Absolutely.

Yeah.

Well, at least it's saying something.

At least it's saying it's being recorded.

Yeah, exactly.

Well, I guess, like, everyone always uses, like, granola and stuff now, right?

And, yeah, we're gonna use that, too.

But,

yeah, well, that's actually, you know, granola was where the, the kind of the frustration also came from because there's all these really great tools, but it's just after the fact, you know, you've, you've already asked the questions and all of that.

But yeah, sorry anyway.

No, no, no.

Yeah, so Reese, my first question is when you've been building with Voice AI.

If.

There was one thing that you could wave a magic wand at, what would it be?

Well, in general or specifically around

trying to use your product or just in general?

Just in general, don't worry about layer code.

We're just trying to understand where your real anything that you would wave for Magic Wand that basically.

Well, I've had a scene for a very long time in different contexts, which is context is king.

And I think that really applies when you are building with AI more so than anything else I've ever known.

It's taken me quite a while with all the tools that I have to start to build to a place where there is enough structure, enough context, enough background,

proper documentation, and then keeping all that documentation up to date and making sure that the AI's got all of that context.

You've got the agents.md set up correctly and whatnot.

All of those things I think has been really key beforehand, it was cool to be able to start from the point of view of code in and then get to that point where you're able to build something, show that it's great, and then you prompt something and it breaks a bunch of features and all that.

It's just like that typical thing.

If I had a magic wand, I would just love

the tools that we're using to be a little bit smarter at being able to capture and document all of that.

So it's less reliant on us.

I'm pretty sure it's gonna, you know, it'll get there.

I'd be amazed if that, if that wasn't already happening.

I know it kind of already is happening to some degree, but I find, like, when I use replit, which I'm moving away from at the moment, but it has that the replit.md file, but it doesn't really capture the things that, that you really need.

It doesn't really remember.

The architecture and the decisions that have been made and the direction that you're going in and all those other things that you really do need when you want to just make a change somewhere, but it doesn't end up breaking something.

Or.

Taking too much agency itself.

So it's kind of around like the,

it's like making sure that the tools you're building will have the context.

So like what you've built so far and how the tools work and stuff.

Yeah, and how you want to work as well.

I don't want the agents to go and have complete autonomy over creating lots of tasks and making decisions in the background.

I wanted to have some autonomy and some agency to go and do things, but I really do want that cycle at the start.

I'm just making sure we've got the plan correct and making sure that the right decisions have been made.

I do find now I'm just moving across to using Curser.

And I'm getting it set up.

I feel like that's really helping.

It certainly is referencing the agents file a lot more than Repl.it used to as well.

But yeah, I feel like that would really help to not necessarily speed things up because speed really isn't the problem anymore, but actually just reduce the amount of issues that might come out at the end of something.

'Cause sometimes it introduces something and you don't even know that actually that's an issue until you're much further in development somewhere else.

I find that's the problem is that I'm picking it.

When did that happen?

Why did that decision get made?

And it can be infuriating to have to go through that.

Yeah, sometimes it just deletes things that it really shouldn't have.

It just removes logic that you had in there or something.

Yeah, because it's just focused on that one task rather than Always having the overarching context of this is always what we're trying to build.

Here are the different decisions that get made and why we've architected it in all this way.

No, you can't go-- it's just recently with them, I was building a real-time gateway for streaming because what Ankit is also doing is that it's building UI where it's able to prompt you with questions in real time for how to move the conversation on, get the right insight, or get more evidence and more detail.

Something in particular.

And so we built this real-time gateway,

the SSE stuff in the background, and it was great.

And then it just completely broke it because it changed one of the key components with it and completely changed the logic because I wanted to change something in the UI.

So it was only when I actually dig down to detail, ask the questions, I was like, well, I don't understand why you made that.

And then it does the usual, you know, AI thing of, that's right.

Yes, I know it's right.

But it's wrong.

You're absolutely right.

I admitted this.

Yes, he did.

Yeah, that's, I relate a lot.

This is, yeah, so true.

Yeah.

Is there anything else that you would wave a magic wand at?

Asking the question again.

Is there anything else that you would wave a magic wand up in build while building with voice AI or.

Just.

Creating voice AI powered products?

Well, you know, one thing that I think you guys have done really well is that I noticed that you, you've got the, the little button.

In the documentation to be able to download everything in markdown, which made it remarkably easier for me to just be able to take that and give it to CURSA so that we could build the plan to integrate all of it rather than pointing it at different websites and hoping that it's able to scrape all the right information and again, make sure that it understands the context of everything.

So, you know, I think more people doing things like that is really good.

More sort of common integrations and patterns.

I see people moving towards that again now.

It's funny, I'm leaving Replit and one of the reasons why I was leaving originally was because it didn't have any real integrations other than forcing you into using Neon and these other things that it's got.

But now it's got all the integrations.

Actually, I would have taken advantage of if I was still there.

Yeah.

Yeah.

I'm not sure otherwise other than that.

I

feel like I've got, there was a lot of issues like going through it and I found

workarounds that have probably now just become sort of so standardized to me that I don't even see them as problems anymore, maybe.

I see.

Yeah, I know what you mean.

It's like, don't even notice the duct tape stuff anymore.

Yeah.

Yeah, I think so.

I'm pretty sure that, you know, as we continue the conversation, things will start to pop into my head.

Yeah, well, if it also, if it's not like a burning problem, it's not a problem, right?

Like, if it's like a, if it's a workaround but it works and it's not causing you issues, it's like, that's fine.

That's how it's supposed to be.

Yeah, that's super interesting.

So it sounds like the stuff that you're getting that's most challenging is actually kind of like the AI powered development, like working with the agents while you're building voice AI products.

Yeah.

Well, just yeah, trying to maintain speed, but also

accuracy, quality.

And, you know, at the back of that, I think those are the real key things.

Yeah.

If you had better,

if that kind of worked better, how would it change your life?

Well, I mean, I have a huge amount of ideas and things that I'd love to be able to implement and the ability to I basically, I have one day a week off work where I dedicate to working on Anika so I can do whatever I can in that one day.

And normally it's a lot, certainly compared to the other job, the amount that we could move internally versus how I move on a Friday is absolute night and day.

It is quite actually, I find it really infuriating a lot of the time now.

See how slow everything is and how difficult it is to try and make everything move.

And I'm like, I literally built that in like two hours.

Same infrastructure, same issues, you know.

But yeah,

I have this fight one day.

I think if AI

had more of that context and that ability to be able to move faster with me, naturally, I think we'd be able to move the product on a lot faster and develop more at speed.

Super cool.

Super cool.

And is there anything that's changed over the last year, six months that has

made it more valuable to be able to move faster and ship faster now?

I think it's probably just the speed at which the market is moving.

So you have to be able to react.

I certainly feel like I'm on a

timeline with a big wall fast approaching where you kind of have to get not just a demo but something that is an MVP

either to market or to a place where actually you can really start to show it to lots of people to secure some funding to be able to make it more of a reality and whatnot.

So I I do think that being able to really get all of these tools working in a way where I can significantly build and move the product forward and react to the feedback that I'm hearing and collating and whatnot is vital.

And I don't think I'd be able to do it at the same, oh, I wouldn't be able to do it at the same speed.

It's called it a year ago, I think six months ago, probably still be in quite a good place for it, but a bit like a year ago.

Yeah, I don't think I probably could have built the product a year or 18 months ago in the way that I have now, certainly to the complexity that it has, considering that although I had some qualifications way like 20 years ago, it's got to be now.

I haven't spent my whole life

in tech.

I was a learning professional by heart for many, many years and then kind of got back into tech.

It was always my passion, really.

So yeah, I know enough to be dangerous, I would say.

I think that makes me a good shoe product officer, certainly one that looks after engineering, but not enough to go and code from scratch my own products, you know, and do DevOps and other bits and pieces as well.

Now that's an area actually, now that I've just said that, I think that's an area of

that I need to explore a little bit more.

Again, something like Replit makes it really easy.

I just deploy.

But I'm like, well, now I'm moving to Casir, what am I going to do?

How are we going to do this in the future?

I think that's probably another area that's really ripe.

Or maybe there's already quite a few products and I just don't know it yet.

But it could really help people like me to make sure we've got the right setup for DevOps and deployments and whatnot.

Yeah, that's a big space.

Yeah, future problem for me rather than an existing one.

That's one of the advantages of using Replit.

I don't like using Level up anymore.

I think it's too simple.

It's too focused on people that just want to create demos and prototypes and all of that.

But Replit does have that enough to be able to really get going.

That's very cool.

Yeah, I need to play around with it more.

What's your setup?

Are you already in Cura?

Are you already using...

I was using Cura, but weirdly, I can't remember why, but I think I went back to VS Code recently, and I'm just using Codex, the

OpenAI version of Claude Code, basically.

Yeah.

And we really like it.

And also the other thing that is quite good is we've got an integration in Slack.

So we could say, like, Ask Cursor.

So even people that don't want to go, like, people in our team that wouldn't want to go mess around with code because it feels very like CLI, they can just say, oh, can you just add this to the docs?

Can you mention this?

This link is broken in the docs or stuff like that.

Or like our website's in Framer, but they could do it on the website if like that was a thing.

And so, yeah, I really like Codex and you can use it on your phone as well, which is really cool.

Yeah.

Yeah, I use Codex for code reviews.

You know, I do really like it.

But I used to appear so got like, what they call bug bot now that they've started to roll out, but I really do like you can use it in Codex in that way.

Yeah, maybe I should go back and check out Cursor again.

It's just like, I think the subscriptions were set up at my layer code for Codex.

And I was paying for Cursor personally, but then I was like, okay, I might as well just use Codex.

But yeah, I should go back and check it out.

Sorry, one last question.

I'd love to see a demo.

How do you discover new tools?

How did you discover Replit, for instance?

Well, I've been setting up my LinkedIn for a long time to subscribe to a huge number of people that are posting about their very latest things in the field, what they're doing with it, a lot of people that now kind of share any 10 templates and videos and demos of what they're doing, that's really interesting.

And I think with my role, especially in a fuse, I'm also trying to make sure I collate all this stuff so I can share it with all the different people internally.

Hey, content team, this is what's happening with day 3.1 has just come out the other day.

They've already got videos and information about it so that they can start to go and explore all of that.

And so I really use LinkedIn as that kind of key place, obviously product hunt and quite a few newsletters.

I think

that's where I see all of these new tools come out.

And that's a lot of the time where I'm actually then going and subscribing and going on to a waitlist and all of that because I'm like, yeah, this is cool.

This could be really interesting for me or for Fuse or whatever.

I will just dive in and and give it a go.

I've done it for loads of stuff.

Any particular newsletters that you like?

Well, I mean, we've got all the usual sort of product ones like Lenny's Newsletter and whatnot.

I'd have to go to my email.

I

wondered if there's any that you really check.

I guess Lenny's is a good one.

It is a good one.

I wouldn't say it's one that

that normally kind of Services, the, the more interesting stuff for me.

But, yeah, I mean, if you are interested, I'd be happy just to, like, get some, get some things, you know, and I can share those with you.

Oh, yeah, that would be great.

Yeah.

But, Reese, I'd love to, if you want to, I'd love to see a demo if you're, if you're, if you're up for it.

Yeah.

Yeah.

Well, sure.

Like I say, I mean, hope, you know, to see whether it It works on that.

I think I'm still in that really sort of nervous new stage.

Oh yeah.

Well hopefully even if it doesn't work, hopefully doing it to me will make you less nervous next time when...

Absolutely.

Yeah.

Get the habits in.

No that's fine.

Well I will share the tab here.

So yeah, I mean, so like I was saying, I mean, with Annika, we started off with, you know, being all about how do we help humans be the best they can be in meetings?

That was the sort of the seed of the idea.

And then where we've gone from there then is, well, actually, you know, it's a much bigger problem than that.

That is part of it, but actually, as a human,

you can only attend so many meetings anyway, right?

I mean, how many meetings can you do every single week with with prospective customers and whatnot.

So what we're building is that a tool that can help you within meetings, but at the same time, you can also then actually go and do autonomous meetings.

So it's not just about doing meetings per se, it's actually about discovery projects.

And where I'm starting to aim this right now is solo entrepreneurs, kind of like myself.

Well, I've got a co-founder, so maybe that's not the right title for me.

But people like me that are able to to get a real product, test it, get it out there, be able to react to feedback, but also people that want to potentially be able to go and get investment.

And so being able to create a project where you can have a series of meetings, some that I do, like maybe they're more in-depth conversations, and then you've got the conversations where you just need more information, maybe it's shorter five, 10 minute conversations, and Anika can actually go and do that and have that meeting.

That's, I should say, roadmap, not really going to see today.

But the autonomous bit is where we want to go and then add different tools into it, surveys and whatnot, so that if you are looking for validation for your idea to take it further, or for your company that's got seed funding and you're going for Series A, or you want to get new, you're trying to get new revenue streams but you don't know which one to go for, to be able to go for Series B, Series C, it's also a tool for that.

When you get into it, and this is the bit that I haven't got installed at the moment, it should be Copilot Lite, which is the browser extension.

Rather than you seeing this screen here, well, I should be having a conversation with you in the browser.

Then on the right-hand side, I would then just see Anika, and it's able to tell me what are the questions that I should be asking.

You go through a process at the start and this is where layer code is going to come in now where you prep for the projects and then for the individual meetings and we want to use layer code so that Anika has a five minute conversation with you to build, you know, what are the, well first of all, what are the goals that you actually want to get out of this particular meeting or this project?

And then within those goals then, what are the objectives of those?

So when we take these things off, we know we've actually then completed that.

And then what Anika is doing is it's having the conversation is it's, it's, you know, there's like a four or five different agents in the background, triaging the transcript as it comes through, it's then passing it to the analyst.

The analyst is then, you know, looking at evidence and updating, it's passing it then to the, you know, the interviewer agent that's then generating and reviewing all the questions and whatnot.

And so in like near real time then it's continuously updating and helping you to be able to make sure that you get the very best and you see how you're doing in the meeting.

It's got all the reason and the logic in the background then to say, hey, you've only got like five minutes left, but you need to maybe ask just this one question.

It'll just get rid of everything else and just say, here's the question that you should just ask next.

So like I said, it started from the seed of the idea of it being about the meeting, but actually for me now, the meeting is just one tool.

In a kit that we're going to be building out.

Voice interaction, I think, is the starting point for this.

So the onboarding flow would be speaking to Anika, what are you trying to actually do?

And then from there, build a project plan that then generates meetings, goes and looks on LinkedIn, finds the very best people that you should probably speak to based on the project that you want to do, hopefully connections of yours, so you can just go and send them an invite.

You know, and then either you turn up or annika can turn up and, and do all of that and send out, you know, surveys and, and other bits and pieces.

Yeah.

Super cool.

Yeah.

So I, i, it's, it's not, you know, i'm sure you the ui, it's not, not actually updating or anything at the moment because this is just a demo, like, meeting that was, was run before, but, but, yeah, I mean, that, that's what we are.

That's what we're building.

At the moment, you know, I want to just be able to help people to be able to validate really quickly, you know?

So rather than spending weeks and months having meetings with people, you could maybe do projects in days or weeks instead.

Yeah, it's crucial.

Yeah, I feel like this is the thing that startups fail at the most, right?

Just

working on something that no one cares about or like kind of just isn't pressing enough or something like that.

It's just Yeah, just like that.

Well, that's the other angle of it as well, isn't it?

So if you just continue to get, you know, you probably won't get negative feedback per se, but nonchalance in meetings and whatnot.

People are just like, yeah, that's cool, you know, but actually there's no buying signals, you know, so we've been doing meetings and I, you know, and it's been prompting me just to say, ask whether they would commit the company's money to it.

And I have, you know, and people are like, yeah, we would, we would do that.

So it's good.

And actually, even it pushes me in the meeting, which is quite cool.

We've had it up and running and set up properly.

Oh, yeah.

That's like a kind of.

Yeah.

Like the co-pilot there for you.

That's you.

Yeah, yeah, exactly.

It's like it just keeps nudging and just saying, well, it's always trying to find buying signals, you know, and it's based on sort of different product discovery Frameworks in the background.

It's not like, oh, you know, like, we're going to use this framework and whatever, you know, it's fine to talk.

So it keeps it really to the business, but in the background it's all structured and thinking about how do we map out things like jobs to be done and other elements within product and capture all of those so that it's then there.

And

then the bit that comes after that then is the insights cloud, which is taking all of these different insights that we've got and then being able to put those together, first of all at the meeting, but then at the project level.

So how do we start to build and show all of the insights and summarize how your project's going or what we've concluded so far from that.

I think the bit that then becomes really interesting is what if you take it up another level?

And so it's not just insights for projects, but it's insights for product or insights for a company potentially because it doesn't just have to be about discovery.

It could be sales and customer service and all of that in the future.

That's where I think it could get really, really interesting is if you can really see what people are actually having conversations about and all the insights you're gathering for customers at scale.

I think that's always the thing, right?

The moment that you put its effort to set up an app, install it in a dozen different places, get everybody to use it and all that, I think that's the difficulty.

So it's kind of working out how you just make it.

That's just hassle, isn't it?

But if you just turn it on and it starts giving you feedback and insights, it's, yeah, no problem.

Absolutely, yeah.

Well, I mean, even with the codebase a bit, if it's, let's say, at a company level, if it's installed at the company level, then we can just autonomously grab the transcript that gets generated if it's like Google Meets or something.

I also just think integrations into the other apps.

Like you say, you know, you mentioned granola earlier on.

Like, I use granola.

I've got loads of people in my company using it now so that we can pull things together.

But if you can just grab all of that and bring it in, it's just more, more insight.

Yeah.

Because I guess granola is just kind of sitting there right now, mostly.

Like, it's very rare that I'll go look at the granola, you know, like it's, I've got it, but it's like, maybe some people will look at it more because they.

Need to do the follow-up calls and stuff like that.

But

for the average person, I feel like it just goes and sits there until it, I guess until you maybe need something, you couldn't remember something, but, like.

That'S how I use it as well.

I mean, that's the same.

It's, you know, we've caught all of my meetings and work, but now and again, I will, somebody will say, hey, we need to go on a call with this customer.

I'm like, great.

Well, let's see what we've spoken to them about.

And we can just see everything and just have a conversation with all the transcripts then, you know, to make sure that we're fully prepared.

I think stuff like that is really good.

It's not quite what we're trying to solve.

That's not the same thing.

No, no, no, no.

You're more about surfacing it.

Sorry.

I feel like you're more about surfacing it and get people getting actions out of it rather than just, like, remembering what they said.

Absolutely, yeah.

Rhys, thank you so much for your time.

Really appreciate it.

We'd love to do a follow-up call in two or three months if it would work for you.

Yeah, I'd love to show you

the prep flow and the tooling.

Once you finish building it, we should be done with that.

Next week or so.

Okay, yeah, yeah, send it to me.

We could do that.

We could do another call once you've got like a demo or something, another like live one, especially a voice, we'd love to see that.

That'd be great, yeah, I'd be very happy to do that.

Yeah, so would you want to let us know when you do that?

And we could also book in one anyway if you want in like two or three months because we're with this technical advisory board, our initial plan was to do like one a month for like six months.

We kind of realized it was like quite intense for founders building stuff and also not necessarily like, you know, a lot.

I think, you know, for like voice AI challenges specifically, it doesn't always change that much every month.

It's like can be like quite big challenges often.

The ones that really matter don't get solved in a month.

So,

typically.

Yeah.

So I'll send a email that with a link, if that works, to have another.

Yeah.

Well, yeah, we'll put that in and then, yeah, I, I'll, I'll get back in touch when we've got something to show it, but love to show it and love to talk a little bit about how it's working and, you know, yeah, experience of building it, you know, and, and what can I have?

We would love to hear that.

Yeah.

That sounds amazing.

Excellent.

All right, well, thanks very much.

Love you to meet you.

Have a great day.

See ya.