The Bookster Podcast: Vibecoding Clients

Transcript

Hello, I'm Simon and welcome back to the Bookster podcast from inside the SaaS Holiday Rental Company, Bookster based in Edinburgh, Scotland.

Follow along as Bookster discuss their inner workings and their journey to where they are today.

Thank you for joining us and we hope that these conversations are insightful but ultimately helpful.

Today as always I'm joined by Robin Morris co-founder of Bookster.

Robin, how are you today?

Minute 1

Good morning, I'm very well.

I'm here on a school holiday so unfortunately that means child interruptions are almost inevitable at some point.

The door will burst open and someone will demand something.

It happens to the best of us, it's happened on BBC TV so if it happens to them it can happen to you.

Again, luckily we're also joined by Adam Aaron, support lead at Bookster.

Adam, how are you this morning? Yeah, I'm good, thanks Simon.

Good morning to you both as well.

I've locked myself in a room and my child is at university currently so no chance of him coming in, although it is his room.

If he does turn up there's something terribly wrong.

Yeah, if he does turn up then he's failed his exam.

Well look, we're back again in the pod booth as they say.

Today we are going to be covering a topic called vibe coding.

Minute 2

Now to me it sounds something like out of a VG song but to you guys it is probably a very important part of your job.

So was one of you like to kick off and tell me and the listener exactly what vibe coding is? Okay, I'll kick off and say what I think it is or what I know it to be I suppose maybe.

And then Adam can interrupt or come in on some stuff that our clients are doing as well.

So I've been fundamentally what it is, is using AI, using these LL models, Gemini, Claude.

There are lots of tools out there that you can use to write applications or write software that can help you function.

Now internally as Bookster we sort of use it a little bit.

We've used some tools that are essentially what you might think of as vibe coding.

Minute 3

As in that's the only thing, the only way that you're creating the code is by typing in prompts to an AI chatbot basically.

You know you're saying "Oh create me this, create that, change this, change that" and you're doing it just through the chat interface.

That really is vibe coding where you're not really going in and actually changing the code much yourself particularly.

So yeah, so I think that maybe outlines what it is I think.

Yeah I suppose I would define it as coding in plain English or whatever language you natively speak and the technology translating it into an actual coding language.

Minute 4

So you don't necessarily know how to read the code even.

You're more interested in the results so it's very much a back and forth between you and a computer or prompt.

I suppose the computer is probably sitting somewhere else ultimately.

And it's yes a very iterative thing so you sort of start with something and it does maybe a bit of what you want and you're just constantly adjusting it just by updating instructions in plain language rather than a programming language.

Okay, that's clear I think in my mind so thank you for that.

How does it then...

What while listening to the Bee Gees? Sorry what was that Adam? You do all that while listening to the Bee Gees.

Yeah exactly.

Minute 5

So how do you...

how does Bookster use this then at the moment with your clients or internally how does it work? I would say we don't use it so much for the core product but from a client's point of view they are now starting to get instances of where you've got clients you wouldn't expect saying can you give me some information about your API.

An API for people who don't know is a way of communicating with Bookster with a programming language essentially so you're asking it for data.

You wouldn't really go on the API as a person and query it for data.

A script or a program would do that so they're getting vibe coding to do that.

So people were generally finding that have some technical know-how but are not programmers themselves so they're just technically minded maybe.

Minute 6

I would say that's the people that are starting to do it but I would say as time goes on we will likely get more and more people...

you know as this becomes more of a thing that everybody's doing potentially it's going to get more and more people with less technical know-how asking to do similar things.

So what are people doing then when they ask you for this API what are they then going on to do with it? Usually I would say they're building their own dashboard is the most common use case so far so they're taking maybe...

so we've created like an arrivals report and that shows things in an opinionated way of what we consider important and what should be presented to the person dealing with arrivals.

Minute 7

That may not suit everybody so they may want to customize that whereas in books that we don't offer a customized view as such they may want to just build their own.

So they'll take the data from our API so the bookings etc and structure that however they see fit and they will generally do that through some sort of vibe coding interface.

And it will create their own dashboard to do whatever they want.

So then does that worry you guys at Bookster that people are going off and doing their own sort of adaptations of what you're providing? Not in the slightest from the point of view they're getting something they need.

We're providing the data they need it's not like they're not using Bookster they're just using it a different way.

Minute 8

Yeah I think it's not...

no I would say definitely not worrying it's interesting, exciting, it's good that people can use the data out of Bookster or use it in a different way and it might solve a problem that we can't solve for them potentially and see things in their data that maybe we haven't been able to see or expose to them.

Ultimately Bookster is kind of one, obviously there's lots of different views and things but can't make it infinitely flexible.

So if you wanted to get the data out and do certain things or manipulate the data in a certain way, if you're technically capable then that's great, brilliant that you can do it.

And I think what vibe coding or what essentially a lot of these AI tools are allowing is the bar to be lowered significantly for people to be able to do that whereas previously you would need to be a developer, you would need to understand

Minute 9

JavaScript or PHP or Java or whatever language you were using whereas now you can use these tools to basically if you've got a basic understanding of the AIs and the data that's coming out of them then you can produce something or make something yourself potentially that gets you something or displays something in an interesting way for you or for your cleaners or for your owners or for whoever happens or whoever, someone in your team.

So I think at the moment it's kind of definitely, you know, they're sort of side projects really rather than anything else but it's definitely something that's happening now that previously the people that asked about our API were developers rather than, you know,

Minute 10

are sort of more regular users and it's now more regular users as Adam says with the sort of technical bent I suppose that are now using the API and using these vibe coding tools.

And does this, I mean, you say you're happy for people to do it but does it give you ultimately any more work, you know, do things go wrong from the client's point of view and they then come back to you and say well can you help fix this or? Sometimes what we found like maybe they don't know like where so an API has sort of endpoints of where you can access certain data so if you want booking data you would go to the bookings endpoint if you want property data you go to the properties endpoints

Minute 11

so they may expect the data to be in one place and it's in another so they'll get questions like that like how do I access this this data.

The other thing we found is it's very quick to write some sort of script or program with with these tools but it's not necessarily efficient as in the program itself is not necessarily efficient so what we have found on occasion is, you know, for example, it's the client will say, you know, get me all the information on my bookings that have come in and in the past two weeks, and it will get all the bookings for all time, and it will do that every every 30 seconds.

So what we end up with is our API or our servers getting hammered with with requests when all this user wants to know is about their latest two weeks of bookings once a day but they're asking it, or maybe they haven't asked it to do it every 30 seconds but the program is just assuming it needs to be done every 30 seconds and they,

Minute 12

because they are not looking through the actual code, they don't know that it's done this.

So they're assuming it's working whereas our resources are getting hammered.

And we need to, you know, super likely say, can you adjust your code so that it's, you know, we kind of see what you're trying to do here.

Maybe suggest a prompt to get them to adjust their code so it's a bit more efficient.

We've had a couple of occasions of that certainly.

In terms of more resources for, so not resources for, you know, time resources, I wouldn't say it's a huge time sink in the interface they are interacting with is doing a lot of the heavy lifting so it's doing a lot of the figuring out

Minute 13

of what they need to do and what sort of questions they need to ask.

So I'd say that side of things is kind of taking care of itself, but we certainly get, you know, where they've hit a roadblock of some sort will get requests of, you know, is there an API for this or I'm trying to get this data what would you suggest.

I, you know, where would you suggest I look.

So there's a lot of that sort of, or a lot more of those sorts of questions.

So it is having an impact on Bookster then.

I would say maybe I mean Robin's probably got more to add to this but I'd say it's maybe altering our direction a little bit in terms of what we choose to open up to our clients as far as you know that the previously the you know most of the the data you would would want would be within site you know within Bookster itself you would log into Bookster.

Minute 14

And, you know, we were, we described the data in various views, depending on what task you're doing.

Whereas now they might not be logged into Bookster at all.

But they're just accessing the data that's within it and the structure that's within it so may it may alter our decisions of how we expose that, you know, in certain ways and influence us, you know, as far as the direction of the product.

But yeah, I don't think Robin's probably better placed ultimately to answer that.

Yeah, I think it probably changed some of our priorities will be to improve our APIs.

Minute 15

And so we're more likely to be doing that, suppose, as opposed to writing other interfaces, but you know, obviously, it's just more a case of prioritization but because we're getting more people asking about the API as was limited before we maybe need to pay more attention to it and make sure it's up to date and.

Yeah, you know, we, the improvements we can make there that perhaps we might not have done so it will change our priorities a little bit.

And, and, yeah, suppose for support, it also means that you are, you know, supporting an API which is completely different to supporting a user interface.

So, and it's potentially a little bit more, you know, well, it can be more difficult because people are doing things you can't see what they're actually doing or what they're trying to do.

Minute 16

So, yeah, trying to support someone's use of it is quite difficult to but no, I think that's the only thing it changes is, you know, our priorities of things that we do will maybe be skewed a bit more towards the API than it would have been before.

I would also say that the level of people doing it right now, as we mentioned before, they're probably slightly more technical minded than I think what's going to happen and maybe not too distant future is you're going to get people that are less and less technically minded.

And we'll just be like, oh, yeah, an LLM can do this.

You know, I'll just ask it to do it and it's just a one liner and we'll get more of that which possibly will be a bit more difficult to support.

But I suspect that's not just going to be our problem.

I suspect that's going to be, you know, throughout various industries are going to be running into that at some point very soon.

Minute 17

And hopefully the technology will keep up.

Does it inspire you or have you had conversations about if a client has asked for something different than you already providing through this? I mean, do you then sit down and go maybe we should actually just natively provide this option for people? Has that sort of invaded your conversations? Sometimes.

I mean, certainly the first time I saw I was pretty wowed by, you know, this has just been built in like a weekend sort of thing.

And I think what the person had done was they'd set up a custom messaging system based on various things around a booking, sending a custom message based on booking dates.

Minute 18

And they just click a button that fills in and we have done similar tools.

I'm not sure if this is a result of that so much as we've done it anyway.

But this is very specific to their needs.

And I was more inspired that that's actually really cool that they've been able to do that.

So they've taken us as a starting point and sort of jumped off from there.

I think it will certainly, they'll be back and forth certainly as far as it will probably influence things we build as people come up with more heads means more ideas.

Obviously, some of those will be good.

Some of those will not be good.

And I think the ones that are good, I think certainly is certainly only a good thing to borrow from that to improve our product.

I don't know that it would directly influence it beyond that.

But I certainly think there would be feedback as a result of that that would give us ideas of where to take books from.

Minute 19

And have you had any feedback from clients that something's just gone horribly wrong.

They've tried to do something and it's not worked and they need you to fix it, for example.

Not horribly wrong.

Yeah, more just more just the I'm trying to build this I'm not getting the right you know the right date is more just you just got to look here instead.

So that wouldn't say horribly wrong.

I haven't had any disasters.

I haven't had anyone.

No one's died yet.

Hopefully, hopefully no one will die as a result of vibe cutting.

I don't think it's that type of activity.

For me who knows nothing about coding in any way.

If I was a client of yours.

Could I go on trying, try and do my own thing and then mess up the interaction between the Bookster and what I'm trying to do.

Minute 20

Is there a way to, I think the worst you would do.

I think the worst you would do is just, is it just wouldn't work.

I don't think it would.

I don't think it would expose any government secrets or anything like that.

I won't get arrested again.

You're not getting that.

Robin, what you're thinking what's the next 12 months with books are there with all this AI stuff going on? Well, I've got a small child here pestering me at the moment.

So that's the one thing I'm thinking about right now.

Ask AI to look after them.

AI can you look after Ben and tell him to go back to his room.

Right.

So, I think as Adam says, the tools that the clients are writing, they tend to be reports and things or like they're not quite so core things that they're doing kind of thing.

Minute 21

So, and they, I suppose what they want to avoid or what we want to make sure that they don't fall into the trap is of something that can be horrendously bad for them.

So we, you know, they maybe use the API to, you know, cancel bookings or something and they don't intend to or something like that.

So, you know, that could be quite disastrous for them.

I mean, that hasn't happened yet, but that's the kind of thing that they, you know, they're not at the moment people are usually using it for, you know, reporting and saying, oh, I can produce these graphs or insights or, you know, people arriving or when people departing, etc.

So, yeah, that's at the moment, it's not too bad.

So, but in terms of what we do next, it's essentially trying to, you probably develop our API as further so that they are easier to use, easier to understand.

Minute 22

They can get more information out, or we can update Bookster in a way that's logical and rational and easy, etc.

So I suspect that's where it will affect us in terms of supporting clients' use of it.

But also, yeah, we can draw inspiration from the types of things that they build as well, I think, and, you know, it's worth keeping an eye on them and trying to understand why they did it and, you know, were they frustrated with Bookster and that's why they did it, or was it more of a kind of inspiration thing? So they're not frustrated, they just kind of want to create something themselves.

So, yeah, when they're frustrated with what we can do, then we need to make sure that we're not, that maybe we should be looking to alter the core product, I suppose, so that it does reflect their requirements.

Minute 23

So, yeah.

So, I mean, all to it then, and I guess a couple of sentences because we're getting towards our sort of end of time.

AI then for you, is it a good thing? Is it helpful? I mean, for your clients, are you happy? A couple of sentences.

I think what it can do is make, we make a product for, I'm not the general public, but obviously, you know, a core audience, what this technology can do is make that product highly individualised for each person.

And I think that's a good thing.

I think that can only be a good thing.

Can they break it? Yeah, but it's not, it's going to break the bit between them and their clients, not between them and us, if that makes sense.

Minute 24

I think going forward, you know, these tools are going to make a lot of software do that.

You know, and anything where you, where we are almost providing a generalised tool set and then you can take that and customise it to your own needs as you see fit.

So, it's empowering, so it's a good thing.

Yeah, I think what, you know, what you, obviously, there are some instances in which case someone can vibe, you know, there might well be levels of competence where they can vibe goes all sorts of things that books are is currently doing and maybe sort of replace a lot of it.

Maybe, I don't know, but there is the kind of maintenance and uptime and making sure that it's available for for themselves and that it's not got errors or it's not doing things in a way that they didn't expect.

Minute 25

And, you know, if they've got a team who's supporting the team on using it.

So there's a whole lot of reasons why you still would potentially want something like Bookster.

Or, you know, something where you're not having to worry about all the things that we have to worry about because we, we will worry about it for lots of people.

So therefore, you know, we can handle that.

Whereas if you just decided to buy code some tool yourself internally, then suddenly all those burdens are falling yourself, which might be fine.

You might be happy with that the upside being is exactly or looks exactly what you want and etc.

And you can change it and you can, you know, so there are some distinct upsides and so people might, you know, they might, there's certainly clients or people who are competent enough or who are willing to take on that responsibility, etc.

That's great.

But for a lot of people that will, I think, mainly the main thing will be is a kind of, you know, to extend or to see data in a different way to order to build a sort of more temporary tool that they see something they use it for a few months and then they, you know,

Minute 26

but like a spreadsheet or something they, they then forget about it or whatever.

So I think, and ultimately it's mainly positive for us really.

And I don't, I'm certainly not too concerned about people just building their own.

But there's still going to be a real strong requirement for us to provide a service that works for them.

Good.

Okay, so ultimately, currently positive.

So, I mean, it's interesting and this episode has been focused on sort of the client side of it really.

Maybe what we should do is come back and do another episode on how AI works within the Bookster itself and how you guys all talk to use that.

Minute 27

So maybe it's a two part episode, maybe who knows.

But for now, let's say thank you very much to the listeners for listening.

Adam, thank you for your time again this morning.

Thank you for having me.

Pleasure.

Nice to see you and Robin, thank you again for coming on as well.

Thank you.

Nice to see you again.

And you and we shall be back shortly with another episode.

Until then, thank you for listening.

Enjoy the rest of your days and we'll see you soon.

Bye bye.

Bye bye.

[Music] [BLANK_AUDIO]