All right.

Welcome.

Well, bad stuff.

We think episode 26 so I haven't checked but it sounds right.

26 Big, Big Yeah 26.

Same jokes, new new material.

We've got.

Yeah, very busy week.

So I've had, I've had both people visiting and the end of Sovereign engineering coming up.

So it's been, it's been pretty hectic.

There's a couple of people here that I still wanted to get out on a podcast walk or something and I just haven't yet.

So you never know.

Maybe I can grab some like 30 minute or 10.

I'll take my microphones.

I reckon maybe I'll get some 20 minute like little stints with a few people.

Couple of quick episodes.

It's nice.

Yeah.

Well, today might be a quick episode, so there's a lot, There's a lot happening.

We're a bit later in the week.

We haven't hit our our usual Wednesday or Thursday release.

It's getting Friday almost like so that's OK.

I think in like 2 weeks I'll be we'll be back in the man.

Back in the van, back at the beach, Good weather.

Well, I mean, you've been getting good.

Weather The weather looks nice.

All right, some good weather and.

Last night and I was like, nice looks, looks like it's a good welcome home sort of spring weather, warm in the day, cold in the evening, just the same on it.

This is the the kind of content you can only get on the good stuff.

What do you want to talk about?

Has where.

The checks were like two different.

Do you know what?

I can't remember when we talked about this.

I feel like it might have been last week.

It might have been like two or three weeks ago.

I can't remember.

But I said I was divorcing Claude code.

Well, I didn't divorce Claude code.

I did step down my plan, but then I realized that it's actually pretty good at doing non coding related stuff.

And I, I feel like maybe this doesn't come as like any great kind of insight to a lot of people that play around with Claude code.

But yeah, I, I kind of just been, I used it to completely like reorganize my financial existence.

So I basically had receipts, I had transaction data, like I had all sorts of stuff that was just sitting in different places on my computer.

And yeah, it just helped me collate all of it into one place, organize it, set up little like, dashboards to track where my spending is.

I spend an ungodly amount on coffee, which we already knew.

But when you see the number next to it, you're just like, oh, shit.

Like, this is actually quite a lot of money.

And yeah, got this like completely streamlined to the point where it was like a very logical next step to just go and prepare my attacks for this year.

So it's done that as well.

And it smashed it out in in no time.

Like, yeah, I would have that whole exercise would have been like 10 minutes.

And so I was like, OK, what else can I turn this to?

I got like a little script written.

So I had all sorts of notes written in Apple notes that you just collect.

And like a lot of time my Apple notes will just have like little, I'll make a note of things that I want to do or need to do at some stage.

And then I don't revisit it because it's lost in my Apple notes.

And then I've added another 10 notes and it's gone forever.

But yeah, I just basically was able to extract everything, pull it into like pull it in by this script and then quote code just went through and extracted things that I actually said I was going to do, build a little To Do List for me in a tracker.

And then just categorized all of my notes into into like a more efficient filing system that I've then popped into Obsidian because I actually do use Obsidian for like more structured note taking.

And Apple notes is just like the everyday stuff.

Like I, I put a note in there like 2 months ago going I need to get my car serviced and put like 3 different places that I was going to look into to get a service, which is never did forgot all about it.

Yeah.

So it's like it pulled those out and then like I was able to do like a quick search on those 3 and like check for like Google reviews and stuff like that.

And this rank ranked and categorized it for me.

And I was like, oh that's really cool.

I didn't ask you to do that, but that's good.

So now I need to book in my car service.

The next logical step would be if it could book it in for me.

Yeah, we didn't quite get that far, but yeah, it got me thinking.

It's like there's a whole heap of like non coding applications that you can just play around with.

And yeah, it's quite interesting.

So another thing that I did was I recently got a this like pretty comprehensive like DNA check done.

So it's like 100 pages of all of these different markers.

And I was really curious to kind of like be able to sit with this document.

I opened up the document.

Like I were you human and none of this what percentage human did you manage to make?

Yeah, I'm well, you get.

You got over the wall like.

Surprisingly.

OK, surprisingly, I've passed the test somehow, but it's a, yeah, it's 100 page document.

And I was just, I started looking at it going, well, I, this is a, there's a lot of stuff in here.

I don't really know how to interpret it.

But then how to take one section of the test and hold that in comparison with another and how to interpret it meaningfully.

Like the, the main objective was I really wanted to build like a diet that was tailored to whatever was in that test.

And I have no idea what was in it.

So I used quote code for that.

It's actually quite good.

So yeah, I just, it just got me thinking a lot about like non coding applications for for chord code.

And I don't quite know why, but it feels like it's more efficient at doing this stuff than just using the app.

And maybe it's just that a bunch of limits are like not being applied to chord code in the same way that the app is restricted.

I'm not sure.

But I just felt like I could run the same experiment through the app through it's nowhere nearest good.

Yeah, so which also actually has the same models, but I think there's the the feel I get from it.

It's hard to know because it's all hidden from me, right?

But I think they, they put a lot more effort into making it terse and quick in every judgement that it makes.

And so, but to have like more turns through the loop.

So you you generally have that agent loop of saying like, OK, well, what, what do I want to do?

What's my plan?

What's the next tool that I'll call or do I respond?

I think it's right and I, I think in the desktop, I think they have similar sort of aging, but it's kind of, it's almost like it's set up to be more interactive.

So they say like, you know, think a bit and then just answer, maybe do a couple of couple of goes around the loop where it's called code.

I think because it's supposed to sort of run on its own in ACLI, you're not really interacting with it in the same way.

You just want to like kick it off and like vaguely monitor it, but not, not really interfere.

It's it's told to just be like, you know, be quick, give me short updates and then just do stuff and keep going.

So I think it's probably in that sort of system setup that they've just given it that slightly different persons.

It is where I know what you mean.

Like they act very differently in each of those situations.

So it clearly is in the application prompts that they're using now that it's.

Yeah, that's interesting.

I've not really tried it for much outside K.

I'm so annoyed that Claude tastes like a Yeah, that's terrible that that.

Was it?

It was more just it was more just like a last ditch attempt to trying to work out if this thing was worth keeping.

Because I do find Claude Desktop useful at times.

Like it will kind of double the search engine for me and if I want to just like bounce around ideas for on, you know, on the product side as well, it's quite good at that.

So I, I kind of just downgraded to the $20.00 a month, I suspect.

What else?

Can I do come back anyway as well?

Within three months?

Codecs will be crap, codecode will be good again.

It's just they just go backwards and forwards, I mean.

Completely so I.

Mean I've I've been doing similar sort of non coding tasks via wingman.

It's kind of a point of that one Goose agent, but I, I have on my list like a different agent stack that will allow you to then flip between sort of four or five different agent systems.

So you could have it set up with Gemini, with glow code, with Goose, with codecs and in the background.

I think I might make a newer version of Wingman that uses that as the, the sort of brain behind the system so that when we do, you can like maybe have like we can abstract away like the recipe type thing.

So you can always set up the right MCP servers, You can always set up the right system prompts and everything else to kick stuff off.

And I think that would be quite powerful because then you can just pick and choose like not only the models that you want to use, but you know, which is the best agent system right now.

So I think, but I, I think that's like, that's the huge one for like the small business type use case of yeah, you don't necessarily want, you need an agent to be quite flexible.

So when I was rebuilding Beacon last week, I talked about it as like, you know, for the like the deep research tasks or anything where I need to look stuff up or do web searches.

Your previous versions, I'd kind of had to hard code in different loops for that.

Like I'd have different pipelines that would do web searches and different pipelines that would do research and you had to be a bit more exact about what you were trying to do.

Whereas now I just got anything that looks like it's more complex and you can answer straight away.

You just give to wing men and it just fires like a Wingman session in CLI and it will figure out like, OK, do I need to use the web search?

Do I need to make a list of stuff?

Like do I need to go read any web pages?

And then it just completes and sends it back.

So it's actually like Wingman in the background for Beacon if you start doing complex stuff.

But then the beauty of that, then you get access to all the different MCP tools that you could have.

I haven't really.

Yeah, exactly.

Having like wing men, a lot of wing men.

So a wing man, a wing man call is really just like a local API that I've got available.

But I can have that the response from wing man back to Beacon is just a local API call.

Once again, I've got I just run the two things locally to each other in our lives and to integrate them less and easily.

But I can also do this where you have a wing man and his job is to figure out which wing man can help you best.

So now you and while you wing man, it's like, OK, well, I'll give you this.

I'll give you the list of like API, like the different recipe IDs and what they do.

And you just choose which one's going to do it best.

And then it will figure out what to do next and respond back to Beacon.

So I think so it's, this is going to make it very quick and easy to add lots of seemingly complex functionality into something that's just super easy.

But again, like I reckon if I can play around with the different AI models and different agent systems for them can just get better, better results.

Yeah, that's cool.

Yeah, finding alignments to different types of tasks where models perform better across different like task categories would be interesting as well.

You know what I was one of the things I've been like really surprised at how good Claude Codes been is just analyzing data and then being able to like report out key findings across that as well.

So I've been using it for like investment decisioning as well.

If I want to do comparisons across, you know, two or three different investment targets in the past, like what I had done is I just had folders on my on my MacBook with just a heap of downloads and well, obviously if you pass through those manually.

So if you do obviously want to do the work, but it's just a really good, efficient way of getting across all sorts of disparate data, you know, long form text financials, and then just being able to spit that out into a, into a report or a memo to get across it quite quickly is good.

Just top line.

And then you can make decisions as to, you know, how much deeper you want to go into it.

But I but I also turned this on our like podcast thumbnails as well for titles and it did review of like all of our YouTube data to work out like, oh, you've got really good click through rates.

But you know, like definitely some of the text titles that we use in thumbnails like perform way better than others.

And so I'm going to start using that as a bit of a guide to improve.

The we're definitely.

Yeah, the thumbnail text as well, stuff that we don't really.

Think about but maybe like, good to have somebody.

It's just like an afterthought mostly.

But that's the thing, you could probably run this on the loop and just, I mean, get it to keep.

You certainly see it often.

Different channels that I would follow where early on the thumbnail is this and then it's this and then it stays as this one.

It's kind of clear that they've probably AB tested a few different thumbnails, figured out which was working, and then it loaded it.

I imagine that's just built into some software.

Some, yeah.

You can do it in YouTube, which I discovered this week.

So you can actually like just add multiple thumbnails then it will it will do the AB.

Test we have this like slight rest so that's quite caring too much about YouTube in the sense that I'm level uploads it's interesting yeah I had AI had a hit post on primal this week, which I was very proud of but top the trending as I yes made it like a you know a joke to Gigi it's like as I'm as almost as famous as you when you say good morning now but it was I mean the wife said oh you know maybe you should put it on LinkedIn and I was like or not but I don't know the kind of stuff I don't think it's a subject matter like what this is I don't think anybody will that's.

Not the place or like.

It's.

I mean was she was she saying to repost it on LinkedIn in its current form or was she saying you should post?

I'm proud and thrilled to have hit the 24 hour nobody trending on on price.

I reckon the latter would do better on things.

Something about how like, oh, you know, it's been an honor and a privilege to be part of something.

But Yep, you know, like, it's hard when you're doing something new.

And something about how somebody else can jump in and say that they're also really passionate about something, something just horrible that I don't want to do.

I'm I'm I'm thrilled to announce I went viral this week.

Probably the post that performed the best on LinkedIn.

Yeah, I think ideally you want to get to a point where you can just like close LinkedIn forever.

That would be the what was interesting.

That's the objective 2.

Days later, like I opened Twitter to look something up, I realized that somebody else had reposted it on Twitter.

It never even occurred to me to post it onto Twitter.

Oh, really?

Oh, right.

Yeah, I'm beginning to forget about it, which is good.

I mean, I think the the one thing that nostril is so good for is not pulling me back to social media every 5 minutes.

Like I just never feel compelled.

Like I'll use it, I'll be able to get some content out, have a quick skin down most recent posts and I'm I'm done in 5 minutes and I'm like, oh, that was actually really good.

That was quite interesting content.

And then I'm good.

I don't feel the like the, the dopamine hits that I need to come back again in, in 10 minutes.

And so that's been that's been amazing.

That's one of the things I love about about Nosta.

But yeah, very nice viral moment for you this week.

No, it was good, but well done.

It was some.

It was interesting because the we have these different captains every week at Sovereign Engineering, and the one this week was just like, right, we need to like, you know, we keep building all this stuff.

It's really good.

It's really exciting when you're here.

Nobody knows anything about what we're doing, so we talk about it.

Yeah.

One of the things you can do is you can just make videos of stuff voice over, like a little clip in the corner.

We should do that.

And then we should tell people that this stuff exists and then other people know it exists, and then I'll use them.

So OK, yeah, that's good.

So I think I already had.

The payment debt for anyone?

Done like I had my whole onboarding and payment thing that I think I'd sent you and I was like, you know, I'll cut this down.

So it's just the the key moment the hockey is it's paid.

It's lightning payments and Bitcoin inside WhatsApp and I'll I'll just ignore my urge to tell you like the five other things that I was in Kharif and cooler than that, but I'll just tease it and just say like there's one single hook.

I think now we can put Bitcoin into WhatsApp and we don't have to ask and then we can do it anyway.

It's good.

Yeah.

I think seeing it as well, like visually occurring inside of WhatsApp, that was like, that was a moment, you know, like when I first watched it, I was just like, oh holy shit, like that is that.

Looks.

It looks so.

Clean and very cool.

We do the onboarding, maybe I'll record the onboarding one later because I've just rebuilt that whole thing.

So I had to re remake the whole gateway architecture for it this week.

I'll tell you about earlier where it was to make it easier for other networks.

There's still a few, what should we say like loose ends to tie up to make sure it all works really seamlessly across different networks.

But but yeah, it's quite nice now that you can, there's a very sweet on boarding flow.

I've got built a new tool this week that will let me do assigned wallets easier as well.

So instead, if you haven't bring your, you can bring your own wallet or you can just get that generated on the fly.

So so something like that, I think would be pretty useful as well, wouldn't it?

So that the onboarding would just be, do you have a wallet?

No, his wallet, right?

Brain's going to be in touch.

Brain gets in touch and you're like, why?

All I did was say hello and that's the onboarding process.

And now I'm going to bank.

Like what the fuck?

That's pretty cool.

Yeah, it's amazing.

So I think that'll be that'd be sweet.

Then we can just harden it up a little bit and then maybe just get it in the field, some prototype, but.

I was, I was going to say, I think the other thing is like, obviously that one post went viral, but I would just be chopping up a heap of content and pushing that out on both obviously Primal, but also Twitter.

I think it's, I think it's just so important that we need to just get that out there, get people talking about.

I mean it just takes.

Time right you got to think about it's even like that video last week like I already had a video and then it was case of like all right I need to yeah put the what this means sorry but it was I already had the video recorded and I was like all right well I need to sit here and cut it down and then I need to figure out like the and then I had to record yeah it's ducked out into this like little base kitchen type thing it's recorded like few versions but of the voice over he's like you know it's it's not perfect it's it's good it's good enough like it's fine like let's go get it out there it's.

Just yeah, I don't think it needs to be that Polish, but the point really.

It's still, it's not polished, still take like an hour or two and then you're like it's.

Still time consuming.

And that's the most obvious one.

Then you're like, right, Well, if you're thinking of it like it will, I'll have like 10 of these like what's the, what are the different hooks?

How do I get the right video?

Like they always just take time.

And at the moment I've just been on like a constant treadmill of like every week is you've got to do something new.

And all right, I'll, I'll probably I'll do this.

That's first.

Yeah.

Let's see.

Yeah, we'll get it on it.

Well, that that's coming to an end soon.

Right now this today.

Is the next week you guys are?

Big last day.

That's why I said I'll probably have to do a quick 1 today because I still haven't read my demo.

I was.

I was.

Mobile Vibe coding.

Yesterday I had this great screenshot from Claude where I was just like, I can't, I can't, I can't see any of this functionality.

There's this like little CLI wallet I was testing on my phone.

It's like I can't see any of this functionality.

It's not even in the list.

Like what is going on?

Like why?

What's this thing?

You just like lightning, but nothing's changed.

And he's like, yeah, you're right, There's no changes.

Like I didn't actually build any of it.

I just built a completely separate standalone demo.

That's why it doesn't show up.

So yeah, I haven't done this.

I haven't connected to your existing wallet.

I haven't done this.

I haven't done that.

Doesn't create any new sub wallets.

It doesn't create any of these.

Not So what connect strings we've been talking about doesn't do any of it.

In fact, I've just like not done anything.

I've been really bad, like fuck you.

And then I just went to codecs and it was, you know, almost straight away it was like, all right, here's the menu item.

OK, this bit doesn't work.

Why doesn't this work?

This one we just it's like again, it's like having a pair programmer.

But again, once again, like with codecs, IE yeah, I think the difference is that I just, I just get rid of the thinking mode.

It's like, I don't, I don't let it think.

I just say you're on your codecs model thinking mode low.

Don't think I will tell you what to do.

I want you to do that and I will test it and then we will itch rake and that works very nicely.

That's interesting.

Can't.

Stop it thinking it.

Just yeah, I've been going for it.

Yeah, that's really interesting because I tend to just do the opposite and it'll obviously sit there and be quite slow and really take its time.

I mean, I haven't noticed it kind of going off on one just yet.

It's been really good.

But yeah, that's I also think it's harder to work at.

All if it doesn't respond quickly, OK, just I don't know, maybe to go back on my earlier point and saying like, you know, code is designed to just run like, I don't think that's the ideal way of doing these things.

I I think you should treat them like little hackathon sessions.

It's like, you know, it's every time you sit down at a computer to build something, it should be like, we've got the pizza in and we're just going to do this.

Like right now, we're going to just, we're going to make this one change.

Then we're going to test it and we're going to see it there and then we're going to do it again and again and again.

You get such good momentum.

But also like as a, you know, as a product manager, you've got so much more intuition about what has been built, how it's been built, like exactly where the code is without having to go through then like review a bunch of code and figure it out after the fact.

You just dictate it as you go and that works really nicely fine.

It's been my work when I need.

To get some interesting.

There's also just the the vibe coding shitting out and there's a quick demo which is good.

So I also managed to create a couple of levels tied on field as well.

So now we have skydiving cams and we have like the man can get power up some different levels and there's like 1, he gets laser eyes and he shoots the carriers out of existence.

That may come back like a couple of seconds later.

It's good, it's very playable.

Got a couple of couple of young playlights in the end of ideas.

The the leaderboard was pretty hectic.

The numbers The numbers on the leaderboard were pretty.

Crazy.

A couple of people just vying for success.

Well, I've got almost.

Just loving it.

It's not.

Working yet?

So I need to test in depot, but it's like a Nip O 7 integration.

So then you can bring your own Noster pub key.

And then once you do that, then you can you can sort of keep the same session over multiple days and you can start to chip away and build your position on the leaderboard or come back to it because.

So I think the top two people on the leaderboard at the moment are me and Jesus.

And I mean, we both burn our keys.

I think #3 is Gigi, I think he's also burned his keys.

And you burn your case by refreshing or closing the browser.

So it's like everyone, like everyone's going to start again every few days.

But I reckon it'll be, I mean, people might be willing to put a bit more into it with the browser login.

Yeah, completely.

There's a good, I think we, I think we talked about this in more detail last week maybe was it last week's episode or the week?

I don't think it was the week before because that was Jeff, but yeah, so we've we've touched on a bunch of like different topics in this episode where we've like actually gone into more detail in the past.

So I think for Beacon, that was last week.

So if anyone wants to dive into that in more detail, Pete goes does a does a deep dive in episode 25.

Did we also talk about retired in a field in 25?

Or maybe it was just like a fun game.

Yeah.

So it's a bit of, I mean the idea.

But it is.

A It was a fun game, you know, deliberately to learn two different concepts being like how do you do multi jurisdiction or multi hosting environments seamlessly inside one app?

So you have one app, but the app itself runs in two completely separate places and neither of those places have intimate knowledge about where they're hosted or any network connectivity between.

So can you do that over Noster?

And yes, you can.

So, and that was like something I was working towards for Beacon.

So it was almost like the last bit that I wanted to make sure I was comfortable with before I went back to Beacon.

Because it's very important in Beacon that the the brain and the ID server don't know each other.

Despite the fact that they come together to do that one service.

They should be two completely different people.

And you can be your own ID server, but you don't necessarily need to do all the hard work.

So somebody else could do the complex stuff.

You can just do the simple stuff and bring it yourself.

So I think it's just a good pattern for app development in general.

It's like bring your own wallet, bring your own database, bring your own identity, bring your own whatever.

And then just plug in the different apps.

You should have more stuff like this.

It's all there.

Like this is the, this is the other stuff, you know, this is what it's all about, some plugging, plugging things in.

It's not just really a shit posting social network.

It's just that bootstraps the identity.

It's a good first use case, but the other stuff is very interesting.

There's going to be a lot of good stuff coming out of it.

Very good.

How are the other what are the other demos like that you're that you're seeing over there?

There's some pretty.

Interesting stuff.

Like, I mean, I'm a big fan of the context VM stuff that keep talking about, mainly because I'm probably one of the like heavy uses of it currently.

Like, you know, vegan's dog can.

I mean, at least about four context.

We have service running inside one episode.

That's that's been good.

There's some interesting stuff around like search.

So Gigi had a very good search system called ants that he'd put together.

There was just like everything was semantic searches where you do a search and then like anything you can click on then triggers a new search, find stuff written nicely.

Like there's been a bunch of work around how you can make everything local first.

So the idea would be like inside your browser you have a reload or inside your desktop, every desktop, your computer comes with its own relay and then that has like the local version of all your events.

So you've automatically got, you know, but whatever your last snapshot Twitter was, you know, like you get on a plane, you've got the whole face and you can interact with everything.

You can do zaps, you can do whatever.

And then we just rebroadcast stuff when it comes out.

Just take, just take advantage of that like local first set up.

But there's a few infrastructure pieces to do just to make it work smoothly.

But that's all getting built out as well.

There's some really interesting stuff around.

So the DNS things, so like how can we start to remove sort of DNS and IP?

It's being like a centralized like naming authority for who gets to have websites and who doesn't.

Like how can we just rag based on public keys?

Things that got to be very powerful just means you could have one service and that service could be anywhere in the world.

It could move around the world, do whatever and just you'd know where it is from.

Its public key and its reputation should be pretty cool.

Yeah, I I think there's a lot, there's a lot of stuff that's just very cool skating works on, which is it's nice, but it's people that like, you know, very nice.

Various servers on demand.

So you do, you know, you can just send like a Noster request for the cashew note in it and it'll spin up a server for we somebody else's infrastructure.

So people being able to then sell their unused resources.

It's kind of like a private cloud for people to sell IRA, you don't worry in the world, but you give me money, I'll give you 10 seconds of silver time and then we can just keep going.

Just keep refreshing it as you go.

So stuff like that would be interesting.

Another guy's done a lot of work on like location based stuff, so looking at how we can get all the different like real time streams of information and pull them all into public maps so you can have like a Internet.

He showed us the IT was like the finish, the Norwegian, like the camera is like where it was.

It's like one of the transport authorities would stream all the things and just pulls up a noster map and then you look at it and see all the trains moving around just like in the city and stuff.

That's kind of interesting yeah that's cool.

You can do location sharing or private location sharing so right then if it could be this week around like private location like location mapping like in offline systems so it would be in bit chat you can share stuff over and offline Bluetooth mesh that says like you know this is happening over here this is happening over there.

So if you're in like, because a lot of people picked up big chat for like protest type situations and in Burma and Thailand, various places, that would be very cool for that sort of use case, be able to share information amongst the crowd.

It's the mesh.

It's it's interesting.

It's just a lot of light, really cool lighting going on.

See.

Be interesting to see what everybody goes today, yeah.

Be nice.

Yeah, it sounds like it's been a very nice four or five weeks, so it's probably been a bit longer even maybe.

Six weeks with.

Yeah, Week 6.

There you go.

Yeah.

There you go ahead.

Nice.

How about yourself?

Very good.

Well, aside from doing all sorts of non coding work and doing a bunch of investing, which you can also go back and listen to Jeff Booth chastise me for not just buying Bitcoin in episode 20.

Was it 24?

I think it was 24.

So there's that.

I have essentially built up a an e-commerce site the last couple of days for us, because we obviously I think we talked about this in the past as well as not having merch for other stuff, but actually like building it out as as its own like kind of stand alone brand.

It kind of taps into some of this like AI nerdism, But so that's called supply drop.

And yeah, I finally learned just why everyone uses Shopify for e-commerce like it it does.

It does save a hell of a lot of time, effort and energy, but it's been fun as an exercise to just kind of build out all the infrastructure for an e-commerce site.

So that's that's getting there.

There's a good bit of learning in there too.

So just doing something different.

I think I was quite keen to just see how far you could get with some more technical website builds as well with Claude code.

I started that one with Claude code and I'm finishing it with with codec.

So that's quite good.

And that all runs off of like a third party, third party manufacturer and distributor.

So it's just API calls to them and the the admin dashboard just populates with that information, feeds it through to the the front end through to the the web store.

And yeah.

So we don't actually have to do anything, capture any payments or yeah, it's quite, it's quite handy.

So I did that.

I kept working on the on the pod graph as well because I think that's just like something I'd like to better understand.

So and also this is an exercise that Quad Code's quite good at is taking all of that just for information and then building out the Jason files, tracking all the entities and relationships from each of the podcast episode transcripts.

So that's been coming together.

I feel like there's a process there that needs to be thought through a little bit more and expanded upon so I can actually provide like a better, almost like a better system prompt to figure out like how you create the optimal Jason based on a, a podcast transcript.

Because I feel like it's a little hit and miss at the moment from episode to episode.

The continuity isn't quite there yet.

So that's been an interesting exercise.

And ultimately, I'd just like to get that to a point where I can have a conversation with the pods because we do tend to surface, I guess, themes and a lot of those themes are recurrent and we'll add context to them or we'll change our ideas or our philosophies a little bit, or they'll evolve over time.

And and you know, if you haven't necessarily caught an early episode that caught the most recent one, you know, it'd be an interesting exercise to go through with an agent and track the path back to how we got to a certain position of philosophy.

So that's the thinking behind it.

And if I can get that to live on on the web, that would be kind of cool.

So, but it's really just an excuse to learn how to use graph databases using Neo 4 J.

So that's a bit of fun as well.

And then obviously, I think we could turn that into like Hype Band V2, which if anyone remembers from an earlier episode, it's just a Twitter bot that takes the transcripts and creates a bunch of tweets and publishes them.

For me, it's just giving it better context, better memory to work from and then create better content because I'm just terrible at doing this.

I'll definitely get into moods where I'm like, yes, I'm going to write some content and I'll do it.

And then a week passes and it's like I haven't done any more.

So the, the continuity from me there isn't, isn't quite up to speed.

So yeah, there's a few interesting projects like that, but I've definitely got a bit of, a bit of ADD across, like too many different projects.

So I'm trying to just get a couple to completion at the moment and and get them off the desk so I can I can move on to.

The it's talks about like when you get back in Perth, we should be we should look to set up some workshops around coding agents and non coding agents.

So you know, that had to vibe code and had to vibe do your Texas, whatever the use cases that that hits people.

But this this is an area that it's very powerful and I don't understood.

I mean, I think I've seen like all the different vibe coded setups in the last sort of six weeks or exactly like what works and had to do.

Yeah, there's a lot of experimentation going on at the moment trying to break things, trying to get across like I started building like a accountability app that I want to use to to track habits.

And again, that's just because I was like, oh, it'd be cool to just try and figure out how to build an app.

I haven't really spent too much time like a mobile app.

I haven't really spent too much time trying to do that.

So let's have a play.

Spent a lot of time in design docs and ideation because I want to, there's a, there's a couple of things I want to track, but I also find like all of the, all of the habit trackers just tend to allow you to stack as many habits as you want.

And then it's just not, I don't find any of these things actually work particularly well.

It's like a bit of a gimmick, bit of a novelty.

So I was like, what if I just lock?

One habit.

Lock you into one habit you have to achieve like 30 odd days or something.

I was trying to figure out exactly what the optimal amount of time I think it's.

20 I think.

It's 21, but it kind of turns out it depends.

21 days is what I was reading.

Yeah, yeah.

I mean, 21 exactly.

It's a good number.

So I was thinking like, I would build it around this.

So depending on how well the habit sticks after 21 days, that will unlock whether or not you can add additional habits to it.

So just looking at like simple ways to game a fight like that and force you to to actually do this stuff.

But then, you know, building in like an accountability buddy into it, which is just a hi that is tracking this stuff and getting to know you and giving you feedback.

So that's again like another little little slides side quest.

There's lots of side questing.

Yeah, I know it's problematic.

It's a.

It is actually starting.

You can feel like haven't got a problem, just need that you just leave this one side quest and then you can stop side quest.

Exactly.

Yeah, stop me side questing.

That's the that's the habit.

Well.

I'm going to have to finish this demo so I've got something to stand up with, but I will.

Let's leave it there.

That's the good stuff.

Solid.