Brainfood Live On Air - Ep409 - Vibe Coding TA Tools - How Viable Is This Really?

0:03I want companies to be sort of used to be aware of who's supporting the show. It's not something that comes for free. You can't do this every week without companies coming in and stepping in and saying, yeah, we're going to put some money behind it. In the same way, when you're doing dinners, you've got to put money behind what you're doing. You know what I'm saying? They should know that the people eating the food, I think, yes, you can absolutely pitch to them. There's no reason why that shouldn't happen. Okay, welcome everybody to Brain Food Live on Air, bringing it to you every Friday, no fail.

0:36It's been constantly rolling forward. We've got 400, I think we're at 400 plus, 409, 410 episodes. This year, we're going to be hitting, I think, 70 episodes of this show. And the reason why is simply because there's so much to talk about in the world of recruiting. It is such a moment of change for us that I think it's so important. and we elevate the conversation, make it accessible to everyone, get more sort of public dialogue going on on the topics that we care about today. So brilliant to have everyone here on the show. Let me first do some sound checks. Firstly, make sure everyone can hear and see me and Brandon here okay.

1:17Firstly, if you're watching this on Crowdcast, let me know whether the audio and visual is fine, please. Just let me know in the chat. We should be live streaming this also on LinkedIn. So if you're watching this on my LinkedIn or anybody else's LinkedIn, do let me know whether the audio and visual is okay. Let me just do a quick check here on my own LinkedIn. I really should know that. And then we'll figure that out. Okay. It looks like a few people can hear and see us, which is good. Thank you, George. Thank you, Glauco. This is all brilliant stuff. Okay. I think we should be ready to roll forward.

1:50You might notice we have a new face sharing the screen with me today. um this is um this is always great to bring on new uh co-hosts and onto the show um and always when we get sponsors on i always get this try and get the sponsors to do a bit of work um because why not you know so anyway we have brandon amaruso here of of scaleless brandon why don't you quickly introduce yourself who are you what it is you do man yeah well thank you for having me on i'm excited to be here so i'm brandon i'm the co-founder and ceo of uh scaleless uh prior to this company, I ran a professional services firm where we primarily use LinkedIn and Indeed for sourcing and an ATS that shall go unnamed that left quite a bit to be desired.

2:35So once I sold that company back in the middle of 2022, we started building Scaleless to solve all the pain points that we had from a hiring and recruiting standpoint, but more specifically what we saw on the job seeker side of the market with people like my younger brother and the just comically bad recommendations that he would get in terms of where he should be applying. My favorite email was his that he received to go be the CTO of a Fortune 500 company, as well as the air traffic controller at the Dallas airport in the same email recommendation. So that's ultimately how Scaleless is born.

3:06Yeah, fantastic. And great that people are taking action and actually launching sort of solutions rather than complain about it. How many times have we complained about how bad technology is? And then we just go on the next day and just use the same products again. So So brilliant to see you take that interaction there, Brandon. Scalist, a lot of people are saying are a little bit different from what they've previously experienced as a product. I wonder whether you could tell us a bit about why they would say that. What's your sort of understanding as to how you guys are different from some of the other great ATS products that are out there?

3:41Yeah, so I think first and foremost, I did not have traditional TA background. So from day one, we brought in our VP of product. who spent the last decade in-house as a TA leader and then surrounded her with advisors from all the incumbents that we're replacing to help facilitate the end-to-end solution. And the most critical thing is that we're not single-sided. So while employers will use Scalist to do everything from job intake, sourcing, all the way through offer letter sign, we have our own talent network. So every application that comes into the Scalist system, we automatically we generate a universal profile for that applicant in the background.

4:17So as soon as they get rejected from that first role that they applied to, we can leverage all the ATS data that none of the other job boards have access to, to train our recommendations model and push them into other open positions across our network that they're a far better fit for. So while our near term sort of competition is the ATS systems, and that's who we displace for our customers and all the other third party tools that bolted on top of it, I built Scalus to go after LinkedIn and Indeed. yeah i love how palsy you just dropped that line it's like the two biggest companies in the recruitment world but you're going for it brandon which is great stuff and i love the vision of it being a two-sided ats that is a new play i think this idea probably uh rolled around you know it rolls around every decade no one's ever been able to execute it but maybe now is the moment imagine how this works folks you're an ats you can manage your applicants through that system but you've also got a marketplace, a talent marketplace where candidates can effectively be recycled into different sort of job opportunities.

5:20And presumably this would increase the chances of good matches happening. So fantastic stuff. Check it out. Scalist.ai also got a ridiculous offer, don't you, of some type, Brandon? What is this about buying out contracts and stuff? i mean i'm hesitant to say what that means but uh what are you offering for companies that want to give a scaleless a shot yeah so because we you know rip out the entire recruitment stack the way that we do our pricing is you send us all your existing contracts we add them up and then we cut them in half so your ats sourcing and note taker all that consolidated into the system yeah very very interesting stuff it sounds like making it easy for people to give it a shot so So if you're in the market taking a look, you should do that.

6:06Okay, Brandon, we're going to drop you off screen for a second. I want to bring on some of our other guests and then bring you on back for the panel where we're going to be discussing how vibe coding TA tools, TA tech tools, and how viable this is going forward. Before we let you go, though, in your break, can you give us a bit of a, I guess, your own take on it? um you know sort of you've entered the moment of uh software development right at the point uh where ai is becoming so diffused across general population everyone is creating agents some people are creating apps out there uh so everyone's like doing a bit of vibe coding what's your general take um on ta people uh vibe coding tech tools is it viable if so where can it go to the point where it replaces the entire tech stack?

6:55What's your take, man? I don't think it can replace the entire tech stack because of the sort of DevOps and infrastructure requirements to support, but it does make it more critical that the system you are using is open and flexible and can plug in with all these other tools so that you can build on top of it. And so we see, you know, the fastest moving teams adopting some of these more modern solutions because they can plug into all the different AI solutions that allow them to build and create, you know, personalized experiences within their organizations. So that's where we've seen the most value add.

7:26That's really interesting. Again, this is why I want to bring you back to the panel here for Brandon. So what you're saying is that actually the traditional software vendors need to now think about building software that is dynamic and flexible, and able to interact with, I guess, DIY built tools from the customers that you know, you can expect them to do it, and they should be doing it right. So you can't be creating a monolithic system and saying, This is everything. You've got to create a system that can flex and integrate with everything. Unpredictable software innovation that would happen on a customer site.

8:04Very, very interesting, Brandon. I'm going to get you back in to talk about this when we go to panel. Firstly, I'm going to bring on Ala Pavlova, though. She's going to help me talk through a few things from the sort of, I guess, from the newsletter side. But firstly, let me say hello to Ala Pavlova. Wonderful to see you, Ala. Is that a new background, a new room? Have you reconfigured your house or something? What's going on? New room, yes, because I have a family visiting and it's a very busy house. So I have to find this place where, you know, nobody's screaming. So you squeezed in granny and grandma sort of into a different part of the building.

8:44You're in the closet somewhere doing this show. I appreciate it, Ala. For the folks who don't know you, Ala, do you want to quickly introduce yourself? Who are you? What it is you do? Of course. Hey, everyone. My name is Ella Pavlova. I'm based in Amsterdam. I still do hands-on recruitments. I do train teams. I do web coding. I try to keep in touch with technologies, although I feel like it's getting a little bit easier in one way, but harder in a different way because things are changing so often. So I'm excited to be here and talk about this long-awaited topic. and I'm excited to see you, Han.

9:23Let's see what he's going to bring us today. Yeah, fantastic. Let's talk a little bit about your experience of this because I actually see you here pretty much as one of the guests on the show rather than some sort of co-host because you've always been an experimenter, an innovator, early user, super user of AI and so on. um what have you done so far with um uh with vibe coding um i mean firstly can you tell us what vibe coding is and secondly what have you done that actually seems to it seems to be effective for you um as a vibe code sure it's a straight away to the point okay um vibe coding is actually not what i thought at the beginning i thought it's like you know i just um open the chat and I write whatever I want and I get things out of it.

10:16Actually, how the engineers are using it, the wipe coding is you try to use technologies in the ways that it feels right. So you know the rules, but you're still experimenting with the code so that you get the results that's probably going to be a little bit different. And now one of the popular tools, this is in the market is lovable. I know we have some guests from this. I'm sorry for the spoiler, but I'm excited about it. And there's also another tool that's called Coursor. I think if anyone tried, please acknowledge in the chat because I'm also very curious if you try this. And my personal philosophy is there's a new tool, install it immediately and check it.

11:04And don't feel embarrassed or don't be scared of trying things. and given access, of course, you have to be careful and thoughtful about it. But try it in your personal computer and see what it can do for you. Because, again, there are so many tools and so many options. And we can be just lost. And usually I start with the problem. What problem I want to solve? Let's say when I wanted to build my own website and I didn't have any skills as an engineer, I went to Lovable and I started to type, I want the website, this and this and that, and it's created something for me. Yesterday, I actually showed a recruiter who was struggling with the hiring manager, who was saying that he has no time to interview.

11:57And I did something very simple with, actually, with Cloud Design. I don't know if anyone is using Cloud Design, but allow me to share my screen. And it's going to just take you a second just to see how easy it is and how funny it can be. Just a second. Sharing the screen is always a challenge. I'm going to do it very quickly.

12:22Only host and speaker can share. Have I not made you a person that is... I need to change your permissions. Give me a sec here. I'm sorry. No, no. I assume that you would be like, oh, you're only an attendee. Let me put you as a speaker. Okay. Yeah. So, yeah, great. I'm going to share the screen. So, basically, very difficult hiring manager saying no time to interview. And, yeah, so what I told this recruiter, use this one. It's like a live board with timing. And so sit down with your hiring manager. He says to you that he can be available only in, like, say, 16 and so forth. October, you can tell that the hiring can happen in the 4th of December.

13:08Oh, there is a panel delay. Okay, so then we're looking there. But okay, this person has eight weeks, not just period. Yeah, we're looking into January. How does it sound to you? And you see it's interactive. So it's one prompt that I use in Google Code Design, just saying, hey, I need to have the visualization that is live so that I can change it in front of the person. and just to show them like, you know, that's, yeah, if you're going to delay it, you're going to see the person next year. So do you want to change it? Let's play with it. So it's just like, you know, super small but relative things.

13:43So I think if you want to solve the problem but you're not an engineer, that's the solution for me. I'm not an engineer and I don't think I want to be an engineer. And that's the reason why I love all these new tools and things that I can solve my problem with. No, I love that example, Ala, because it's very useful. Folks, did you see what it was? It was basically like a kind of a visualization of starting time and kind of empty seat time. So the manager here probably didn't realize his or her kind of unavailability to do interview would have obviously impact as to when this person is going to start, but you've got to just visualize it for them and say, hey, listen, yeah, this is happening now in January.

14:31And that may kind of change that person's mind and prioritization. That tool doesn't exist as a product, but you can make it into a little product simply with a prompt. And we can see that this, how do I want to call it? I would call it maybe like a micro app in some way, right? I mean, you can never, yeah, it has a single use case, maybe even a single sort of circumstance. So it's only maybe a single manager in that circumstance that you might use this scenario. So it's almost like an ephemeral tool. But because the kind of cost of production of making this today is what, as you said, one prompt, right?

15:11So how many minutes do you reckon it took before you got that? Five. I can share the prompt. I just even I use the whisper flow just to talk my prompt so that I'm not typing. My typing skills became horrible after using all the note takers. so basically a few minutes to um do the inputs and then a few minutes wait for quote to design it and then like boom and you have it in front like by the way hiring manager get scared see in january 2027 you know made the point clear i was gonna say everyone be terrified so that's super useful and i think i mean folks you don't think that's useful let me know in the comments also can you also see um how this is different from if you like software engineering right this is basically creating a tool for a single use maybe a single user and maybe you never use it again it's like you just need that thing solved maybe you park it you can redeploy it at a later point it's a useful thing to have maybe the team can use it at some point but it's not something you'd ever think is okay has any kind of impact on your your your core kind of tech platform.

16:20It's kind of a useful tool for right now, isolated, localized to what you're doing. So I think all of us can agree that's good vibe coding. That's not something that's going to freak out anybody illegal, I hope. Okay, Whisperflow, Randy. It is basically an app that listens to you, essentially. And you can just remove typing from it. So give it a shot if you think your talking is better than your typing. And it might be very useful. Joe Atkinson, Selfish Software. I kind of like that as well, even though I'm not sure the connotation is quite correct. Okay, let's bring on some more people that know how to vibe code.

16:58I'm going to bring on some engineers later, by the way. I've got some software folks coming on to join us in the show, as well as George LaRock, who's written intelligently about this also. But I want to just explore a little bit from, okay, internal TA, what are we doing about sort of vibe coding? And hey, we talked about Lovable before, so no more better reason than to bring on a recruiter from Lovable. It's Glauco Leal. Glauco, great to see you, sir. Welcome to the show. Thank you very much. It's real to be here. Great to see you, man. Can you quickly introduce yourself? Who are you? What it is you do?

17:32Of course. So my name is Glauco Leal. I do go by Glau, so it's simpler this way. Originally born in Brazil, been living abroad for 13 years, currently living in Stockholm. That's my country number eight. I joined Lovable about 10 months ago, roughly. One of the initial recruiters of the company when we were like about 60 people and now we're way past those numbers. So I had the opportunity to build some really, really interesting things in here. Yeah, I was going to say, you've got a unique pressure, don't you, Glau? Because firstly, you're a recruiter, but you're also a recruiter working for Lovable.

18:06So I guess you better be good at some of this vibe coding stuff. kind of a unique pressure. Okay, you arrive. What did you see on the landscape sort of tech-wise, tool-wise? And what did you decide? Okay, first thing I've got to try and like just experiment with is this thing. Can you tell us a little bit about the first thing that you tried to build from a recruitment perspective here at Lovable? Of course. I feel like being at Lovable, I'm in a very privileged point because I understand that when you're not in a company that's so into AI, you might have a certain amount of, you know, red taping around what you can build, what you can do, which resource you have to even start to have a conversation.

18:50So the culture of trust is pretty much fundamental, right? But then at Lovable, when I joined, well, there were basically no recruiters in the team, right? So everything was kind of like a greenfield. So in our particular scenario, the main urgency is like, okay, how do we hire and at the very same time how do we understand what is out there in the market that could potentially be noise so a few months after i have joined i built something that i'm super super super um excited about which is an internal tool for talent intelligence and competitive intelligence which essentially helps us to minimize noise from the outside help us to make better hiring decisions and stuff like if we're trying to hire talent from xyz company what is this company trying to build for the next six months, right?

19:35Which talking points we need to have with the candidate. And none of this is based on data that's sensible. It's all public data.

19:46You're muted, hon.

19:52Schoolboy error, my God. Okay, Claude, that's brilliant stuff. So can you tell us a little bit, is there anything you can show us? Is this thing still live? Is it something that we can look at? It is live, but not something that I would be willing to show on this stage because there is some internal information. So I'll leave it to imagination, but it even went as far to a point that truth lovable, I could turn this app into a agent that lives on Slack and essentially queries information to our equipment team, also does some sourcing. So it's evolving as time goes by. And I would also love the opportunity to talk about how applications can evolve as the team grows.

20:33as necessary grows, right? Because while I agree that tools that are quote-unquote fiber-coded can have some specific usage, as your team grows, as your company matures, you can continue to evolve that. And I think that's one of the beauty behind being able to vibe code things because essentially you do not necessarily need a huge software engineering team behind you to help you to do this. Yeah, very interesting. So you might start off with a single-use problem, such as as ala demonstrated uh my kind of encapsulation was initially okay maybe it stays that way but what you're saying is actually no if if it is potentially useful maybe you build additional stuff on top of it no different from how software evolution might actually work um and it may just become more of a dynamic sort of tool going forward um and evolve as you go uh go on Is that a decision you make sort of in advance, Glau, or is it something that is more emergent, do you think?

21:33Like you decide, for instance, to build with a view that the application is going to change, or do you kind of just sort of use it and then see what happens? Yeah, to answer this question, I'll try to split this into two things. So the first one is sort of like how I go about the decision of what is sensible to be built, since you can technically build whatever your heart desires in a platform like Lovable or Claude or whatever. And also, how do you maintain that? How do you upscale that if necessary? So on one hand, I feel like the most important thing to understand is, one, who are going to be using this tool?

22:16how often this tool is going to be used, and which type of data would this tool have. If this has any type of sensible information, any PII territory, I would say, probably not the best usage of your time or direction to build, right? And then you also have to think about change management, right? Change management in a startup could take a couple of days, in a scale-up, maybe a couple of weeks. In a company, certainly a couple of months. So if you're building anything for whatever that it is, you just have to be very mindful about that. Who's going to use, how long is it going to take to implement, and which benefit, like how are we moving the needle by building this, right?

22:55And then I think the second aspect of this question is not necessarily about you thinking so much ahead of time on the roadmap of this tool, because businesses evolve, hiring needs evolve. and I feel like the beauty of being able to iterate on these things is to understand like how this tool that maybe you built six months ago might now be obsolete but if there's still lots of buy-in in the company you can still iterate on that and improve that over time yeah really interesting man I mean it sounds to me though Glau you're getting to the point where your just your psychology is moving quite some distant away from like traditional recruiting right um because traditional recruiting just psychology wise you're going to come in your immediate job is okay where the jobs where the vacancies i've got to just fill them right and maybe you then develop a tool to help you do that and then okay but you're still going to fill it but how you're speaking sounds more like okay you are becoming more like an engineer um did you feel that that is it do you think that was a transition that you consciously made or do you think that it simply emerged as a result of you you know being in an environment like in in in lovable I'll definitely answer this question and maybe we can jump to Allah's question because I see that she has her hands up for quite some time.

24:10I feel like Allah will know this, so it's not going to be a surprise to her. But prior to being a recruiter, I used to be a developer. And I also did about two years of dedicated project management for global companies around talent acquisition. So whatever I build, I build thinking on how can I scale this over time, right? And sometimes it's honestly, transparently, not that obvious. What you think in your mind might not necessarily be instantly bought by your team. So it's also like how we communicate, what's the intent behind, what's the use case. And whatever you're building, you just need to be like super humble and open to listen because maybe in your head, that was the perfect use case.

24:54But if you're extending that tool to the team, it might not. Maybe you're doing technical recruitment. Maybe your colleague is doing go-to-market. And maybe for go-to-market, what you're building might have absolutely no usage. So you just have to iterate. Yeah, very, very interesting. Okay, Alec, go ahead. You wanted to say something? Yeah, I'm sorry. I will lower the hands. I didn't see my face. Sorry. I actually wanted to add something to what Glau was saying. to build something for the team. Sometimes it's better to do it together and have a scheduled, let's say, Friday, casual 30-minute build-up call and choose whatever challenge you're trying to solve.

25:37Because sometimes, as you're an engineer, you know that sometimes we think that, oh, there's a lot of technology, so it's like we're going to build this very easy. But then we face some challenges, something doesn't work, et cetera. I also wanted to share my screen very quickly just to show you what's a simple thing as well that's done with the team. Like we were playing. Let's create somebody's website, like landing page, just based on the LinkedIn profile. So it's lovable. This is the whole prompt, like create landing page. it's already suggesting you a few designs asking you a few questions and asking me about the domain and stuff but it's built it for me just in a few seconds so it's um that's what i'm saying about very fast um uh results uh with web coding like the whole website can be done or something else if you want to yeah it's super interesting in the sense that the capability is now being diffused across many more people.

26:45I mean, even the two things that you've shared right now, Al, over the last five minutes, I would say three years or so ago, no non-technical person could have done it. Maybe a website, you could have used some sort of no-code product, I guess. But certainly the little app that you had before, you probably need to get an engineer to build something for you there. So now one person, non-technical, can actually do it. AI is doing the heavy lift, which of course is an existential crisis is for software engineers, isn't it? So now is the time to bring in some software engineers and they can help tell us about sort of how they're dealing with this.

27:19And I wanted to also mix this up because basically I didn't want sort of all of us to, because we're all in it together, but looking at it from slightly different angles, aren't we? So I wanted to bring on some of my friends that are in the software dev world. They build software, but now I guess they're moving into, they're also confronting this reality that maybe that isn't the primary thing that they do anymore um in which case how does all that work um so okay we've got andrew lawson here let's bring him on um andrew good to see you you're on mute my friend um let's get you off that there's a big test of tech skills here let's give that a shot uh there's pedro let's get him on as well are you there andy yeah can you hear me yeah we can hear you good um great to see you mate um Andrew, can you quickly introduce yourself?

28:09Who are you? What it is you do? Of course. Hi, everyone. I'm Andrew. I'm a senior staff engineer and director and engineer at Monzo Bank in the UK. I kind of work in the platform end of Monzo, working on things like AI and security and stuff like that. That's what I do. Fantastic stuff. Great to have you on the show, Andrew. Long time since you've been on here. The last time you were on the show, by the way, do you remember we started the live screen and killed the live stream within like 10 seconds and then had to re-spin it up in disastrous fashion. So you're part of that experience, my friend.

28:45Still got trauma from that, but never mind. It happened to me too. It happens to us all. We have the very handsome Pedro Oliveira here as well. Pedro, great to see you, sir. Can you quickly introduce yourself? Who are you? What it is you do, mate? I do a lot of things, but thank you for the compliment, man. And I'll give it back to you. You look great today. I've been building in the HR tech for a long while. And recently I was going to say I moved on, but no, not really. You never move on from HR tech. I started building something in the physical AI space, which is these intelligence robots for search and rescue missions.

29:31It's still very early, like two months in. But, you know, I fell in love with robots for a long time, but more recently going all into that. But, yeah, I guess I'm supposed to share stuff with the group on my AI use stuff, but I'll follow your lead. No, I mean, what I want you guys to do is to just give opinion as to how software engineering has changed. and what do you think when you've got non-technical people kind of intruding onto your turf, I guess? I mean, I wonder whether there's some element where you think, hang on, isn't that our job? But isn't there also an element where, okay, interesting, how do we build software, as Brandon mentioned earlier, such that we kind of anticipate that users will self-build stuff and we build software that would help support that issue?

30:27So, Andrew, you want to kick us off on that topic. How does it make you feel when you hear non-technical people build what looks like pretty viable bits of software? And how does that inform you as a quote-unquote real software engineer and how you think about building your stuff? Now you know it's gone into the civilian population, so to say. It's a really good question. I think it's great when people are building stuff where they typically wouldn't be building, or they're somewhere outside of engineering. I think that it's helpful to see where, I guess, with user research and product and a few different roles, usually we try and figure out what it is that users need or people need or customers need.

31:14But instead, that's sort of being flipped around and we can see what people need by them explaining it and expressing it to AI, coming up with some artifacts where it's not quite right for them, then they're iterating on like the the way they explain it and the like the result of that is really good indication of what people actually want um the the ways in which it kind of like starts to interact with engineering is when we start to ask questions around like security and availability and resiliency and stuff like that but like typically when someone's trying to solve a problem for them or for like them and their direct peer it kind of doesn't matter so much um so long as it doesn't leak data and it you know it is somewhat secure i think that's kind of fine So I think it's a really good thing.

31:56Yeah, fantastic. So there's a dividing line, again, sort of the scope of usage. If the scope is small and it's not touching confidential stuff or doing anything crazy, go for your life, no worries. Maybe the vibe coding is a good signal as to what engineering could also be doing. So potentially it becomes a form of prototyping, I guess. um you know if you see let's say you've got a team of ta people and they're all creating job you know uh job description generators then you can say yeah we probably need to create a job description generator for these folks yeah like so inside of monzo we try and figure out like what it is that people like internally at monzo and our customers need and when we see folks in uh in our hiring um teams by coding things we try and ask ourselves why like why is it they're trying to vibe code these things what is it that we're not doing or our our sas providers aren't doing how do we help them sort of like elevate this vibe code thing to something that's a bit more you know robust that we can depend on is that something we want to do do we want to spend the time or is it like is there already a sas thing that provides this thing what's the gap and we try and like have that kind of conversation yeah very very good how do you get visibility of that though, Andrew.

33:08Is it something that you're waiting for? How do you get visibility of VibeCoding stuff? I mean, you can't really see it, can you? The way that we can see it at Monzo is that we don't let everyone bring their own AI tools. We have a bunch of providers like Anthropic and OpenAI and stuff that we can effectively observe how people are using those tools. We can see where people are spending lots, where they're using lots, where they're like trying to share artifacts they create and see the like the effects of that, where it's not entirely safe when they're doing it. And you can kind of see where like the like the edge of safety is sort of being teased.

33:50And that's where we start to like zoom in and like try and figure out what it is people are doing, what people want and try and solve a problem. Yeah, fantastic stuff. Shout out to Cy, by the way. He's the CEO of Crowdcast right here. So brilliant to see you on the show. By the way, this is episode 409. We've done 409 of these on your platform, Si. So, yeah, big kudos to you. I might actually bring you on the show in a bit, actually, because if you've got vibe coding stuff in the background, I want to hear about it. Pedro, your view of things. How realistic, viable is vibe coding? Basically, is it controllable?

34:27We hear about shadow use of AI, right, in a big corporate. People are vibe coding away, possibly dangerously, possibly exposing things? What are the guardrails that we need to put into place to make sure we activate people to do the vibe coding, but also don't put the business at risk in case you get over-enthusiastic people just charging away and doing stuff that they shouldn't? Okay, just a quick note on the previous points. In tech in general, we've always had people not from a formal background, if you will. Like most of the people building in our space, Andrew, has been people that didn't really do the computer science course, you know, in our uni.

35:11Casing point, exhibit A, right? Just wanted to drop that one. So we've always been like, even though sometimes it didn't look like to the space, oh, those guys in tech, they're like, yeah. But it always been an open group, if you will. So I saw this very naturally. Honestly, I do have a formal education in the visual science. And from the first moment I stepped my first foot in the uni, I was like, damn, you know, this coding thing is really in between me and building what I want. I always start coding as a roadblock, not really as something that will give me. Yeah, I would get a lot of happiness when fixing a bug and then another and another.

35:56and at some point the whole day passed by and you're like, fuck, I need to have lunch. And that would happen to me many, many, countless days, if you will. Now, to your question, sorry for taking a detour there, to your question on like honesty is the same old, same old. I mean, we've had IT managers in the past with card rails on equipment. You're going to have guard rails in here, especially in corporates. Like that's going to happen. I don't expect that to happen in new startups starting up. Like, I mean, one guy, two guys, three guys in a startup. I don't expect that to happen as much. But in a corporate, hell yeah, you're going to have guardrails.

36:42You don't want people vibe coding, whatever, and then jeopardizing the whole security. Especially in places where data is paramount. Let's say biotech or anything like that. Even those guys, they don't want the frontier LLMs knowing what they're doing. They're going on-prem. So do expect a new wave on security paranoia, which is good. And by the way, in my previous career, if you will, I was very paranoid about security. and like those first use cases of, you know, bad actors getting hired to then, you know, get access to confidential information, it happened in the space I was working on before.

37:33Like I've been targeted by, I think, I'm pretty sure, North Korea agents previously. And I'm pretty sure a lot of people in this call have been, but they have no clue they have been. So, and you can be sure that they are using AI. any AI tool that they can get access to to break in. Just my last thought will be there's an asymmetric approach to offense and defense in here. So offense always takes the lead. The red team always takes the lead. And I think AI has kind of widened the gap a little bit, but at some point it's going to shorten up again. So honestly, I'm not sure if this was what you expected from me, but I think it's same old, same old.

38:22Pedro, I expect nothing but long monologues from you, mate. So yes, you did indeed come in and deliver that. No problem. No, but great. I love what you're saying. It may just feel different to us, but actually it's kind of similar, right? It's like there's always been somebody trying to say, look, don't do this. Do you remember when we had to hand mobile phones into the office? And very early on, I mean, the internet was banned. I'm old enough to remember where the internet was banned because you can only use the email. What are you doing on a web browser? Do you know what I mean? So all of this craziness, maybe AI is just the next generation of this.

39:00And we're going to get to the point where we treat AI much the same way as most of us have grown up with the internet, free to use. But there's certain websites you don't use and there's certain things you don't install, right? In the same way we kind of all intuitively got to that point, maybe we'll get there with AI as well. Okay, we've got Sai here as well. I'm going to bring him on stage. I want to get the view from... He doesn't want it either. I can see the grimace on his face, but it's too late, my friend. He's on. Sai, unexpectedly. Good to see you. Yeah, can you quickly introduce yourself, Sai?

39:35Who are you? What is it you do? Yeah, I'm Sai. I'm the founder and CEO of Crowdcast. I've been working on this for many years. and I'm also an engineer. Fantastic stuff. Wonderful product. Great to have you on the show. The reason why I wanted to bring you on just serendipitously right now is because you are the founder of this business. And you mentioned in the comments there that AI has been great for activating and kind of improving the productivity of your team. How have you kind of balanced this between sort of the engineering folks that you probably have, I guess, full license, a greater license to use AI compared to some of the non-technical people where, I don't know, do you have a different policy, for instance?

40:20How do you kind of think about AI policy within your business? Yeah, I mean, funny enough, it's been a process of trusting the AI more and more and being surprised at how much can happen when you give it more tools that you previously would have been too scared to give to. uh like these days actually a lot of my job is designing the agents that the team interacts with and deciding who can ask that agent to do what so for example we recently rolled out like a shared agent that is in slack that has like even production read only access to to our servers to be able to ask you know the questions that you typically need like a like a data scientist right to answer.

41:06And so that's been a huge unlock for us at the team is now our support engineers or support team, non-technical people can just ask, hey, super smart engineer who's always available and never gets annoyed. Can you look into the session and see what happened and investigate? And that AI can dig through many logs, thousands of lines of logs to identify the issue without having to bother like a human, right? And so we don't, we also don't want that support engineer to be like, hey, can you delete all of our users and have the AI actually try to do something like that. So learning how to give the right guardrails for that agent has actually been a lot of like the way that I've been sort of thinking about working with AI these days.

41:48And I think this is kind of the direction a lot of companies are going to go in the long term is not just designing the applications but designing the agents that the humans are using to build the application. Yeah, very good. a brilliant example by the way hopefully this means things like you know bug fixing is going to be super super fast right so suddenly you remember like when you have to report into a so anybody using product uh a product uh you try and report a bug fix you'd have to do screenshots you have to wait for someone to review it someone's going to review it and say oh we'll get back to you and then you think okay how long is this going to wait like i don't i don't really care that much truly.

42:25And then suddenly you're in this like never ending sort of pipeline where you just don't know or anticipate when it's going to finish. Now, I guess with AI constantly being able to do it and be diagnosed with high speed, you can get much quicker resolution. Fantastic example, Sai, but keeping with you for a second, tell me about how you treat vibe coding in the business. So I guess this is different in the sense vibe coding for me is kind of like undirected use of AI. Like your CEO is not saying, hey folks, Vibecode this. It's more like, okay, everyone's getting worth the job and it kind of goes ground up.

43:03So someone operationally, a salesperson, support person actually just produces something by speaking to an AI. How do you sort of activate this, but also make sure it doesn't sort of put Crowdcast at risk? yeah so there's sort of i guess three levels to this right um there's a an employee vibe coding something that just helps them as a tool uh just to do their work a little bit better there's currently no restrictions on that so if you feel like you want to take on that that i guess that project that'll help you out great it does take some some level like engineering mindset right to still kind of think about how to build something So I find that most of our not technical team doesn't actually like an app or something like that, but they might have an idea for someone who's a little bit more technical to help them do that.

43:57So then the level two would be, what if we created a tool that's easily bi-coded for the team to kind of use? And so we have an internal admin tool that anyone on the team can access for support related stuff. And so if we need to like create a tool there, that's pretty low effort. like just go ahead and you know have someone add another page there maybe we can see a little bit more data on what plan this person's on or something like that um so that's also a pretty low bar it doesn't affect production it doesn't affect customers um but yeah when it comes to like vibe coding something that's like on the app itself that everyone else is experiencing that's when the bar is the highest and only our engineers can do that um and then anyone who's in design is like kind of the final site like sign off before it gets to the customer side but But yeah, I think it's really about what's the exposure?

44:50Is it just affecting the individual or is the rest of the team getting affected or is the customers getting affected? So just being cautious about how wide that impact is. Yeah, it sounds pretty straightforward, doesn't it, folks? It's got a lot of common sense stuff here. It's almost like, OK, let's not overthink how big a challenge this might be. It's simply, okay, you can use this, but what's the impact on others? And the wider that circle becomes, the more sort of policing and guardrails and permissions you might need. Or is that right or wrong? I don't know. Let's bring on George LaRock. I think he's got a particular job to do here, George.

45:33George, good to see you. Good to see you. Thanks for being patient and waiting along. I just want to firstly introduce yourself. Who are you? What is it you do, George? George Laroque, I'm the founder of WorkTech. We do market intelligence and advisory across the HR tech and work tech universe. And so what does that mean? In the last week, I've talked to founders of, I don't know, 30 startups, looking at, you know, what they're building and where it's going. Brandon's one of them, by the way. I've talked to the chief product officers and CEOs at some of the biggest platforms in the space about what are they building?

46:16What are they looking to buy and where are they looking to partner? And what's the impact of all this on them, right? The conversation that we're having and others. So that's WorkTech. That's me. Yeah, fantastic. George Lerock, by the way, must follow on LinkedIn, must follow also on his sub stack. Make sure you do that. Also, great YouTube channel, by the way. Lots of great interviews with founders on that. So follow all of those sort of channels, please. George, you have commentary to make on this concept of vibe coding. And the title of this show is, look, Vibe Coding TA Tech Tools. How viable is it really?

46:56What's the answer, man? How would you respond to that question? I think it's viable. I think what everybody before me here has done is I was so happy to hear you've got such thoughtful guests, such experienced guests that I mean, that that's my message to the market is I look at it as you talked about that circle. I call it, you know, if something goes wrong, what's the blast radius going to be? So is there, you know, PII involved? Is there candidate data? Is there sensitive corporate data? You know, what are we building? And I think everybody's done a great job of explaining the difference between, you know, internal tools helping me or my colleagues and I be more productive versus thinking that I'm vibe coding a new platform to, you know, get involved with candidate screening and selection and so forth.

47:53There are some lines that we just, you know, can't cross. So I think it's viable at the right time, in the right place with the right data and security. Yeah, I think so. I mean, basically, I kind of, again, we're talking about the size and scope of these tools. We also, by the way, have a very prominent commentator in the chat, Dan Georgescu. Dan, thank you so much, by the way, for all of your commentary in there. And I'm going to try and bring Dan on the show. So make sure you get your webcam on in a sec. Dan, you are saying you've actually done all of this. So basically, Bill everything from basically has replaced the SaaS stack with his own Vibe code.

48:35I think that's viable. But is it a question of scale? Like, for instance, solo recruiter doing it himself. You could probably get away with that. No different really from, I guess, using Google Sheets in my day, you know, building some sort of database. I used Google Sheets actually when I was recruiting Andrew back in the BNT. No ATS. So, yeah, we did that. But obviously the world has changed. You can't do this in the same way. You have better tools, but you also have, I guess, more legislation to be careful of. Now, I want to also be, I don't want to expose anybody here, but in reality, when you're sort of building a vibe coding platform and you're putting in sort of personal data and stuff, what is the, I guess, what's the compliance sort of things that we're breaking here, George?

49:31Well, I mean, think about GDPR, right? And I think it's not about the size of the organization, but the context of what we're dealing with. So that's a great example. A recruiter working on a freelance basis, they can go nuts. And when something goes wrong, how many lives are impacted, how much data is impacted, they're not going to have a work council show up in the office as they solve their problem. But at a large enterprise, you know, we're talking about a very different context, a larger, you know, bigger risk, bigger issues. I mean, there's a lot to think about. And most of the enterprise platforms in our space that are, you know, doing, they're innovating and they're on this.

50:25So like if you think about MCPs that are putting out servers that are being put out, those come along with all of the security guidelines, all the roles and access sort of guidelines that are in the SaaS platform that they've emerged from. So if you're a recruiter at one of their customers and you're using Claude to build an app, Vibecode something, what you have access to and what you're able to input to is going to be the same as if you log into their web interface, if you will. So I think we're progressing along this. At the same time, if you're, you know, those large number of startups, I run the pitch fest for HR tech.

51:13And I'll tell you, I've looked at so many clawed websites in the last two months as I've evaluated about 100 startups for this thing. And a lot of folks that didn't make the cut have basically vibe coded a pretty cool product. But like I hear questions like, you know, what is what is SOC 2? What is that? Tell me what that is. These are the things that we can't put in the enterprise, but we can allow responsible recruiters to develop cool tools like Aula's with the right controls and access. Yeah, very, very interesting. I mean, the vision I'm getting from all of this conversation, folks, and I just want to get everyone's sort of view on this, is that we might be seeing some core platforms at enterprise that are going to be still SaaS vended, right?

52:05They're going to be core tools, you're not going to get rid of them. They've got to be compliance ready, they're going to be multi location, no one's going to vibe code that. But they will be more flexible to be interacting with a kind of unimaginable explosion of little vibe coded apps that everyone within that organization might then want to interact with in some kind of safe way. So you can just imagine, let's say, a global organization, 100 locations, 20 ,000 people, employees, whatever it might be, core platform, they may have it, but how many apps might actually be existing in their universe?

52:43Maybe 10 ,000, I guess. That will all be very small teams, single use, maybe. They'll pop up in and out. Maybe they'll continue to land and grow, as Glau, you mentioned, and turn into something bigger. Maybe there's a gating system where over time, maybe it gets elevated into something else. Perhaps engineering, take a look at it and say, okay, let's build this properly, quote unquote, with all of the, you know, I guess the robustness of enterprise grade software. And then it goes out into, it sounds like a very happy sort of situation, doesn't it? Can anybody kind of put a disaster scenario to this view?

53:23I mean, is this overly optimistic? what's your take um uh brandon amruzzo let's go with you yeah i think i was actually speaking with an agency owner a couple weeks ago who built everything himself um but his organization has grown and scaled quite a bit over the last two years now he's managing you know four recruiters underneath him and what he told me is you know 80 of my time is now not spent recruiting it's meant maintaining and managing the systems that I built to run the company. And so he reached out to us to see, and it actually was a helpful exercise because he's already built it all himself.

54:00So we literally just took what he has today and mapped it into Scalist to see where it fits one-to-one, where it doesn't, what he could build around Scalist to help augment some of the gaps. But he is done having to be, you know, play engineer and run all the maintenance and get taken away from actually being able to support his clients. So I think that goes back to an earlier comment where, you know, if you're like a solo recruiter, you might be able to, you know, get away with it. But as you continue to grow and scale that organization, it becomes too cumbersome to maintain. Really good point.

54:31The more vibe coding you do, the more effort it is, you're going to have to decommission a lot of that. Glau, I think that's true, isn't it? Like, if you're very enthusiastic about sort of vibe coding, you might just end up with 100 tools. And before you know it, like the mental load of trying to just figure all that stuff out, you think, you know what, I don't need that for this. You got to clean it out. What's your take on sort of the future as it looks like of how the tech stack is going to be. If you could project forward, you know, let's say 12 months, how does the tech stack look architecturally for you?

55:04Yeah, I like to think that I'm both pragmatic in a sense that there are things that are not possible to be done today just because of how companies work, how red taping works and how data processing essentially works. That's law, not something that we can do anything about it, right? But on the other hand, I'm also optimistic in a sense that I believe technology will continue to evolve. The capacity of what we can build and how data is stored and how everything is safeguarded will also continue to evolve. My personal taking here is before you build absolutely anything, also connect what Allah has said.

55:41Connect with your team. Understand exactly what you're trying to build for. Make sure that if you're going to be building a tool that's going to be part of your stack, of your day-to-day, that it is easy to maintain. And on top of this, this has credibility amongst your team and it fits within your processes. It's not something that's going to create more workload for you because otherwise, instead of becoming something that's meant to support you to be quicker, to do your job better and so on, you're just going to create a snowball type of effect. And then on top of this, this will make people that are seeing your efforts into building your own tools be like, oh, but maybe let's stay away from AI.

56:17Too complicated, too cumbersome, not for me. Yeah, yeah. Really, really interesting, Glau. I love that. I love that view of it. You can't let us get away with it. You can get over enthusiastic about the tool building. But don't forget, you know, you've got a job to do. and these things are meant to support the outcomes that you're there to deliver against. Okay, good stuff. How about you, Pedro? How do you see the future of the TA Tech Stack from what you kind of know about the industry and software development in general? Tolman's down the line. What does it look like?

56:56I think one recruiter can do the job of 1 ,000 in the future. Of course, it depends on the recruiter himself or herself. Honestly, what everybody said, I agree with it. We're going to have different capabilities. You can build your own. You can buy them. There's a maintenance cost to them. But honestly, you're amplified. As a recruiter, as a TA professional, you are amplified more than ever. It's insane the amount of things that you can do. With a caveat, and I don't want to go on the monologue here, but with a caveat that there's a lot of stuff going on in the market that, you know, like AI driven applications is through the roof.

57:43It's insane. So even though you're amplified, now you have new types of challenges to deal with as well. Yeah, quite right. And yeah, we are getting better. I think some of the stats that some of the ATSs have come out with, it seems that yes, there's less recruiters kind of currently doing the job than they were three years ago. But actually, sort of the placement rate is generally similar, even though the volume of applications is high, the number of interviews is high. So the more work is being done, but we haven't actually not delivered the candidates. So we are doing more or less, even though it doesn't sometimes feel like that.

58:22Okay, let's go to you, Andrew, from an engineering point of view, hearing what you've heard so far and what you think is going to happen 12 months down the line. Everyone is going to get even more ai enabled we presume uh 12 months down that pipe how does that affect software development um uh going forward the relationship between engineers and non-engineers i think um maybe i kind of want to reflect on what brandon said before about the the one uh ta whose team grew and now he's he's like supporting all this like this system that is is quite large now i think that like the the The problems that AI will be solving for in 12 months time are much larger and much deeper and much more able than they are today.

59:04So like that one person, the stuff that he's built will be able to do much more. But it doesn't mean that the effort that he has to maintain it all is much less. It's just now the system he has built is much more capable. And maybe instead of having to hire those four people, he might have been able to hire just one more person. I think that kind of like scales up a bit more broadly. I think that the stuff that I think if we think about the stuff that AI is able to solve for today I think that that's too I think we can come out with a very optimistic view on like the kinds of things that are true and that we'll only need to like you know if we vibe code the gaps in between all these SaaS platforms that's it that's like that's that's us done and I think in 12 months time that's going to be less true I think that like the gaps in between become much broader because AI can fill much bigger gaps, which means that maybe there are fewer SaaS products or SaaS platforms that do fewer things, but they're still there because they are really robust and AI is still not able to solve for those problems.

1:00:03So it's like the gaps become a bit broader and AI is able to fill those gaps a bit more. Yeah, it's really interesting implication on the SaaS model, isn't it? I mean, there is this general, one of the conversational threads in industry is that AI is going to lead to the end of SaaS. probably not, but it will kind of damage some elements of the SaaS world, particularly maybe the entry level, I guess, you know, the entry level stuff, the stuff that you would use, as I said, you know, you need to get off Excel and just improve something. Maybe that is an area of SaaS which gets eaten a little bit.

1:00:37Science go to you. I was going to say, sorry, if you just think about like the things that AI is able to do today, sure, it can like quite simply replace some very simple SaaS tools, but I think that like the complexity will only increase or the complexity I was able to solve for will only increase. And so I think if we only think it's just the small edges of SaaS that disappear, I think that is very optimistic. I think it will not be quite that optimistic in 12 months time. Right, right. Very interesting, mate. Sai, go to you. You're leading a tech company. how do you see the sort of 12 months down the pipe in your own internal sort of tech stack whatever it might be doing CRM stuff doing sales stuff doing all that kind of work that you have running your business do you see sort of the vibe coding and self-coding stuff kind of increasingly replacing more of those tools or do you see actually no some of those tools are going to be fairly resilient and we don't have a program to delete them yeah like like i just mentioned in the chat um i think a lot of the vibe coding stuff is really around not adding new apps to the business um there's already like so many tools that any company's navigating so the if you can just kind of unify it across a couple of different solid ones and use the vibe coding skills to link them together that i think is really the main value like we're a notion based company so our company brain lives on notion and notion is very flexible and if you just add a little vibe coding you know with notion and connecting with your emails and things like that you can pretty much build any application just on top of notion you know so um a really thorough crm can be built on top of that so um yeah so yeah i do think in the long run um yeah specialized uh apps might actually have a it's kind of interesting my sense is that the era of sass that specialized might actually be the one that's going to get the most you know difficult time to grow as sass companies that are more generalized and more like primitive blocking like blocks that you building blocks that you give people those tools might actually become the ones that are ultimately the ones that people are building on top of most.

1:02:57That's kind of the direction that we're also seeing our company going. We're starting to build more and more things so that Crowdcast is becoming more of like a platform as opposed to just like trying to solve one specific problem. And I think that's the general trend that a lot of other platforms are going to go to towards as well. Yeah, very interesting. Expansion of scope. Like the AI is basically enabled everyone and everything to expand what they were able to do before. So we're going to investigate those new territories. I see kind of an ecological metaphor. You know, it's like the new land has opened up and some fish are going to crawl out, you know, and colonize those spaces.

1:03:34Okay, Ala Pavlova, last comment to you. You're a recruiter still, right? You're still doing the job of recruiting. You've got all this vibe coding going on. Where do you see sort of, I guess, the tech stack in tournament's time, the TA tech stack in tournament's downline? are we going to go to the level that Dan has gone and self-build everything or do you think certain core tools are going to stay maybe even certain point solutions are going to stay how do you kind of feel it's going to end up for us until this time thanks so much I believe that Vibe coding is giving us the opportunity to close some of the gaps and you saw my example was just closing one small gap and so before I had to wait for the engineers to do something.

1:04:22And now I can do a small like MVP myself or try to demonstrate the problem so that I can get a better solution with maybe like engineering teams. I see that our TA can have more opportunities to showcase the problems and maybe be the first one to initiate the change and then engineers can help out. I do see biggest challenges recorder right now. I get thousands of perfect CVs. Thousands perfect keywords, perfect job description matching CVs. And I do think that I see myself as a first person facing this problem. And I hope that the tools that we are given will be flexible for such things like me adjusting to my needs.

1:05:14Like if I get like a thousand CV for one job and one day it's actually happening sometimes in some regions. I need to be able to probably just, you know, have this webcoding small solution to solve this problem on my level, I would say, maybe to go further. Yeah, I just, with webcoding, I also don't need to wait for somebody to help. I can just go and test things and be proud of myself that, you know, I did like a small thing, but it's worked out. and change something in the whole organization. This is how I see it, yeah, as a hands-on recruiter. No, I think it's such an exciting time to be a recruiter, such an exciting time to be an engineer, even though it is trepidatious because the change is happening whether we like it or not.

1:06:09So brilliant conversation, everyone. Thanks for joining. It's been a fantastic learning experience for me personally. So I want to thank everyone on the panel for coming on. It's been brilliant. Sai, particularly jumping on without solicitation, really appreciate you coming on. And congrats on the tour, by the way. Fantastic stuff. And everyone, thanks for watching. I hope you've enjoyed the show. No show for us next week. I will be traveling. I'll be back in Nashville next week. I'll be in Tennessee. So I'll be seeing everyone there. If you're going to Wreckfest, do let me know. Come up and say hello.

1:06:44I then will be in Mexico for a couple of days for a friend's wedding and then I'll be back in London after that so yeah a lot of bouncing around so no time to do a show outside of an airport so I didn't fancy my chances of giving that a shot so we'll be back though full schedule in October so make sure you follow the channel if you're interested in this type of conversation we usually bring it to you every Friday no fail okay that's it um good to see uh see everyone here and we'll see you next time take it easy thanks for watching thank you thank you