you share a little bit of context about your role and what do basically day by day?

00:02:49: Yeah.

00:02:49: Of course!

00:02:50: So my job position on paper is stated AI architect.

00:02:56: that's the official nice name which we also use in more normal ways.

00:03:01: but I call myself a FAFO engineer.

00:03:03: That's F around and find out Mostly At the agency, so I work at a Dutch Agency called Advice.

00:03:12: It's digital marketing agency and my role is to help our clients integrate AI in way that makes sense And it also useful and possible.

00:03:23: Also same way other fifty percent is to have the agency itself.

00:03:28: So our teams services how we can make something faster How you do with AI.

00:03:34: And it's a lot of experimenting, implementing things and not just playing around.

00:03:41: but we also have already running systems in production.

00:03:46: Either at us or either some our clients.

00:03:49: So yeah basically that is in the nutshell.

00:03:52: But if there anything else you want to know feel free to ask.

00:03:56: Of course, I always loved these videos on YouTube where it's like day in the life of a... I remember software engineer at Google.

00:04:05: Day In The Life of even investment banker because sometimes you generally understand what this people do but don't have clear visual manifestation how their days actually looks.

00:04:18: so can take us through typical one-day AI architect at advice.

00:04:26: Yeah, yeah I can't...I start really early.

00:04:30: so my wake-up is around four forty five four fifty depends a bit on when the dog wakes up and then i just quickly go out take the doc to tour.

00:04:44: basically after breakfast i go straight away to work.

00:04:48: what's exciting but also at the same time challenging about this role.

00:04:53: I really don't have two days that are the same.

00:04:56: So it's hard to find like, obviously...I try to have a return where i work in the morning-i had two or three hours focused work.

00:05:04: but the topics projects i worked on is constantly changing and sometimes i don't even know Monday what will come up on Wednesday.

00:05:16: One day can be, for example if I take a typical day.

00:05:19: For example yesterday i was three hours at the client talking about their new plans on how we could integrate AI.

00:05:26: We already made proof of concept and tested it And now had to present that proof-of-concept with the client himself To his team's management.

00:05:39: This is what we've been building in past couple weeks.

00:05:43: It's working!

00:05:44: Made a demo And then, yeah.

00:05:47: We just had to present it and now we are waiting for approval.

00:05:50: so he is waiting for the person that's building inside a company.

00:05:55: So we can take it further because not as just proof of concept.

00:06:00: That was in the morning.

00:06:01: Then I had to finish up something... We were working on reporting agent inside an agency.

00:06:09: One of my agents took me into certain level.

00:06:14: take it to deploy into a Docker server so we can run certain things also on the server side.

00:06:21: My colleague, which is a lot more technical than I am finished up part of it and then i finish that in afternoon uh.

00:06:28: And then rest day.

00:06:30: they was just looking for some issues and bugs.

00:06:34: because one thing nobody talks about AI systems are in production.

00:06:38: That's going break alot types And the more system you have running, The more systems You need to maintain and fix.

00:06:46: Since we are at the agency right now, Fourteen active systems that are running in certain ways.

00:06:52: Some of them run in NA-X, some outside NA-x.

00:06:56: There is always something broken.

00:06:58: Somebody sends a slack with an image That's not working.

00:07:02: Usually I spend like one hour A day trying To fix those as much As i can.

00:07:10: If it's a sport day, I go to sports and after that i do some work in the evening.

00:07:14: And basically thats it.

00:07:15: but yeah mostly its like ten-thirteen hours of work per day about learning about AI and then trying projects and things like that.

00:07:24: What would you say?

00:07:25: Why do AIAgents break so often?

00:07:28: Do feel they break more than traditional software products pre-AI era?

00:07:36: Not really...So software wise don't really break, it's more like the response they give.

00:07:42: It's not always what is expected.

00:07:47: What I see from our examples is that forty percent of users are not using it in a way that was intended to be used with every software i guess.

00:07:59: But no but people need use how to prompt.

00:08:02: so most times colleagues put something into it and then they get something, but the way they asked for it or the data that they gave to AI had no chance of giving a correct answer.

00:08:15: The problem is even if you set this system from these systems try to help in any possible ways.

00:08:21: so they tried to come up with an answer.

00:08:23: sometimes its wrong and then that frustrates people.

00:08:26: Other things are just timeout errors Something breaking in production, thirty percent of their error.

00:08:33: probably me making a mistake when I created it.

00:08:36: because yeah, these are most of them.

00:08:40: I would say wipe-coded with an intention.

00:08:42: so i'm vibe coding in the way...I know exactly each file what needs to do on their project and where the folders are database.

00:08:50: So understand that this is not completely wipe coded but obviously the code.

00:08:54: ninety percent of the code is written by something we break here there you also have fixed And

00:09:02: can you share an example of a product or an agent, etc.

00:09:07: You build maybe some that is best for people also to understand what's possible?

00:09:13: So probably not the most complicated thing you've built but may be something just very easy to imagine how it works out in day-to-day either at agency and clients.

00:09:31: Yeah, I can share an example from our agency.

00:09:35: You listen to also digital marketing so it's gonna be easy to understand.

00:09:40: the social team needed a quick tool where they can A-B test certain creatives and just making small changes to image itself.

00:09:50: So when that design team or client creates creative sometimes you want button on the image or something in an image changes from green to blue and then just small changes.

00:10:01: And those, yeah instead of asking someone for a design team are going yourself into Canva.

00:10:07: we're making the changes.

00:10:09: Nano Banana Pro is exceptionally good at those things.

00:10:13: so I created this small tool where they can upload images give context about what they want to change, select the client if it's available also in the database.

00:10:23: So we already have some context of how they are doing and their brand colors or things like that And then yeah... The user just has to say okay this is an image.

00:10:32: I wanted to change That button from green to red.

00:10:36: They get the output, download it immediately.

00:10:38: a new version almost same creative.

00:10:42: What would you tell people?

00:10:44: say, hey okay but I can also just do this in Gemini and then select Nano Banana Pro.

00:10:50: So why should i build a tool for this use case?

00:10:55: Yeah one thing is that in Geminai yeah you can create the gem where you can also upload the brand colors and things like.

00:11:03: That's nice for one client But when you need to create gems For each of your clients it will take always bit longer.

00:11:11: so with these tools we have So they just need to get a selector.

00:11:18: I'm now working with this client, so there are already most of the colors that he can use.

00:11:23: and brand tone of voice because it's not only color changing.

00:11:27: for the A-B test.

00:11:28: you also write in new text.

00:11:29: You don't have to explain what words we want.

00:11:33: some clients like these Type of text or not and then that's already over.

00:11:39: there is an accessible And yeah, it's in this way.

00:11:42: It's better.

00:11:43: the other thing In nano banana you can get rate-limited with this weekend because we use straight API.

00:11:49: so if Ten colleagues are working on it at the same time when everyone sends a request.

00:11:55: Yeah We just scale with that and in Gemini At some point run into some rate limits as well.

00:12:04: And how do you decide if for example someone from the team or from a client comes with a request.

00:12:11: For hey, couldn't we do this in a better way?

00:12:14: Or couldn't maybe use AI for that?

00:12:18: How did he decided between different options like okay We can just use like chat GBT Gemini Claude maybe a little bit better prompt, but we use the normal interface and then We can use a cloud project custom GPT or Gemini Jam.

00:12:38: So create dedicated instance so to say in tool for it.

00:12:43: And okay?

00:12:44: We have to build A standalone agent our standalone tools.

00:12:48: what's your Decision?

00:12:51: making checklist so-to-say for these requests?

00:12:54: yeah

00:12:56: It always starts with the request and what they actually want as output.

00:13:01: So when I start this, And i have like a talk or small meeting With that client Or team members That wants something to do with AI My first question is Always What would you Like To Be The Output?

00:13:14: How often Do You Need That When You Talk With This Tool Agent?

00:13:19: Whatever?

00:13:21: What Is That You Want To See As A Response?

00:13:25: most of the time.

00:13:26: that really helps deciding on which way we need to go.

00:13:30: So if they say usually I want have a conversation about something, so let's say i want talk my data then building full agent is not always best option.

00:13:41: Then you just need find away for this person.

00:13:45: They can get their data in safe ways To already existing platform like Cloud or Gemini or Chelsea Pity and then they can talk there.

00:13:55: The reason for this example is that inside all these tools, a lot of other tools the agents can call.

00:14:02: so if you talk with chatGPT he can search internet things like that And people are used to it.

00:14:08: If I recreate as an agent, stand-alone agent i need to recreate those too.

00:14:12: That takes lots time A lot experimenting.

00:14:15: They won't work.

00:14:16: good what openly created.

00:14:19: So when someone wants to talk today Then usually I recommend we find a way for you to get that actual data from BigQuery, Google Analytics or whatever.

00:14:35: If someone wants an automated process every day then those platforms automatically are not an option.

00:14:44: So then you need to look into, for example if it's not hard to make.

00:14:47: the NA-TEN is a nice option depends.

00:14:50: so when they say okay I need this every day at certain hour delivered to a certain place.

00:14:57: Then next question what comes?

00:14:58: Okay where is your data?

00:15:01: how much reasoning the agent has?

00:15:06: And based on that, I can already see if it's like a multi-agent process.

00:15:10: So we need two or three of four agents to get the answer.

00:15:14: then probably NATEM won't be good fit because of timeouts and crashes.

00:15:19: It is not always reliable.

00:15:21: Then you have to think about another framework.

00:15:23: but If its just one agent with certain API calls and send as an email than you dialed exactly ok then go with NATEN.

00:15:31: Its easy to build and fire up.

00:15:34: So yeah, mostly this is my checklist.

00:15:36: I try to find out what they want as an answer.

00:15:38: do They need it timed?

00:15:39: Do they need in a certain place or they just want to talk with you?

00:15:42: then that decides from In which direction we go from there.

00:15:46: Hmm when did you become an AI architect?

00:15:49: Or when did like your role change?

00:15:52: maybe also if officially

00:15:54: Officially changed in July last year and so obviously people don't know me but i have no AI background at all.

00:16:05: So I am not even a computer scientist or something like that, i started out as a regular normal google specialist.

00:16:13: uh...i have nine years experience in that field and thats what i did the agency before i start working on ai projects.

00:16:20: And this is came pew from..I loved it ,and saw an opportunity here.

00:16:27: when chad gpt got released On the day of the release on Twitter, someone was posting about it and I also immediately tried it.

00:16:38: And i was instantly hooked!

00:16:40: I was the first guy who said to the agents because next time Monday i went in... ...I showed everyone hey this is a nice tool you can talk.. You get answers and he can do lot's things.

00:16:49: Nobody understood at that point what am talking about.

00:16:51: two weeks later everybody was talking about Chagapit.

00:16:55: And ..i just loved it !

00:16:56: I saw opportunity ,things that you could create.

00:16:59: how many chances you have.

00:17:02: I spent every morning, before work after-work weekends everything building certain things trying it and playing around with the tools that are available.

00:17:11: when a new mother came out.

00:17:13: I tried it instantly.

00:17:14: When i saw a new tool like try this instantly.

00:17:16: And I shared all these processes inside the agencies so talked about it.

00:17:21: I shared certain tools that I made.

00:17:23: then obviously The leadership at the agency said okay Yeah, maybe we should give him some more time to actually have dedicated time for this.

00:17:35: And then that increase from ten percent to thirty to forty-fifty and in July we decided okay I should go there probably full on this.

00:17:44: Do you think or do your... Maybe even see This role also popping up at other companies because i feel like it's still fairly new?

00:17:57: Probably I can dare to say we were the first agency in The Netherlands.

00:18:01: And, uh...I don't know about other parts of the world but i probably.

00:18:05: We Were the First Agency at the Netherlands that had a full-time someone That was working with AI related things.

00:18:11: Um..i'm not sure if this is gonna be standard or Not?

00:18:14: I am definitely Sure you will need Someone that Focuses on AI more than Other Things For.

00:18:21: so If I take the agency model I really see the benefit Of of having someone like this at the company.

00:18:28: The thing is that if you have some full-time doing things like these, then they can have all kinds time to experiment new tools find out what's working and see patterns.

00:18:39: What do we need to try?

00:18:41: Because when other team members come from their industries vertically with a question You know exactly which tool could help them because So many already.

00:18:53: And you know, do we need to build something or don't need to be at something?

00:18:57: Uh and You can help them a lot easier.

00:18:59: so if you expect like A dedicated SEO specialist Or someone that's working with social ads To also create the tool while they are Managing clients and things Like That.

00:19:09: it is going Take a Lot Longer.

00:19:12: Doing It in tandem With Someone that's Focusing on creating The Tool Getting feedback from the actual Specialist Can Help a Lot.

00:19:19: Yeah!

00:19:19: You Need a team or one person at least that understands all the things that comes with AI.

00:19:25: I still don't understand lot of it, but... ...I do understand.

00:19:29: i think a lot compared to colleagues who dont spend.. ..a lot time playing around these two.

00:19:37: And what helps you keeping up with all the news?

00:19:40: Because every week probably there is new release.

00:19:44: either Claude releases a new model or there's, I don't know.

00:19:49: A completely new set of features coming out.

00:19:52: then There is like New agent platforms popping up etc.

00:19:57: so i feel Like when I talk to people that are maybe still So they're way earlier in their AI journey than you for example.

00:20:07: One Of the things That I hear The most Is that They Struggle with Keeping Up And then you already feel left behind if your not like on the cutting edge.

00:20:18: Where probably that's no true, but how do handle this?

00:20:22: maybe even more mindset wise or mentally?

00:20:28: Yeah!

00:20:28: The number one thing to accept is that You can't keep up.

00:20:33: It's impossible.

00:20:34: So If you accept it with all of these tools its my job and I cant' keep up While doing all the things and creating projects, there's no chance keeping up.

00:20:45: I have an AI agent running on for myself.

00:20:48: that sends me sometimes news... ...I don't even have some time to read what he sent.

00:20:51: so let alone keep up with the news You can!

00:20:55: So i think we need to accept at some point you won't be able to keep up it all the tools.

00:21:00: My thing is how do this once you get into it?

00:21:05: At a beginning its really hard.

00:21:07: You're going to see a lot of tools, you are gonna see ways to do it and everything like that.

00:21:12: As soon as you get more into it... ...you will see most of the tools are just behind.

00:21:19: what they do is different.

00:21:20: So if for example you understand Context window And how this works It doesn't matter which coding agent your using.

00:21:27: You use.

00:21:28: Cursor, Cloudcode, Gemini, Codex Works in the same way.

00:21:31: One has bigger Context windows one has lower Context but The logic behind it is the same.

00:21:39: As soon as you understand these things, like with everything... Getting up on speed in a new model or tool isn't really difficult after that.

00:21:50: Usually what we can expect from a new module is to do more and have better context window understanding.

00:21:55: so What I usually do when doing something issue, for example with coding projects and AI can't solve it And I see that we just running around its circles.

00:22:06: I make a note of It i've write down what?

00:22:08: I tried and why it failed and then the new mother comes out.

00:22:12: Usually go back to those notes and I try A few of those issues again With The New Mother To See What Changed.

00:22:17: Does The Thinking Logic Changes Of The Model Or Not To See?

00:22:24: But Yeah This Obviously you need this first to start using it.

00:22:28: so I think You don't have to be like scared or something.

00:22:32: I think the best is pick one of the bigger players, Gemini, Claude whatever and start actually making something.

00:22:41: do a coding project.

00:22:43: Everyone has ideas.

00:22:44: everyone has A tool that they want Or try To make.

00:22:48: That see where it takes you to understand how you work with these tools because as soon As sooner you Understand It's The Easier We'll Be A bit.

00:23:00: I don't know if this makes sense?

00:23:03: Yeah, it make sense totally but there is always a new kid on the block as i also said in the introduction and now its like this open-cloth thing.

00:23:12: so... And you said that first things to accept are that You can't keep up But As far as I know.. You're someone who immediately jumps into It.

00:23:24: So yeah!

00:23:26: This is what I forgot.

00:23:27: say If you are on Twitter or LinkedIn, I recommend if you really want to be up-to-date with AI stuff go to Twitter.

00:23:35: Because what i see on LinkedIn sometimes is two or three days later.

00:23:40: so people sharing things on LinkedIn and that's all news posted yesterday.

00:23:48: when something big comes out because there's so much noise everyone is testing everything.

00:23:53: When Something Big Comes Out we will see it everywhere.

00:23:55: So instantly okay...I need to try this.

00:23:58: Every small tool release, you won't be able to try it.

00:24:02: But when for example Open Claw is now the name but then launch was Claude... When this came out my Twitter feed instantly got flooded with multiple people talking about that in like first two or three hours.

00:24:15: Then I knew instantly okay This something.

00:24:17: tomorrow i need to book Two Hours In The Morning.

00:24:20: I Need To Test It Immediately.

00:24:22: If like ten of other people talk on X I know Okay!

00:24:26: This probably serious So I need to look into it.

00:24:30: And yeah, this is what helps.

00:24:32: so people are helping actually each other with keeping up the news because there's so much tool come releases every day and something that's mediocre doesn't work at the time.

00:24:43: People won't talk about you that much.

00:24:45: when something big jumps everyone talks about It?

00:24:47: Then You Need To Probably Test.

00:24:50: So please take us With You two these Two Hours That You Blocked After Open Claw or back then called ClaudeBot came out.

00:24:58: What did you see?

00:25:01: Why is this substantial, why should people

00:25:03: care?".

00:25:04: Yeah so first of all what.

00:25:08: OpenClaw is one the most capable AI assistants at the moment that's on the market.

00:25:16: You have to think about it as a regular chat GPT or Claude.

00:25:22: It can do more.

00:25:23: certain things those platforms can do for you, but it has a lot of extras to it.

00:25:29: So first off all... You can install OpenClaw on your device.

00:25:34: I don't recommend doing that Definitely not in the main devices which we are using daily Because the biggest difference between OpenCla and ChagPPT-like platforms is that OpenCLA does everything with your device So it can open a browser, delete files and search for files.

00:25:55: It does this.

00:25:56: so if you have experience with Cloud Code then you understand OpenCloud quite fast because OpenCloud can run terminal commands in your device And with terminal commands.

00:26:09: basically you do almost everything on your laptop.

00:26:14: So this is one of the biggest differences.

00:26:17: I made some notes for myself to not forget, To mention most important things!

00:26:23: Let me just make sure that i covered all these... Yeah so This Is One Of The Biggest That It Can Run The PC Actually As A User.

00:26:36: So That Gives So Much Freedom With it That Its Insane.

00:26:42: The other part is that it's really autonomous, so you don't have to explain a lot of things.

00:26:47: For example when I was setting up OpenClaw usually You can set this by just talking with it.

00:26:53: So once its running on the server or your device and start chatting With it then say for example i want to activate nano-banana pro skill in your toolbase so you Can generate images from me And use it in discord.

00:27:12: And I saw that he's typing, typing, thinking and then it came back with the answer.

00:27:15: Yeah, Nanobanana is activated.

00:27:17: all i need from you Is to put on the server The API credentials.

00:27:21: cool okay?

00:27:22: He made It!

00:27:23: I Put the api credential in and it was working.

00:27:26: Honestly, I have no idea what he did In the background but its working since.

00:27:30: So Its really autonomous so You don't Have To explain everything What he needs.

00:27:34: do things like That.

00:27:37: Also The other part, the third thing that makes it super interesting is that... It has built-in wakeup moments.

00:27:47: So there are two things.

00:27:49: they're called and we can go into detail later on obviously but one is called a hard beat And others are cron jobs so hardbeat!

00:27:56: The standard setting for the heartbeat Is that it wakes up every thirty minutes And it sends you like a DM message or he checks on certain things.

00:28:05: That is configurable by the user, so what do want?

00:28:09: The agent to order AI assistant doing every thirty minutes You can make?

00:28:13: each hour everyday depends on your settings and the crown jobs are like tasks that you can schedule just Like for example with any tenant Every day at eight o'clock.

00:28:22: send me report about this.

00:28:24: I think these are the three biggest differences compared to just a chat GPT account.

00:28:35: Okay, cool and why do people now talk to Cloudbot via WhatsApp?

00:28:42: Because this is something that felt fairly new compared using ChatGPT or compared to cloud.

00:28:49: so like what's behind it?

00:28:56: Why is this happening?

00:28:57: And it's also something where you feel like, This part of the reason why got so popular because It feels so natural.

00:29:06: Like I can just message this bot on WhatsApp and will actually do

00:29:10: things.".

00:29:12: I think that helped to become...it made more personal.

00:29:16: So for example i don't use it with Whatsapp Because the recommendation if using it with whatsapp get for the agent a mobile phone number, separate numbers so you don't run from your own number.

00:29:27: I use it via Discord but that's also the same.

00:29:32: It just fun because we talk with an AI on a platform which is already familiar to us.

00:29:36: So like WhatsApp everyone uses it and talks about someone And then when they are on their way to work You message your assistant to look up something before getting into meetings.

00:29:46: It can do it for you.

00:29:47: So once to get there, You already have the message on your WhatsApp?

00:29:50: If we really feel like I think For people...it feels Like for me that you are talking To a virtual assistant.

00:29:56: That's there for you twenty four seven and it Can Do basically anything for you that you ask And that probably Helped A lot to make it more.

00:30:07: Yeah We can Get attached to it faster than other chatbots.

00:30:12: And can you check some of the use cases that you already explored?

00:30:16: Like things, maybe even things where your felt like okay this is not working out so well yet.

00:30:25: So we're doing a little bit of expectation management here also with people but then also things were you felt like hey this is really great and I'm surprised by how well it's going.

00:30:37: Yeah At the moment I call her Lyara, so you have to give the assistant a name.

00:30:44: When we are setting it up and gave him his name then they have what's called Soul.MD.

00:30:50: So like how do they need to act?

00:30:52: Yeah It is interesting.

00:30:56: We quickly explain here for people.

00:30:58: Soul.md because their people might hear it... ...so its file that is called Soul And MD stands from Markdown.

00:31:07: So it's basically the soul of this bot written down how... It is shaped, so to say.

00:31:15: And I laugh because its so human!

00:31:18: Yeah exactly in that sole file standard comes out when you install it already there has a bit of instructions and one of the instruction over here i think literally says This Is Your Soul keep it maintained add things to it, remove things from as you see fit.

00:31:40: So the creator of Open Cloud deliberately created in a way that like can change its soul and what's fascinating is really working so soon when we start talking about it adapts.

00:31:52: for example now he talks with me exactly how I want Liara talk with me.

00:31:58: i get short messages to the point explaining What do I really want?

00:32:02: No fluff no nothing.

00:32:03: That's How I communicate As well.

00:32:05: And it's so nice to see that he changes over time and remembers things because It also has a memory markdown file.

00:32:13: In the system prompt of this agent, is stated That needs to log for itself The memory files.

00:32:21: So when you think something important will make an entry automatically in the memory file.

00:32:26: When we talk with Based on the question.

00:32:29: The model already understands, okay I might need to look up the memory because he's asking something about the past.

00:32:35: then He searches the memory fire quickly and then ultimately remembers oh ok we talked About that with me made these certain things And this two things makes it really personal.

00:32:47: but the original question one for why i use it.

00:32:53: It's important to explain the solar family Because I used it.

00:32:57: One of the things are you sit forward.

00:33:00: It's like Lyara is just now at the moment, as a co-worker that backdoor boosts.

00:33:05: That's an app I'm working on and don't have too much time to work in the app.

00:33:09: so I realized okay this is nice use case for open cloud.

00:33:12: So she can pick up things from me that i dont' have time For And then we move one form there.

00:33:18: At the moment We are using it for creating block entries for us To generate content.

00:33:24: Creating images.

00:33:26: that was my biggest time killer.

00:33:30: So when I created the blog even with AI because most of it was already automated, The problem is that always had to think about the prompt.

00:33:39: so what image can you use in that blog article?

00:33:41: Because we need an image and since i'm creating a image creator app... ...I've always have to create images as examples for example from this product image using our app And then took lot's of time thinking about the prompts.

00:33:56: I just send a Discord message, hey write this blog and then create these example images.

00:34:02: Everything is uploaded directly to our CDN.

00:34:05: everything's over there.

00:34:06: it's done!

00:34:07: It's awesome.

00:34:08: so... This one of the things that she really good at writing blogs its hit-and-miss.

00:34:17: Its getting better due to SoulMD because i always try to give feedback on what's working, was not working.

00:34:24: So I scanned the memory and saw file a few days ago And it is really noting down things that i gave as feedback.

00:34:33: The blocks are getting better.

00:34:35: so if you install this for content writing expect at first weeks or days Terrible output.

00:34:42: then you need to give it feedback just like your employee.

00:34:46: It gets better.

00:34:48: now are really good, so I don't even have to give any feedback on them.

00:34:55: These are the two use cases that are most heavily used.

00:34:58: and third one is reviewing our current codebase.

00:35:01: So she has access full code of the app And goes through it then comes with suggestions...so i had a Chrome job setup.

00:35:10: She wakes up every day at nine o'clock.

00:35:12: She checks the whole code base and checks the metrics.

00:35:15: we've got How many people signed-up in past seven days?

00:35:20: Did we make any revenue or not?

00:35:21: And then she tries to help us with some features, what we can add and remove from the app.

00:35:27: Because I said your goal is get us to four hundred MRR at the end of February.

00:35:32: So since that time we added twenty five euros MRR.

00:35:36: so it's working...I don't think you will make a four-hundred but hey

00:35:41: Interesting!

00:35:42: What are something maybe didn't work out as well as you would have liked?

00:35:49: Code changes.

00:35:51: Okay, that's why you moved to just having viewing rights on the code base and making suggestions?

00:35:59: So she has only viewing rights but can make changes because like made a clone of the repository She can send poor requests.

00:36:07: so hey You should add this as we actually rise the code And then we decide okay do it or not We give feedback.

00:36:14: The problem is that the codebase we have, it's already so big.

00:36:18: That even when you are coding with Claude You'll need to direct in a certain way.

00:36:23: So AI doesn't get lost because It has so much code and context there that the Context window gets full immediately if you don't manage it correctly.

00:36:31: And thats something I see with OpenClaude.

00:36:35: Yeah its just messes up the Contxt window.

00:36:38: It creates really weird things.

00:36:40: Sometimes it writes a code.

00:36:43: I'm testing on Discord.

00:36:44: Did you actually test what you wrote?

00:36:46: Yeah, yeah!

00:36:46: I tested it and was like are you sure?

00:36:48: because i just tested then it crashed And then goes back with the message oh yeah sorry yeah i tested.

00:36:53: yeah It doesn't work...I need to figure this out.

00:36:57: So code changes are really a hit or miss.

00:36:59: Up until now in Total i think she created thirteen pull requests and i was able To use four of those Actually to push through production Something.

00:37:11: so That saved me a lot of time, actually.

00:37:13: And the other ones are good ideas.

00:37:15: just execution was bad.

00:37:16: so I save them and i will give it another shot.

00:37:23: might be my mistake as well... ...I don't use the most expensive model with OpenClub because I'm still not a millionaire to spend thousands and thousands of euros on API calls Because It makes some costs there.

00:37:37: hook it up to Opus.

00:37:38: four point six for example, because probably I will go bankrupt in a week.

00:37:42: So i'm using it right now In two ways.

00:37:45: Using with my open AI account You can log-in and then Liara uses the same account as I do With OpenAI.

00:37:53: And she is using Codex.

00:37:55: five point three which Is really good model.

00:37:57: Only thing that yeah It gets quite fast rate limited Because we are using The Same Account.

00:38:03: And then I use with API the Gemini a three point zero flash model because it's cheap It's fast.

00:38:10: It's quite good, but for coding is not The best choice.

00:38:12: so i still need to try it probably with four point six opus.

00:38:19: Yeah, I didn't dare.

00:38:20: at the moment who kick off?

00:38:21: two point six office

00:38:23: If anyone listening has a million spare, Alfred would be delighted if he gets a donation to... I'm

00:38:32: already happy with ten thousand.

00:38:34: Okay!

00:38:34: Ten thousand is enough.

00:38:36: so people let's make it work.

00:38:38: So Alfred can actually test Opus.

00:38:39: four point six with Lyara.

00:38:43: But i think here's one critical point and you mentioned that from the thirteen Pull requests you could use for now.

00:38:52: I can imagine people listening to this and saying, oh This is very disappointing.

00:38:57: so i would Stop using it then?

00:39:01: I Could imagine that You think about this differently.

00:39:05: Yeah, I know there's from the thirteen at least six was my fault.

00:39:10: giving bad instructions And setting up the other two fail.

00:39:14: basically The model used immediately explains why some of those fails failed.

00:39:20: So when she ran out off usage on the open AI model, The fallback module is the flash.

00:39:29: so I assume She was halfway through the task.

00:39:32: then the rate limit came in that she fell back to three point zero flash.

00:39:36: and if you ever coded with three points zero flash You know it's rubbish?

00:39:43: I am sure if i really like give it a good model and can run on the good model,I'm sure that four will jump immediately to six or eight.

00:39:59: But even with the four ,i'am already satisfied because those were also bad instructions.

00:40:04: but Those are things that didn't have to care about at all.

00:40:07: so If you code with cloud code It's fast!

00:40:12: You need be there talk into microphone and then transcribe it, accept things check code.

00:40:20: But with this all I did is sent a discord message to work And i came back and had the response say its done!

00:40:28: And just have to check the poor request.

00:40:32: If you bump up better model.

00:40:34: that's only win in situation.

00:40:38: And can you share something about the speed of execution?

00:40:41: Because I tried Claude for Chrome.

00:40:44: I haven't tried Claud co-work yet, but for Claude For Chrome i felt like it's really fascinating and generally working well But It takes Really long.

00:40:54: so it Takes forever to complete a relatively simple task.

00:40:57: So if You for example ask Yara To I don't know publish A new blog post and create an image for the app or the blog, for the apps website.

00:41:13: How long does it actually take to complete a job?

00:41:17: Honestly I'm not sure because the blog creation is scheduled jobs.

00:41:23: so she has one blog per week but i don't want too much content on our web site.

00:41:30: It runs every Monday morning at eight o'clock And then I just get the message that's done.

00:41:35: So, i'm not sure how long she takes but The message usually writes around like half Like eight thirty something like That.

00:41:42: But i'm Not sure if actually needs half an hour for the whole process or Just a message gets sent later?

00:41:47: I assume it's half An hour let's say Half an Hour!

00:41:50: But If you just chat with With the agent depends A bit on the task what You are asking but Usually It is quite fast.

00:41:56: so Uh...I remember Two days ago..i needed something figured out based on our data.

00:42:05: I was planning a meeting with my co-founder and i had no time to look up certain things in analytics, And...I was just stepping into the car from the office to come home before I drove away by the car.. I sent him a message saying hey!

00:42:21: Look at this.

00:42:22: and that I barely drove away and already got the message with the data.

00:42:26: So it took like two or three minutes, not even to go to analytics.

00:42:30: get all of that because she's using mostly API so its quite fast.

00:42:35: getting the data fast depends a bit on their request but It usually really really fast.

00:42:39: so i can't complain about speed.

00:42:41: Okay nice And which other use cases have you planned to test?

00:42:45: Because now your obviously settled already little bit.

00:42:48: You know how it works.

00:42:50: Yeah Adopted its soul communication style, it feels like you're through the early days of testing.

00:43:00: So I could imagine that based on how i know you... That at least in your brain there are already a lot of ideas cooking.

00:43:08: what's coming next?

00:43:10: Yeah so I upgraded this setup recently and for first one-and-a-half week I was running her on VPS which is not the best Option, it's probably the good option to start.

00:43:25: But if you really want to have clod open claw unleashed You need to give it a device because then he can use browser on away.

00:43:34: Yeah That exactly what I have.

00:43:36: a Mac mini Good job!

00:43:39: I'm just holding a mac mini into camera for everybody that is listening Because obviously i also saw someone unlinked in Posting about it..you should have separate machine blah blah blah.

00:43:48: but yeah

00:43:51: I totally agree with that take.

00:43:52: A lot of people say, i don't need a Mac mini you need a VPS.

00:43:56: so yeah if you are like an nerd vps is probably good because you can figure out the server and everything.

00:44:02: for me The choice was to go with a mac mini Because i understood okay If i needed this next level i constantly needs to log into my server make there certain changes...i have no idea how to run a server.

00:44:16: I barely done it few times..I had to ask Cloudcode To Make This.

00:44:19: For Me It's a lot work and I still feel that i'm limiting the potential.

00:44:24: With the Mac mini you also get The browser use case unlocked so she can really Use a browser like you do then, You Can say hey go to verser create an account put the project up And Do something or check Twitter?

00:44:39: Get back To me about this and This because otherwise you need to give in a VPS all kinds of API access and things Like That.

00:44:45: yeah it I don't think it's the ideal.

00:44:48: And also if you have like a separate device, It is bit more secure because You do not have to mess around with securing the VPS.

00:44:54: so... ...you know what you put on that seperate device and.. ..that only thing open clock can access.

00:45:00: So If you put there an API key then she has access to that API key.

00:45:04: Yeah!

00:45:05: If u see something going wrong just take that API Key away and you know That its more or less isolated On her VPS.

00:45:13: make sure Securing the server itself.

00:45:17: That's also a lot of work.

00:45:18: if you don't know what you are doing You can get in trouble people and getting to your server Get all your credentials things like that, so that's a risky move which we're not technical But I'm planning right now.

00:45:30: I have three plans for the upcoming period.

00:45:37: The second one is that I want her to create a project.

00:46:01: That will create, uh...I would say revenue but at least get users.

00:46:06: let's say that this so i won't.

00:46:08: I was.

00:46:09: I'll be saying hey This has come up with Project ideas.

00:46:12: start coding it and push it to virtual.

00:46:15: Create whatever you want make.

00:46:16: Make a business?

00:46:17: I'm curious if It can actually do that because I saw someone on Twitter that Was able To Do something like that.

00:46:24: So I wanted test that as well.

00:46:27: And then the third one is just taking it further with a coding part for backdrop boost because if she's running properly on separate device, than my research isn't wrong.

00:46:41: Then I can use her cloud account where she can log in and then she could use Cloud Code via an account instead of direct API which is much cheaper!

00:46:56: Pandora's box.

00:46:58: Yeah, I

00:47:00: was about to say the same pandora box.

00:47:03: We were approaching the end of The Rainbow Nice.

00:47:07: a quick question before we dive into this set up.

00:47:10: i know people want To Know how to Set This whole Claude But Thing Up because like it It sounds cool.

00:47:17: but then the Question is okay?

00:47:19: How do i get started?

00:47:20: and we obviously Want to deliver also on that promise.

00:47:22: um yeah why Is it a Mac?

00:47:25: Why is it not a Windows or like a Microsoft PC?

00:47:28: Like, why's everybody buying Mac mini now.

00:47:32: Because Mac Mini is stupidly strong for the price that you have to pay.

00:47:37: one of the reasons and The other one-the main reason for me For example Is I can use an iMessage To talk with him And that's even cooler.

00:47:50: So, the other thing is I can share a lot of things from my Macbook to her so Apple Notes and things like this because she will have an iCloud account as well.

00:48:03: It unlocks lots for me Because it makes it easier.

00:48:09: But I think the main reason is that Macminis are really cheap at the moment.

00:48:14: I think Mac Minis probably are a good investment for short term.

00:48:17: So if you buy them now and everyone runs out of stock, You can resell it.

00:48:20: probably few months later.

00:48:24: Not sure...I don't take this as financial advice but yeah i think that's one of the main reasons because mac minis are cheap and strong.

00:48:34: Got It!

00:48:35: I guess there will be someone who buys like hundred mac mini then sets all up And provides like as EA, as a service.

00:48:48: Like executive assistant-as-a-service thing where you can basically have your own AI Assistant which in the back end is a cloudbot running on a Mac mini that will then be personalized to you and you can share what do want to share?

00:49:03: And they set up couple of things already pre-made...

00:49:07: Already people are doing this.

00:49:08: I saw someone.

00:49:09: Yeah, I saw someone on Twitter.

00:49:11: There is a platform called TrustMRR.

00:49:15: It's a platform where you can.

00:49:17: So lot of people say love things and internet.

00:49:19: then mark Lou the creator of the platform created this platform.

00:49:23: You can connect your stripe key to it And then you can actually see if people talking about their app Actually make money or not?

00:49:32: Trust MRR that they created an open claw something so you can use it or, uh... You can set up there quickly.

00:49:39: I saw one of the guys made twenty thousand in like a week or something?

00:49:44: Crazy!

00:49:45: And i think if you go to platform and look at like Open Something Or Claudebot If we search with this We probably already see lot's services That are here People making serious money Already With The Idea That You Had That's coming.

00:50:00: Yeah It is already Here.

00:50:02: Interesting Very, very interesting.

00:50:05: So now let's get into how to set this up.

00:50:08: I know you prepped a little something for people too also better understand it.

00:50:12: Can You guide us through?

00:50:13: so if i feel like This is Interesting Alfred Convinced me?

00:50:18: I want To Get Going.

00:50:19: How can I get Started Tomorrow?

00:50:22: yeah If you are rich then buy A back mini.

00:50:24: uh-huh if you can afford Do It Straightaway if you Are Like A bit more cautious and don't want to spend four hundred five Hundred euros straight away Then You can probably do it the way I did.

00:50:35: So that is go to AVS AWS this Amazon's servers stuff.

00:50:42: They have a super generous free tier.

00:50:46: Basically, you get if you are making your first time account you got two hundred euros?

00:50:50: I think as budget which which is plenty enough so you can run your Open Claw for three four at least a long period.

00:51:01: I didn't even spend ten euros of that free budget.

00:51:08: You can create that and then on there you start experimenting, okay is this something that i like?

00:51:13: Is it something that don't like?

00:51:15: or nice?

00:51:16: ?You will probably have a lot of headache managing the server.

00:51:18: but for that ,I created some things to listeners And share my screen if u dont mind To walk everyone through this small thing.

00:51:30: For everybody that is only listening to this, we will also share what Alfred is showing us with a link in the show notes.

00:51:38: So you can check this out.

00:51:41: otherwise highly recommended going to YouTube and checking it out there.

00:51:48: I don't know the time stamp exactly.

00:51:50: That way We'll be on but probably somewhere around the fifty minute mark.

00:51:54: then You Can Check It Out On Youtube Also.

00:51:57: Yeah.

00:51:57: So what I'm sharing.

00:51:58: right now for the people that are not watching this screen is, i created a small repository on github with the cloud md file.

00:52:08: so if you're not familiar with CloudMD it's like... It's a markdown file.

00:52:12: That If You Are Using Cloud Code For Coding ,it will be sent every time you send a message to cloud in the background!

00:52:22: What I did with the AWS server, i set it up quickly on the interface and then use cloud code to configure the server in a secure way.

00:52:32: Because cloud code because you can't use a terminal just like your OpenClobule It can log into the server via the terminal.

00:52:41: You need keys for the proper place.

00:52:43: So what created here?

00:52:44: I open now the cloud.md.

00:52:47: It's basically template CloudMD that you can use in your project if you want to do it the way I did.

00:52:55: So this will explain for cloud code, the SSH access... Cloud will help set up the PM file.

00:53:03: so i don't even explain what's in this file and he'll explain it.

00:53:09: And then Cloud will also guide through which server needs to be selected or installed on that AWS Server when setting it up.

00:53:16: just have a chat about.

00:53:17: basically In this cloud.md file, I mostly try to put everything that's I already use and it's important And the one of the biggest things That i think is most value can share right now Is OpenClaw command map.

00:53:34: What did?

00:53:34: I scraped the whole GitHub repository from openclaw And I scrapped all commands you can run on the server to set up certain things.

00:53:44: So like changing models, logging into certain models I have here for example open claw channels list.

00:53:50: that's the command you can run in the server terminal window and then it will do something.

00:53:56: but this is not for you!

00:53:57: This is for cloud code.

00:53:59: so if we go back quickly to the clod.md That's the thing that every time he sent to Cloudcode there is a line at end of it, openclaw commands.

00:54:10: See OpenClawCommands.md for the full CLI reference.

00:54:15: so when you are talking with Claude and say to Claude hey I want switch models in my OpenClaude agent can do that for me yeah Claude then will check command map.

00:54:27: he knows exactly which command needs to run on server if we make changes enter the server every time, write certain commands because trust me that's a terrible thing if you are not familiar with using terminal.

00:54:41: So yeah!

00:54:41: That is one of those things I also have set up guide.

00:54:44: this probably where we need to start.

00:54:46: just give it your AI assistant which i recommend using cloud code as well for setting up the server and its one of best things at the moment.

00:54:58: But it's another conversation, but back to the set of files.

00:55:01: You just give this to Claude and he will walk you through that whole setup.

00:55:05: how you set this up on AWS On The Free Tier which storage you need to select Which instance?

00:55:12: You need to Select How many ramps And things like That.

00:55:16: if your don't know Don't have an idea He Will guide you Through and If You don't want To use a WS I'll Set It.

00:55:21: Other Options or Such as Digital Ocean and other providers.

00:55:26: I also saw yesterday that Hostinger, actually they made a service for this.

00:55:33: so if you search on Hostinger Open Claw They have one click setup For i think seven euros per month.

00:55:41: So just buy it And then create the server.

00:55:44: where openclaw is there You need to configure after that.

00:55:47: It's secure probably fast way But then you already need to pay for that.

00:55:54: and this what I show.

00:55:55: You it's free if you do it with AWS And the rest of the file is just for Claude explaining What he needs to do.

00:56:02: once?

00:56:02: He gets into the server.

00:56:04: basically, they instructions are not even for you And i have here in This file.

00:56:11: i think i have a small part Of there.

00:56:16: yeah It's a small snippet of the soul dot md That i was mentioning at the beginning.

00:56:21: But this has the core values, personality traits and basically what's in the sole MD.

00:56:26: It is like explaining to an open-claw agent who she or he is identity and things like that.

00:56:33: And a nice thing about the SoleMD.

00:56:35: This is the core identity but it is editable by the agent himself.

00:56:38: So these will improve over time based on conversation set you are doing.

00:56:48: The last thing in the repository is a logger.

00:56:50: That's also explained for Claude.

00:56:53: This is just that Claude, when you are configuring the server using Claude code he will make here logs what things have been changed on the server.

00:57:03: it's like a memory of Claude Code.

00:57:04: so I created this project in my device which is called Liara and thats me using Claud code to configure the server every time i need to change on the Server Once the server is up & running.

00:57:17: It sounds really complicated but Trust me, you can do it in like... I don't know.

00:57:22: I did it for about fifteen minutes using Cloud Code.

00:57:25: Once the server is up and running And the OpenCloud agent is live.

00:57:30: It's living.

00:57:30: You can talk with it on Discord.

00:57:33: After that The whole setup happens when talking to it.

00:57:36: So we just say I want you to be able do this and then openclaw, we just fix it.

00:57:44: He will maybe ask a few things like... You'll say hey i want you can browse on the internet And he would come back OKi need an API key for Brave and the rest.

00:57:54: I fixed The API keys.

00:57:57: don't directly give them to the OpenClaw so not copy pasted When you have to add an API Key To the Open Claw agent.

00:58:05: Then you go to your cloud code setup and then you tell Claude, hey I need this server a new API key.

00:58:12: in the instructions he knows how to do that.

00:58:15: Then Claude will give you command so you don't have to run yourself.

00:58:19: also don't give your API key to Claude himself.

00:58:22: Don't just copy paste it into the interface.

00:58:25: Claude gives you commands which ask Claude how to do that, it will help you.

00:58:30: And then we just run the terminal command and after the terminal commands You just paste the API key.

00:58:35: now once your on that The API key will be added To the ENV file in the server which is an open clock can use.

00:58:43: That's security

00:58:46: more or less nice.

00:58:49: Um, it's very good that we have this repository.

00:58:52: That we can share with people and thanks so much.

00:58:54: um that you uh build it actually for us to share.

00:58:59: I highly appreciated what i wanted to ask you because I feel like So I feel the sense of excitement You Have About It.

00:59:08: Yeah And Drawing back on your experience you had when you saw the launch of chat GBT and that feeling off Monday morning going into the office talking to people.

00:59:19: Hey, have you already seen it?

00:59:20: blah blah blah.

00:59:22: How would you rate?

00:59:23: And I know since this inception moment You have been deeper and deeper in deeper with AI.

00:59:29: how Would you rate this moment now with open claw like In comparison to that and everything that happened before?

00:59:37: yeah If that moment was a ten, I think it wasn't then back in twenty-twenty three So I was super excited about it.

00:59:47: Then i would say this opens law.

00:59:48: It's like probably seven half eight ish on the whole scale?

00:59:56: I Think I will go quickly to nine once I have the whole Mac mini set up properly.

01:00:03: So I might change my score after that, but just based on the early experiments i made.

01:00:08: I would say seven and a half eight which is quite high because The last tool that made me give such high school was the launch of nano banana pro And there was a big one as well.

01:00:21: so yeah it's...I think this Is probably really I wouldn't say Probably it's revolutionary in a way that I think the bigger AI platforms such as OpenAI and Tropic, they saw now.

01:00:40: okay there is need for this.

01:00:42: And if we make something like this in a secure way then... That's key to lot of other new users because the platform currently can't do something like these but they have their strongest models.

01:00:55: so imagine running Lyara natively built by entropic on an opus, four point six.

01:01:00: That would be like absolutely bonkers.

01:01:06: I'm not sure if they can make it because there are a lot of security problems with open claw at the moment.

01:01:11: so probably that's why They didn't do it Because for them is Like A big risk If they mess up something but At some point they will figure It out how to Do it properly and then we have proper AI assistance that We Can use By The native model creators.

01:01:25: And that's Probably biggest that we will have.

01:01:28: again, this would be the biggest thing at that point.

01:01:32: Because thats what I wanted to ask.

01:01:36: if i remember the chat GPT it was for second half of twenty five strategy document which got leaked in a legal process with Google.

01:01:54: They actually said the division for Cheshire PT is to become this super assistant.

01:01:59: And I feel like so, The more I listen to you talking about it... ...I feel like This Is The Closest We Got To Having The Super Assistant Now.

01:02:07: So You Feel Like Security Is The Biggest Problem?

01:02:11: So This Coming From The Community Basically As An Open Source Project.

01:02:15: Yeah

01:02:16: It's Due To An Individual Creator.

01:02:19: The Project

01:02:19: Was Vibecoded.

01:02:22: So the guy who created Open Cloud, basically he said it in an interview.

01:02:26: I have no idea what every file is in there so i was just talking to... He created it with Codex and some of it with Cloudcode.

01:02:34: He was just taking his laptop when he was programming.

01:02:37: it then started work.

01:02:39: from that it grew.

01:02:41: but you can see as a Wipecrow project.

01:02:44: this really good But yeah.. There are some security stuff.

01:02:51: You have to be careful how you use it, and I think that should be a totally different episode as well.

01:02:56: But i'm not sure if we have time or I can go like quickly five minutes into our security

01:03:00: course.

01:03:01: Let's do please.

01:03:01: yeah

01:03:02: um Be careful what?

01:03:05: You give access to your open claw If you set it up at the beginning.

01:03:09: so if you do It on a separate device then you are more or less good in A way That you know What's On That Device And She has Access Only To That Things.

01:03:19: Don't sign up your agent to platforms where other agents are.

01:03:24: One of the platforms that got really popular is The Mold Book, That's like Reddit for all these open-claw agents people created and thats number one most unsecured place.

01:03:38: you can give your agent access because people do prompt injections from there.

01:03:44: so then its bad.

01:03:48: obviously.

01:03:50: For example, I wouldn't recommend allowing your agent emailing people other than you or talking to others at the beginning because if someone knows what they are doing then it can also do prompt injections.

01:04:02: If somebody sends an e-mail now with a prompt injection through Lyara It will probably go though and she could leak my API keys.

01:04:11: so i would not want that.

01:04:13: So at the beginning, just keep it on your device.

01:04:15: If you are running it in a VPS or private server make sure that the server is secured and after then give access to tools which first want for use.

01:04:26: I would say using Discord at the begining.

01:04:28: super easy to setup.

01:04:30: Telegram also super easy.

01:04:31: set up what i heard but didn't test it.

01:04:33: But shouldn't be hard And start talking with.

01:04:39: Make sure you have one clear use case in your mind what you want to test and just give access To the tools at the beginning.

01:04:45: that that you won't do on to us And nothing more.

01:04:48: be mindful how you gave access via API keys.

01:04:51: create separate API keys with clear budget.

01:04:55: So if for example, If you used nano banana pro so you want your assistant.

01:05:00: Just like I did two create images Create a new project on Google console For that set up budget alert of ten euros over there and an instant kill if the budget is depleted.

01:05:13: And create a separate API, obviously so you can take precautions like this.

01:05:17: If something goes wrong she burns love credits The most you lose.

01:05:21: it's a tenner that nobody gonna have a headache about But if you don't do it then she burns one thousand.

01:05:28: You might have different conversation at home

01:05:32: Maybe depending on which account is charged.

01:05:35: Exactly, exactly.

01:05:37: I know my wife wouldn't be happy.

01:05:39: Yeah

01:05:39: these are very important additions.

01:05:41: thanks so much for that.

01:05:42: i guess Are there any resources That you draw when You think about this?

01:05:47: So Is There Anything Like URAD like a blog post or A post?

01:05:53: We can also maybe put in the show notes That have some helpful security guidelines Because I could imagine people thinking Okay, yeah.

01:06:02: That sounds really good but wait a second.

01:06:04: what was it?

01:06:05: again and I know how people are they probably won't re-listen the whole part.

01:06:10: so is there anything that we might be able to share?

01:06:14: Yeah obviously their own website is really good.

01:06:17: They have a lot of docs.

01:06:19: So openclaw.ai i think its domain or com i'm not sure.

01:06:28: There's alot of good stuff about set up and other things that can be helpful.

01:06:36: at the beginning, I recommend also checking a YouTuber called... I don't want to butcher his name.

01:06:43: Let me just quickly look it up

01:06:45: because

01:06:47: i'm subscribed And he should be on top of this video yesterday evening.

01:06:53: Matthew Berman He created lot's good videos how to setup.

01:06:59: So there is also a lot of good content and he has lots links in those videos.

01:07:03: I can share maybe link with you the video that i've watched, then it will help as well.

01:07:08: And these are mostly like setting up security.

01:07:12: if you want to get inspired on how to use Openclaw or what u could do with openclaw Then follow Alex Finn on Twitter.

01:07:20: The guy already bought two Mac Studios.

01:07:24: That's a ten thousand euro device.

01:07:28: He bought two of them to run OpenCloud.

01:07:30: The guy is creating an army of OpenCloud agents, he's doing a lot things... I think fifty percent his hype just generate content because he was the content creator.

01:07:39: so probably there are lots like that.

01:07:41: but it really inspires you.

01:07:43: open your mind and see certain use cases which can be done with your OpenCloud.

01:07:47: So i wish for also link in this profile so we could put into show notes.

01:07:52: Wow nice thanks.

01:07:54: very much.

01:07:58: wrapping up the conversation a little bit.

01:08:00: I always want to ensure that people will leave the episode with most actionable advice, and i think we already delivered on that like three times.

01:08:10: so thanks for much of it!

01:08:13: But... We touched getting started with AI, how do you think about structuring problems?

01:08:21: which route to go?

01:08:22: So is there problem they can solve basically in the UI of chat GPT, et cetera.

01:08:27: Can I solve it with cloud projects?

01:08:29: Custom GPT etc.

01:08:30: do an agent?

01:08:31: we talked about Cloudbot setting out on VPS Mac mini so we covered a lot.

01:08:38: now if there are people listening that are thinking a lot about AI but they're still early in their journey like the early.

01:08:48: okay yeah i have built a custom GPT.

01:08:50: But other than that and just prompting for example, for stuff.

01:08:55: What would be your maybe two to three top pieces of advice.

01:09:01: actionable advice?

01:09:02: For people that want to take their next step at least towards something That yeah you're building.

01:09:10: for example

01:09:11: Yeah number one I will say start the coding project very.

01:09:19: use AI.

01:09:21: You will learn so much about how these models work and what are their capabilities.

01:09:25: And everything you do in code, it can be translated later to other use cases... ...you don't have to understand the code that you're writing.

01:09:32: just Use AI for these things So you can understand How To Talk With Them.

01:09:37: That's What You Will Learn By Creating!

01:09:39: You Don't Have To Recreate The New Facebook.

01:09:42: Do Something Small.

01:09:44: I can't believe, i think everyone has like an idea or something for a app.

01:09:50: Maybe you have problem at home that you are frustrated about?

01:09:54: How I started?

01:09:55: in actual example my very first project that I coded with the AI was a gym tracker because all the gym trackers out there were rubbish but not from my use case and program which is following.

01:10:07: So I created it on afternoon.

01:10:09: back then It was like, feels like it was ten years ago.

01:10:16: And that teach me so much about how to talk with these.

01:10:19: what are limitations?

01:10:20: things like this and its really something simple you can set up quite fast with AI If your idea is not as simple.

01:10:27: or not ask AI Take a microphone So don't have type.

01:10:34: Install handy.computer.

01:10:36: I can share the link.

01:10:37: That's free transcriber.

01:10:42: It's awesome, but you have to pay for it.

01:10:44: And if only need a transcriber then handy.computer is an app that you just install and runs locally and completely free.

01:10:51: Then turn it on Talk for ten minutes into your microphone about the idea Send it to Claude, Gemini, ChagePT whatever.

01:11:02: Ask this tool okay Is difficult or not to build?

01:11:06: If get like plan of more less than ideal Okay This looks manageable.

01:11:11: Install Cloud Code or Cursor.

01:11:14: I recommend if you are doing a coding project, don't do it in the chat interface.

01:11:17: It's gonna be lot harder and lose motivation.

01:11:21: so install cursor Or get cloud code.

01:11:25: that probably would be the wiser option.

01:11:31: Anti-gravity is also really good option And i think its free.

01:11:34: So If You Don't Want To Spend Money On This Then Use Anti-Gravity From Google zero pro there for free, if I'm not mistaken.

01:11:46: For a certain period of time and you can create probably an ice project And you understand most of things.

01:11:51: You don't have to understand like code and things Like that but you will understand A lot about how these works context windows Things like That because you Will encounter a Lot faster over There These issues then in just the chat interface?

01:12:04: Those things are playing everywhere when you Are working with they are.

01:12:07: i think This is one of the biggest things that you can take as next step.

01:12:11: And, another thing which should already be done and should stay doing write everything down if it's in your head doesn't exist.

01:12:20: So If You Need Help from an AI Assistant Doesn't Matter Which Tool About Your Fitness Course Personal Projects?

01:12:29: If The AI Doesn't Know What Is Your Current Status?

01:12:31: What Are You Doing At The Moment?

01:12:32: How The Heck Do You Expect To Get A Proper Tip From AI?

01:12:38: When I realized this, I started to track everything.

01:12:41: Also a gym nerd.

01:12:42: so... ...I created the tool and tracked daily.

01:12:45: what do i eat?

01:12:46: What's my weight?

01:12:48: And My AI is connected via API.

01:12:51: So when ask Claude can you get data from my fitness and check it for my goals for twenty-twenty six?

01:12:56: Am I on track or not?

01:12:57: He knows exactly where am at right now.

01:13:02: but if u don't have this context just get slopped.

01:13:05: Write everything down.

01:13:07: Write projects done that you are working on challenges.

01:13:10: That's your face with the project Everything basically.

01:13:14: so what I use its Apple notes just because it's accessible then from all my devices and i have a folder for each Project, but i'm working on an also areas of My life.

01:13:25: if You read The book build A second Brain From Tiago Forte Then?

01:13:30: Listeners That Read The Book Probably Understand What i'm Talking About.

01:13:33: It's like A system Where you Take Notes.

01:13:35: in Certain Way it helped a lot also with AI stuff.

01:13:39: So I work on the project, i write down everything.

01:13:41: what works and doesn't.

01:13:42: What are the challenges?

01:13:43: And when I ask Claude about that he knows the whole context...I don´t have to explain things so He know's what I already checked or did.

01:13:51: It is alot easier to write these things as you go.

01:13:57: You get proper stuff out of this AI tools But if its only in your head then it´s nowhere.

01:14:07: Have you always been like an explorer type person?

01:14:11: Like there's a new thing and I'm intrigued.

01:14:35: So one day we just sold everything and we stepped into the car, then drove to The Netherlands.

01:14:41: To start a new life!

01:14:43: I only knew where i was going because my cousin was living here And he said you can live for ONE month at us... I had no idea when im gonna work!

01:14:52: Three days later I had to work.

01:14:53: so yeah.. My whole life was exploring Jeff trying things.

01:14:58: Yeah, I feel like that's a trait that lot of people share on the frontier off these technological thing because they don't wait for someone else to show them away but actually find their way themselves.

01:15:13: so your personal story with you moving to another lands is perfect testament to it.

01:15:20: couldn't be better example.

01:15:22: Yeah, I'm not afraid making mistakes.

01:15:24: Breaking things and you will do a lot when you are working with AI.

01:15:27: so... ...I see a lot of colleagues doing something.

01:15:29: then it breaks doesn't work.

01:15:31: they come to panic and i don´t know wow!

01:15:33: I made some mistake.

01:15:34: like I broke only fourteen things today that's no bad.

01:15:37: yeah yesterday is sixty.

01:15:39: Was was hard to get buy-in from the agency leadership?

01:15:44: That in future would be normal if things break?

01:15:47: Happily not.

01:15:49: So they are really open-minded.

01:15:51: They understood what this brings, I explained it at the beginning... ...they already knew me.

01:15:55: so i broke a lot of things before that!

01:15:59: It was easy for them to accept.

01:16:02: okay This probably just will give more problems when offered is unleashed But also gave a lot benefits..They said it's not my words, they said it too me.

01:16:16: Happily it wasn't easy.

01:16:17: get on board

01:16:19: Very cool.

01:16:20: I've got to have a final question for today, which is what?

01:16:24: didn't we talk about that?

01:16:25: We should have talked about.

01:16:29: um i think we covered most of it.

01:16:34: but i would say one important thing Which i saw a few days ago on twitter.

01:16:41: someone wrote an article That got like seventy something million impressions About the new coveties here and its ai And then got love attention.

01:16:50: and What i hate at the moment on whole social media is the fear mongering about AI.

01:16:58: And I saw someone else reacting to that post with one line and it was brilliant, he said A.I doesn't come for your job Fear is coming for you.

01:17:07: so if You are just sitting still and won't do anything then close-minded Then don't test these tools and not open using them.

01:17:16: yet Most of the people are not like that.

01:17:24: Some... People, like me or explorers.

01:17:28: we're way ahead but everyone will catch up at some point.

01:17:31: and if you look around in real world how fast things get implemented How open-minded our boards to implement AI.

01:17:41: When you go into a real company with rules and corporate stuff It's not easy for an AI project to be running.

01:17:48: You see all these guys on LinkedIn talking about it like explorers, they have.

01:17:53: this is their job.

01:17:54: Most of them are also content creators who need to talk about something and generate impressions.

01:17:59: but I would say don't be afraid...I'm a positive person!

01:18:04: Really nice things coming towards us.

01:18:08: we will solve so complex things that were not able to solve before And it'll unlock a lot of things.

01:18:13: probably for some people bring negative thing.

01:18:18: But yeah just dont let fear disappoint you, or don't get overwhelmed by the fear posts and things like that.

01:18:26: And I'm already behind them not going to try

01:18:27: it.".

01:18:28: That's the worst you can do!

01:18:29: So just be open-minded.

01:18:31: ignore everything.

01:18:32: go try yourself.

01:18:33: see what he can do in your field In your area for yours before.

01:18:37: You will understand a lot better than reading on Twitter when they say someone Just or LinkedIn?

01:18:43: See some of us talking about fear and things Like that.

01:18:45: yeah close.

01:18:45: that's designed To Close LinkedIn and Go Try It Yourself.

01:18:51: Very well put.

01:18:52: Thanks so much, um I very much enjoyed the conversation was very insightful.

01:18:58: i like literally have this Mac mini here standing in front of me on my desk and

01:19:03: the podcast and install it.

01:19:04: yeah

01:19:05: The more we talk.

01:19:06: unfortunately I only have time on the weekend but the more We talked the more I was like craving going to because It felt like this moment when, I don't know.

01:19:16: When i was a teenager and had this new computer game And uh...I was literally sitting in school very nervous about getting home installing it setting everything up or just going.

01:19:30: Oh!

01:19:30: When I have the new computer or something even crazier Like oh can I install my graphic card etc?

01:19:39: So thanks so much for giving me that.

01:19:41: Thanks

01:19:41: for your chance.

01:19:43: Yeah, it was great.

01:19:44: And if people want to follow you around so obviously we will put all the links that I talked about in the description.

01:19:50: but what's best place for following?

01:19:51: because i know they also share a lot of your thoughts and projects you are working on.

01:19:56: LinkedIn is nice way to follow me.

01:19:58: If You Want To See A Bit More Filter Than Normal Offer Then If You Want To Get A More Raw Experience Than Twitter Is The Place like different content on Twitter than LinkedIn.

01:20:12: But if you follow me and LinkedIn, that's already nice.

01:20:16: there also shared a lot of things.

01:20:17: when I have time obviously it is not daily.

01:20:21: so

01:20:22: yeah hopefully Liara can make more time in your day So that you could post.

01:20:27: Yeah

01:20:27: would be nice.

01:20:28: the problem is If she makes some times then i already come up with new project which will take time again.

01:20:33: but yeah thats life.

01:20:35: That s actually The New Struggle.

01:20:38: You Make Time And Then You have more time, but then you immediately start working on something new.

01:20:44: Yeah exactly!

01:20:46: Cool.

01:20:47: so then Alfred thanks so much.

01:20:50: very much enjoyed it.

01:20:52: all the best for you and for advice for backdrop boost.

01:20:57: I hope you will get to the, what was it?

01:20:59: Four hundred euros

01:21:01: per hour.

01:21:01: Yeah small numbers.

01:21:02: so

01:21:03: yeah no worries i don't think that always has to be a shiny big number.

01:21:07: So um wish you much success in that.

01:21:10: and uh yeah thanks so much from making good time today!

01:21:14: Thanks for inviting me and same for all of your successes as well.

01:21:18: Hop into the ride!

01:21:22: Bye bye.