Episode Transcript
[00:00:00] Speaker A: Foreign.
I sit down with Jesse Hanley, founder of Bento, to talk about testing new ideas.
Our current AI workflow and kind of tool set set up to allow us to be as productive as possible.
Mindset around the transition that we're all kind of going through with it comes to AI, AI tooling, keeping up with technology and the value that we as founders and the technology itself brings to our businesses and how we're thinking about the longevity and the trajectory of our businesses, whether we're growing or we're not growing. And we're finding space to test new things, which to me is giving like a lot of life and reinvigorating my passion in the business as I'm testing and building new products. I hope you enjoy this conversation with Jesse Hanley from Bento.
We did our first like onboarding of a new customer, a potential new customer on Monday, two days ago.
And what I'm finding because I developed the whole thing, right, like start to start to finish. I'm not a developer so I vibe coded the whole thing.
What I'm finding and I'd love to hear how you're dealing with this because you're like kind of mostly solo too, right? Like, how are folks dealing with like testing in qa? Because like it sure as shit is easy to like vibe code stuff up and claude code or cursor or whatever is just doing a whole bunch of output and you can have like unit tests and you can have end to end tests, you can do all this kind of stuff, but like you can't go like send an email or record a piece of content and see like what happens. Like this seems to be the biggest bottleneck in AI software right now is testing.
[00:01:57] Speaker B: Maybe, maybe.
I think for me with Bento, it's like the stuff that has tests is the stuff that is most critical. So sending email, making sure those emails are all correct. When it comes to the features that I'm building at the moment, which arguably is a lot of AI features, not everything there needs to be super rigorously tested.
And I'm trying to build a lot of those features in a more looser, malleable way where I'm likely going to be changing them and I'm not necessarily going to be doing aggressive test driven development on these types of features. I'll do that on the critical components. If it's like I'm tinkering with Event Pipeline or I'm tinkering with the email stuff, or I'm tinkering with anything to do that touches a user like yes. I'm going to be very, you know, I'm going to work on my. Yeah. Deliberate there. But if it's more kind of like front endy stuff or if it's more experiencing stuff, like, I've overhauled, I've overhauled so much stuff since we last chatted and it's like a lot of that work I am not necessarily doing really big TDD stuff with. Okay. I'm refactoring designs, I'm moving things around. I'm creating a lot like front end experiences which I don't want to do like end to end, you know, kind of click tests in some of these UIs, especially in our drag and drop workflow product.
That trying to do tests on that is honestly just fried. Like, it's just so. It's so hard and so I really can't. So I'm not being.
Yeah, I'm not aggressively testing a lot of front end stuff which.
Yeah. I don't know where, where are you getting, where are you getting bottlenecked or, or where. Where do you feel anxiety around it? Like, do you feel anxiety around not having a lot of these tests? Do you feel anxiety around creating the test?
[00:03:56] Speaker A: Yeah, like, so, so you, you know, I was on the call with this guy Monday. He gave me some interesting feedback. I went and I spent about an hour working on stuff and I was like, oh, this is a pretty big change in like how we do stuff.
I was like, cool, I want to push this out. It's like I haven't even like clicked through and tested this thing yet.
And I was just like.
I think, I think how, okay, this is going to like abstract this away to like slightly higher level. I think.
I think the way I feel about this is how like this week the topic of like, should we be using a CMS has come out because we want to just like vibe our website, right? Like the cursor guy's talking about like they ripped out whatever their CMS was and they're just like, they talked cursor and it changed the website.
I feel like we're doing that with everything else, but then we have to go in and manually do stuff to, to kind of test and it just feels like there's this disconnect. And I guess that's kind of like, that's the whole experience that we're all going through is like there's an expectation of like the unit of work I do. And in one way it's really good and in one way it's super old school and so there's this disconnect right now.
I don't know if you agree with that, but I think that's consistent with the website experience and testing and kind of just how we use SaaS in some ways is like how we use SaaS is not AI first like a lot of the rest of our life.
[00:05:26] Speaker B: Sure.
With the tool that you're building now is that like vive coded from the start, basically like Claude code. You're just like ripping into it, running tokens and then pushing it. Yeah. So I think maybe that's like a large difference. So a lot of my developing is like in cursor and my pattern is I use plan mode is default. So plan mode is always default. New prompt, plan mode, hash it out with cursor. I'm using Opus 4.5 predominantly and for me Opus 4.5 feels like the first time where I am out of the drive in seat, but I'm still watching it. And so I basically am running multiple prompts, often at the same time iterating on one feature. So something could be building. Like in plan mode you can actually do all the tasks and you can run multiple agents on all those tasks. So run them kind of like async and stuff.
But because I'm in the ide, I'm looking at all the files, I'm telling it to roll back and refix stuff. And so I'm kind of treating a lot more like a junior dev.
I'm still intimately involved in the building.
Whereas I find open code, chord code, codecs and all that. You're too abstracted away. And I actually felt when I was running background agents with cursor, I was too abstracted.
And that led to anxiety because I wasn't meticulously going over file by file making sure everything was correct. And I wasn't testing stuff, running a prompt, testing stuff, iterating that kind of feedback loop.
And when I'm doing that and I'm intimately involved checking all the files and I'm not, hands off the wheels, eyes closed, I don't really have any anxiety around it. Does that kind of make sense?
[00:07:22] Speaker A: Yeah, it does, it does. And I guess that's like long term kind of consistent with how I imagine this all going is like I get kind of a proof of concept, you know, an MVP going, we get a couple customers and then like the plan is definitely, I'm going to hand this off to our development team and like they're just going to, they're going to own it, they're going to have to Figure out all the shit I did and they're going to have to fix it, but in a lot of ways, like time to market.
The fastest way is for me to just live in Claude code for a couple weeks and get something going.
[00:07:53] Speaker B: Yeah.
[00:07:53] Speaker A: Otherwise it's like I'm gonna fucking spec this thing out and design and all this. Like it's just easier for me to just talk to Claude code and build it.
[00:08:01] Speaker B: Yeah. So I think the like with these tools it's like you build really far. You can use Claude code and stuff to get like up and running or whatever, but your goal should also be to build very.
To build building blocks for the product. Like when I think about Bento, Bento's got like a couple of really core building blocks in it and those things can't be messed around. So the user building block, the events building block and the emails building block. Right. So those are all very firm, tested, hard components that have been very well thought about and arguably they were thought out by Andrew Culver in the beginning of Bento, like when we were doing a lot of the domain modeling.
[00:08:39] Speaker A: Right.
[00:08:40] Speaker B: And ever since then, a lot of those models have never like they haven't had too many migrations run against them. They haven't really changed too much, but how they interact and flow within the app does and has changed quite significantly.
And so that's kind of like how I think about it. I'm just. Yeah. Like when I'm kind of coding a lot of these features out, I try and make things very like lightweight, but they're interacting with these more harder physical objects.
[00:09:10] Speaker A: Yeah.
[00:09:10] Speaker B: That is kind of streamed through the system. So. Yeah, yeah, that's, that's how I kind of think of it.
[00:09:15] Speaker A: No, makes sense. I'll tell you, like, we'll wrap up this part of it. I think one thing I've definitely learned is I want to just like I want to get the thing running and get it working as fast as possible. And that is definitely not the right way. Definitely the right way is like you're saying like, let's get these couple of really important things done. Right. Use plan mode, develop a plan like in Claude Web first and then bring that in and plan it and like be really deliberate.
Anyways, for my non technical founder vibe, Cody buddies, like go slow early to let you go fast later. I think is like what I'm learning.
[00:09:49] Speaker B: Well, yeah. And also try and stay very true to like the problem that you are solving.
[00:09:55] Speaker A: Yeah.
[00:09:55] Speaker B: Because I think like when you know it's kind of similar to you Know, like, photographers, they always talk about, like, they were most creative when they didn't have all the gear, and then they get all the gear and they're suddenly a lot less creative.
I think that is very true here. And this is also true in, like, the beginning of Bento, when I had Andrew Culver, because, like, we. I didn't really know that I was building the email marketing tool. I didn't know that I was building a marketing automation tool. I was just giving. I was paying, contracting hourly rates and stuff. It wasn't me. So we could build everything, right? Like, you know, I had cash flow from the agency. Let's just build everything.
That was dumb. And now with, like, all these tools, it's like, I've got Andrew Culver on tap, and I can just run him against, you know, build whatever we want. And that is not smart, I think. And so you. You still have to stay very true to the product category you're in, the problem you're solving, the actual customer true problems, all that, and try and work out, all right, like, what are the fundamental building blocks of this? And build that really rigorously and then move on to more kind of malleable other stuff. I think, if. If that makes sense. I think. Yeah, yeah, yeah.
[00:11:06] Speaker A: You were saying how, just in general with, like, kind of AI and tooling, like, gosh, like, it. It seems like it was moving fast, and now it's, like, kind of crazy. The. The speed and the options and everything that we have. Like, you know, I was. I was doing my 100 days of AI, so I was covering this stuff, like, and learning it really in depth, which was really good. Now I'm like, not in a GPT5 and Gemini 3 fast came out today, and Google Opal is inside Gemini now, and you're like, what the is going on? Like, how are you?
I don't know. How are you? How are you kind of keeping up to speed? This is a question I get from everyone, and I think, like, someone kind of your caliber is like, if you're not feeling comfortable with the pace of change, like, we're all in a world of hurt.
[00:11:57] Speaker B: Yeah, things are moving really fast. And I think in the beginning, like, in the past couple years, I was bullish on the tools and I was using them and I was directly benefiting and stuff.
But recently, in the last month or so, things have felt pretty different for me.
I even was replying to a tweet with Derek Reimer, savvy cow, and kind of made the comment that it was like, I Kind of feel like we're moved into product manager roles in some regards. And I'm like, this feels really good. I was working on a pretty complex feature yesterday morning whilst making my kid breakfast with Opus 4.5 running on the kitchen table. I'm prompt in, I'm scrambling some eggs and making some pancakes and stuff. I'm watching it spit out insane code and then watching the live reload and just watching the page get built and I'm just like, what is this? I've never worked in a tech company. I'm like, is this what product managers felt like?
[00:13:04] Speaker A: They just dishing out stuff, get done?
[00:13:06] Speaker B: Yeah, basically. And I was like, this is amazing. I was like, you know, you see those things online, those tiktokers of all, you know, a day in the life of a product manager. And I was like, okay, I feel like that now.
And before I didn't really feel like that because it would make mistakes. It wasn't as like confident so it would make mistakes, it wouldn't do the full thing. It didn't have all the tools and stuff. But now, especially for me, it was like, yeah, I think the 5 models GPT 5 models and then 5.1 and then Opus 4.5 and then I've kind of like settled there and it really does feel like I've got a couple of full time engineers on tap whenever I want. And yeah, I don't know, there's so many like hard performance problems. Like Cursor just released the debug mode which I don't know if you've played with and no, it's insane. It's insane. I like, I had a really hard problem in Flows product.
It's a canvas, you can have unlimited nodes on it. That's already a problem. Unlimited.
There was a problem with some of our power users where they'd have potentially 500 nodes in an automation.
When you're panning around, okay, that's fine, but when you zoom in, zoom out, you got react re renders running, you've got drawing of the canvas a mess and so it would lag under some circumstances.
And trying to get that to be perfect was always really hard. So I go to debug mode. I'm like, I'm going to chuck at this. This one. This is hard. I'm not a front end guy. I think I'm okay at it but like I don't know. This is complex. Anyway, put the, put, put my problem in. It's pretty vague. It then adds instrument like finds all the files, adds Instrumentation. So it basically puts kind of like a console logger where it'll do. Something will happen. It'll do like a post request of data into some server, which is just like a debug log. It'll then ask you, hey, can you go just replicate that issue. So you go to workflow, zoom in, zoom out, add a whole bunch of nodes, come back and go, I've done it. And it goes, okay, looks at all the debug logs, goes, oh, I got it, fixes. It goes, try it again. You try it again. It works. And you go, what?
And then another one. I was like, oh, our JavaScript payload is too large on production. And so it goes, all right, give me the HAR file. So I give it the HAR file, has all the network requests and stuff.
It then again does the whole debugging thing on all the JS files, debugs it, figures out all the things that we're not using, then goes, hey, like, can I just remove all that stuff? And because we actually don't need all of these components, yada yada, does it. Payload drops down, site gets a lot faster, everything's better. I'm just. It's just. I don't know, man, it's just insane. And that wasn't necessarily possible before without a lot of prompting and tweaking and stuff, but now it's just like, everything's very seamless. So I have two.
[00:16:21] Speaker A: I have. I have two kind of surreal experiences, actually, both from today, but. But they're representative of kind of like the last couple of.
The first one is I'm still publishing quite a bit of YouTube and I'm using nanobanana for all of my thumbnails.
And it's really good.
One, the biggest. I was going to hire a graphic designer, but the real reason I want to do this is YouTube has built in split testing now for thumbnails and thumbnails is by far the most important part of a video doing well.
Um, because, like, the title, like, it's pretty easy to get the title right. It's just words, the thumbnail. Like, there's so many elements to it. And so now I can go in and like this morning I sit at the coffee shop.
I put into Gemini. I want three. This is the video that I'm doing. Give me three prompts for Nano Banana to create three different thumbnails.
Like, two of them were one shot. One of them was like. It was like Mr. Beast on one side and me on the other side.
Because the video is like how I Reverse engineer viral thumbnails. And like my facial expression was really weird and I was like, ah, that's too, like, that's wrong. And I had to do it a couple times, but it was like 15 minutes to get three pretty good thumbnails.
Like definitely not possible. Like a few weeks ago, about a month ago.
The other one was way more valuable. And come to find out later in the day, I totally fucking did this wrong.
We're running some ads now. And so I was like, hey, it's like our weekly time to Review ads.
Opus 4.5. Just walk me through what you want to see to make sure that we're like as optimized as possible. And it's like, cool. Here's the five screenshots I want. I take the screenshots, I put them into Claude and it tells me, okay, that's good. This is a problem. Do this. Your tracking's fucked up here. We spent about 45 minutes on it and that was like.
I was on Twitter later in the day. I was like, I want to be able to like talk to my Google and my, my Google AdWords. And so my buddy Stuart's like, yeah, there's MCPS. You literally can just connect your fucking AdWords to Claude and talk to it.
Which is how I'll do it next week.
[00:18:31] Speaker B: I should, I should do that. I mean my, my, my largest, my one area that yeah, isn't tapped at the moment is, is marketing and add stuff. So I actually, actually might, might look at the MCPS for that. That's actually, that's, that's a cool idea.
[00:18:47] Speaker A: Yeah, it like stripe chart, mogul, Google Analytics, search console. They all like AdWords, all have MCPS. Yeah.
[00:18:56] Speaker B: Yeah, that's cool. I mean unlike the Nano Banana stuff, that stuff's been very useful actually. Like all the, all the stuff that's coming out from Google has been super useful. So I think like when, when we were chatting a few weeks ago, the initial discussion we were gonna have was around the Gemini releases for front end stuff because after that got released, I've just been on like a marketing rip. So I've kind of been revitalized again to do marketing. And that's funny for me because I've used to be a marketer, I used to be an SEO and I hated that. I just didn't wanna do that stuff anymore. I've avoided SEO for five plus years. So I've like literally done no search stuff and I've just been ripping into it. We've got new landing pages, the designs I think are pretty great.
I migrated our entire docs into a subdirectory on our main marketing site. Again, one prompt with Opus 4.5, I was like, hey, our docs are over there.
Put them over there.
[00:19:59] Speaker A: Do it.
[00:20:00] Speaker B: And it did it all perfectly. That used to be like a consultant, you know, handled all the redirects flawlessly. Actually came up with a really cool idea with some of the redirects too. I used to be like a consultant and multiple weeks of work, Right. And then now it's like it was done in an hour in the morning again. I think I did that when I was, like, looking after my kid. So insane.
And then on the nano banana stuff.
A cool example that I've done recently is there's a few cool examples. One was we're sponsoring the Tropical MBA podcast in January, and I wanted to do a landing page for them. And so I threw up one of our main images and was like, ah, make a Tropical.
And they did it perfectly. And so now we got, like, a tailored landing page just for TMBA listeners and stuff. So, yeah, there's stuff like that that is useful. And even in app, I've started to animate all of our icons.
So when you hover over, like, the AI icon, the character starts pretending to talk. If you go over the search, the guy's got a little microphone and does stuff like that.
It's making things more whimsical and fun. And so that was all contractors I would add to pay to animate, which I guess is bad for that category of work.
But, yeah, front end components. I've got all these examples on our marketing page. I had an active upwork job going that I was happy to pay, like, multiple thousands. And then Gemini just did all the work, created all those components for me. So I. I don't know, man.
I saw a lot of. I don't know.
[00:21:45] Speaker A: Yeah, I saw. I saw a tweet, and it was in 2026. It will be.
It'll be really hard to get a job, and there will be a lot of jobs available.
I was like, yeah, yeah. Like, I think there's just gonna be this huge discrepancy of, like, talent and need out there.
Yeah. For folks who are not, like, super savvy with this stuff.
[00:22:12] Speaker B: How. How's. How's your team? How's your team? I think I asked this on the last time we chatted, but, like, how's your team now adapting to a lot of the tooling? Like, do you.
Yeah. What. What are those? What are those? Discussions at the moment. Are you guys actively moving faster? Are you not moving faster? What. What's.
[00:22:29] Speaker A: In a lot of ways, in. In a lot of ways,
[00:22:34] Speaker B: I would
[00:22:35] Speaker A: say we're 80 or 90%, like, full Aon N. AI. Native AI is the answer, the first answer to everything.
Um, I think that.
I think there's some.
I think there's a healthy amount of, like, not skepticism, but, like, reserve about, like, hey, I'm not just gonna throw Claude at this thing and let it go ham. Uh, so. So I think, like, the development team specifically is a little more reserved with, like, all right, we're gonna use, like, our tools to do this. A lot of them use, oh, my gosh, what's the GitHub copilot?
Because they want to be in, like, in the file, working on it and having it help them there.
So they're a little, like. They're not like, using Claude code and cursor.
They're not all using it. Some of them are, but the rest, like, yeah, marketing. It's like, AI is everything.
Support AI is the first answer. And then, like, people just because, like, we solve some really, probably, like, you. Some really complex stuff in support, like, oh, I migrated my domain and now WordPress is fucked up, and all this stuff is, like, not matching up with the database on Castos, and you're like, oh, my gosh, like, AI is just not going to solve that.
But I think, like, the answer, what you're really asking is like, what's the mindset of the team?
And it's very solidly like, AI for forward.
I don't have hardly any, like, oh, is this going to take my job kind of discussions anymore.
[00:24:16] Speaker B: What about. What about your category? So, like, with podcast hosting, is there any impact on the category as a whole?
Do you see any, like, yeah, is your category going to be impacted at all?
And yeah, open. Open question.
[00:24:39] Speaker A: I don't.
I don't see it, like, our podcasting is not like, a massively growing space right now.
Our business is pretty flat.
So that's, like, frustrating. But that's not like AI. That's just kind of like, it's pretty competitive. It's pretty tough to differentiate.
Probably sounds like email marketing space. Right. It's just kind of tough.
Um, I don't think that's an AI. That's just a business thing.
[00:25:08] Speaker B: Yeah, but what. What about, like, with. You get a lot of customers through WordPress, right? And through, like, the. The. The plugin and that ecosystem. So how do you see that category changing significantly?
Therefore, installs of Castos. Yeah, so, so that, that, that's kind of, I think where I'm getting. I, I understand, like recording audio and all that. Like, sure, maybe there's like, people will start creating AI audio podcasts or whatever. Like. Okay, but like, what about the WordPress installation market and how people are hosting websites and creating websites and how that kind of flows?
[00:25:45] Speaker A: Yeah, that's not a tailwind for sure.
That's like, that's not a tailwind. At the same time, like, you and I and the Twitter sphere are like on the bleeding edge, right? Like this whole concept of like, oh, we don't need a CMS anymore, we can just talk to Cursor and our copy content team can just. No. What? No, like non tech company is going to do that. So like, I think we have years of WordPress being like a stable partner for us. Like, you know, whatever. They're an investor and they're, they're like a significant source of customers for us.
Five years from now, like, yeah, that, that probably will be different.
Unless we're. I mean, WordPress is a.
Even if you don't love them, which, like, I don't always love them, there's a lot of momentum behind WordPress. And so like, you gotta believe that, like, even if it's not automatic figuring it out, somebody's gonna figure out how to basically put cursor into WordPress.
[00:26:50] Speaker B: Yes. Yeah, I, I agree. I've already seen a lot of people do plugin stuff.
Like a lot of, A lot of WP devs starting to just not reach for plugins and they're just vibe coding a plugin, shoving into their directory, seeing if it works, which there's probably some security concerns there, but yeah, yeah, there seems to be a little direction in that regard. You also see that in.
I think we're covering a lot, a lot of different topics here, but it's just busy. So I don't know, things seem to be pretty easy. I saw, I don't know if you saw Shopify did like, they do these releases.
Did you follow this at all?
[00:27:29] Speaker A: No, I don't follow Shopify at all.
[00:27:31] Speaker B: Yeah, okay, so Shopify do these like, releases every quarter or whatever and they did like a Q4 release lately.
The most interesting part about their release and why I bring it up is they announced that they've got an AI in Shopify called Sidekick. And before it was like, hey, you know, you could ask a couple of questions or whatever. They gave it the ability now to make apps.
And so that's really Interesting to me. So now an E commerce owner can go, I need a returns app that takes returns, refunds customers and does xyz and in a couple of minutes there's an app deployed custom to that person that they can then iterate on and they no longer need a third party app. Now I never leaned really heavily into Shopify. It's a very single digit percent of our customer base. But it's something that I've been chatting to a lot of friends about who are in ecosystems just to kind of be a little bit cautious, I think.
[00:28:33] Speaker A: Yeah.
[00:28:34] Speaker B: Because I think a lot of these large players, Shopify, maybe not WordPress at the moment, but may maybe that HubSpot, Atlassian. Yeah, yeah. They're incentivized to get rid of their marketplace and install their own kind of apps because they can now build these mini apps and these mini use cases. So it's.
[00:28:58] Speaker A: Well, it was like a constraint problem before, right? Like Shopify couldn't build, recharge or whatever, all these apps because it's just like a bandwidth problem now. Like, yeah, they have an agent that can just build these that's like you said, custom to what the customer wants. It's basically free.
There's not a bunch of liability. They don't have all the approval process. Yeah, I think, yeah, I mean I think that's a good example of like you like Jordan Gall talks about this with like his, with Rosie is like he like opened a third door. You know, before it was like the phone doesn't get answered or Betty answers it.
Now it's like a different thing.
And I think that like that third door opening is like that paradigm change is happening everywhere, man. Like we've talked about a dozen of them already in just like you know, 30 minutes on this, on this recording. But like you extrapolate that out to like the, the economy.
I, I think, I think what's interesting is I am less fearful and more excited than I've ever been. Even as kind of like weird as stuff is right now, I, I have the, like, I have the belief that like overall it's just going to be like a ton of abundance, like massive abundance because of this. Like, I'm not scared right now.
[00:30:26] Speaker B: Yeah. But okay.
I kind of feel that how I feel now is actually really similar to how I felt when Covid was happening. So I don't know if this is a thing about this recently is when, when Covid was happening. There's a lot of like anxiety and. But there was also a lot of like quite crazy Optimism. And so people were trying to really like seize opportunities and grow through it. And at the time I had the agency and so I was really working really hard to not, not just survive, but I kind of diluted myself into, hey, I'm going to thrive here.
And I think delusion is kind of the key word there. I deluded myself to get through it and I got through it and it was good, but it wasn't necessarily good for everyone during that time. Like a lot of people did lose their businesses and didn't kind of get through that period. And I do feel like a pretty similar shift is happening now. For example, I was talking yesterday to someone and they have like a email marketing countdown SaaS. You know those countdown timers. Yep. And I, I'm like, okay, like that doesn't exist.
Like, like that doesn't. That doesn't exist very soon. Like that, that, that. Those types of businesses where they're little add ons, it's a feature.
It's a feature.
You're in an ecosystem, you're bit of sticky tape. Those were great SaaS businesses from 2010 to 2025.
I just, I just don't think that's the case.
And so I think if you're in those categories, I think you're somewhat anxious and you may have to build something else or you may have to move into a larger, more competitive, scarier category to survive the long term. And that's kind of actually all that really matters. It doesn't really matter about surviving this year. Next year it's like, all right, am I still in the game in five to 10 years?
And that's what I think most people should be really, really thinking about. Because I, I do see a lot of people are like holding, you know, onto their businesses, but they're stagnating. They're not growing. They've hit like some ceiling.
And if, if the maths, which is like, you know, their churn stays consistent but their net new customers starts to decline, they'll just naturally contract. And so you do have to kind of re energize yourself to get back into kind of like a growth building stage. I think you've almost got to delude yourself a little bit to get into that mindset, to push through whatever's kind of going on now.
At least that's how I feel about it at the moment.
Even with Bento, it's like I, I feel really re energized at the moment. I feel really excited.
Like things are going good and we are growing, but Yeah, I do have to push. I. I have to. I have to actually make that happen. Yeah. Because if I. If I pull off the gas a little bit too much, I think. I think that company will go into a decline or stagnate and in a couple of years may not be that. Anyway. Bit ranty, but.
[00:33:43] Speaker A: No, Yeah. I mean, we. I talked about this with Rob on startups. The rest of us, like, two weeks ago, I think it came out, like. Yeah, we've been. We've been like, pretty flat the last year or so.
I feel like we've tried a lot of things to like, get quite, you know, significant growth, and the. The. The place we've landed is like, the easiest way to grow at this point isn't, like, pushing harder, more marketing or optimizing pricing or with the landing page or whatever. It's just, like a whole new thing. And that's why we're, like, experimenting with, like, new products is like, we are where we are. With Castos, I believe it's like a relative, like, Max. Like, we. We. We can probably get a little more, but, like, if we want another, like, 2 million bucks, like, it ain't coming from Castos.
And so. So I think that's, like, It's taken me two years to admit that, which is, like, it's. It's. It's. It's like, it's a lot of pride to swallow that, like, for whatever reason, like, I just can't make it happen.
That's fine.
At this point. Like, that's fine.
Similarly, I am super energized by new products, which, like. And I told my investors this month, I was like, I don't know if it's right or not, but the fact that I'm excited is, like, pretty significant.
Like, even if it's, like, not absolutely the best plan, like, my momentum behind a thing, like, kind of means a lot, I think.
And so, like, for you, that's, like, you're in marketing right now.
I'm into new products and, like, that. Whatever that thing is, like, that's the right answer.
[00:35:28] Speaker B: No, it's a good plan. I think it's a good plan.
I think it's a good plan.
It's like an ATM for you to explore a lot of these ideas, and I think you're totally right. Some categories and some niches and some businesses, they hit a ceiling. And the Jason Cohen Max MRO article, I think is such a phenomenal thing that everyone should read because it really is just a maths thing.
[00:35:55] Speaker A: Okay.
[00:35:55] Speaker B: Like, I've probably done as much as I can on my Churn. And the net new customers is consistently this much. Therefore I'm basically going to hit this ceiling. And so for me to hit a larger ceiling, I need to either like Bento's Churn I think is pretty, pretty good, but growth is mostly capped at new customer acquisition. And so I either need to radically rethink how we market, radically rethink how we do our positioning, or build a significantly better product.
And so I'm trying to step on the gas on all of these to see what works. And I did try a new product which was like tatami and that was.
[00:36:37] Speaker A: And I've heard you talk about that, but I don't know what it is.
[00:36:40] Speaker B: Yeah. Cool. So it's super niche.
It is deliverability monitoring for large senders. So it's basically a tool for myself.
And we've got a couple pretty large other ESPs, competitors to Bento that use us, and they use it to monitor all their email traffic so they could send in sendgrid, what is that postmark? SES everyone.
And then look at all their customers and go, hey, that guy's spamming. That guy's getting a lot of bouncers. Da da, da. And like police it and monitor it.
[00:37:17] Speaker A: Cool.
[00:37:18] Speaker B: And. Yeah. And. And that's got up to about 10k ish a month and seems to have kept. But I'm not, I'm not pushing it so much anymore because I kind of partially realize there's a ceiling to that. That category.
[00:37:34] Speaker A: The. The TAM is not huge and all that.
[00:37:37] Speaker B: Right.
The tam's not huge.
It's not.
Yeah, the tab's not huge. It's really hard to get in front of the, the buyers of that.
And it's not something to.
Yeah, it's just not. It doesn't have as much potential as something like Bento does.
[00:37:56] Speaker A: Yeah.
[00:37:57] Speaker B: And it's more just fun to work with kind of competitors and, and such. And actually working on that re invitalized me to. To bring a lot of those new ideas and stuff back into Bento as well. So, yeah, it was, it was good from that, that exploration perspective.
[00:38:12] Speaker A: Yeah. Hey, I want to, I want to go. I want to zoom back into like what our original topic was going to be, which is like front end stuff. Right. When Gemini 3 came out, we were going to talk about front end stuff.
I did a video on my YouTube channel where I. The Bento website was like one of the, one of the subjects. Like, I was like, hey, that the design on this site kind of sucks. I want to take some inspiration from a couple different sites and Bento is one of them.
And I want to make the design better at using these as examples.
I'm just not that good at this, evidently, because the design still was pretty average in the end. I'd love to hear how you are using AI to make better design. I think that's a lot of people's limitation.
[00:39:02] Speaker B: How are you prompting?
[00:39:05] Speaker A: Oh, poorly probably. I don't know. Like, I. Yeah, just like natural language stuff. Yeah, yeah.
[00:39:10] Speaker B: So the way that I do the front end stuff, I'm a big inspired by guy, so I'm generally pretty inspired by certain designs. And then because I'm pulling from a few different sources, it comes across as like, I think a little bit more unique or whatever.
And so I'll generally find I like Vercel designs. I like kind of like that black on black stuff. So take a couple of screenshots and then I'll usually get it to write like a cursor rule.
So like, when I started working on these components, these animated components to demonstrate features in the product, what I did was I created a cursor rule first. And the rule would be like components md.
And so it's going to refer to it anytime it builds that component.
And then I would usually go back and forward with a Gemini to write that rule. And then it's more or less always going to be using that components rule whenever it's building those components. So that's typically how I did it. And then I am usually using screenshots, which Gemini is very good for because it's multimodal. So it'll take those screenshots and adhere to those design guidelines and stuff.
And then I just kind of keep iterating on those rule files until I'm pretty happy with the output, even with Bento. So at the moment, I'm on a mission with Bento's main app to clean it up because I get feedback from people. It's just nice that it's like a little bit too chaotic or, I don't know, kooky. And so I'm really trying to tighten that up. And I'm doing the same process. I'm like designing stuff that I really like, then asking the model, hey, I really like how we design this component. Can we generalize some of those design rules into a cursor rule file?
[00:41:06] Speaker A: Yeah.
[00:41:07] Speaker B: And then let's go refactor this, refactor that, refactor that. And then, yeah, that's how I'm doing it.
[00:41:12] Speaker A: Okay. So Just to rephrase that or restate that. So taking inspiration from other places.
Screenshots.
[00:41:21] Speaker B: Screenshots. And.
[00:41:22] Speaker A: And then like asking it to create a style guide or whatever cursor rule to say like, this. This is how you do this thing. And then it just knows that's how it does this thing all the time.
Yeah.
[00:41:33] Speaker B: Yep. Yeah, yeah, yeah, yeah. And then I generally have a few of those. You can also do cursor commands.
So you can basically create like an MD file for.
Yeah. As a command. So I can do like/ Component Build 1 on the Build 3 on this app page.
[00:41:52] Speaker A: Oh.
[00:41:53] Speaker B: And then reference the page and then it will just always rip into that as well. So that's also kind of nice.
[00:42:00] Speaker A: Yeah.
[00:42:02] Speaker B: But, yeah, that's how I'm doing it. I'm trying to think what else.
Yeah. Generally Gemini for the screenshots plus and my prompts are very vague as well.
Like, you know, I talk to it like a friend. I'll be like, me like you. These.
[00:42:22] Speaker A: I just don't. Yeah, I don't think you. I mean, you know, whatever. Two years ago, we were all into prompt engineering, and I just don't think it's important, like, at all. Like, the Nano Banana stuff, I think you need to be, like, pretty specific there. But I mean, you're asking it to do so much lighting and text and background and like.
[00:42:43] Speaker B: Yeah, I. I think if you want to get very specific. If you want to get very specific, then yes. But yeah, you can. You can still keep them. Like, with the Nano Banana stuff, I'm still pretty broad. I'd be like, make this tropical.
[00:42:58] Speaker A: Okay.
[00:42:59] Speaker B: Then it one shots it. I'm like, whoa. Okay.
[00:43:07] Speaker A: Yeah. So what do you think about. I mean, the kind of. The kind of. The big thing that I see this week was like the cursor guy saying, like, we're not using a CMS anymore. We're just like, the whole thing is marked down and like HTML and CSS and everybody can manage it. We don't have a CMS at all, I guess. Like, what is the current stack of Bento like? I know you're like, next js, right? Like, the marketing site dashboard have like, separate repos. There's everything in one place and you just push live stuff all the time or what? What's the current state of affairs?
[00:43:42] Speaker B: Marketing site is Next JS main website is a Rails app.
Chat app is Rails. So anything application is Rails. So I think we've got a total of four Rails. Separate apps, separate db, separate everything. So that's that. And then the main app is Next js. We had two Next JS sites, one for the docs and then the marketing. I've Sunset docs moved that into marketing for the blog. I have Contentful as the CMS because I have external contractors that write articles and so they're doing that though docs are entirely just markdown files now. And that works really, really well because yeah, often a doc will come up and then I'll just cursor background agent, go write a doc on XYZ and then it will do it. And especially when you're making like we shipped, let's say we ship like a new library. So a new Python library.
I can then go at cursor. Here's the GitHub, go read all the functions and then add them into our docs and it will do it perfectly. So there's stuff like that for like a technical blog. Like, sorry, technical docs.
It's just so good. But we're missing this. We're missing this human editing layer on top of it still.
So I actually explored moving to sanity at one point and it was just like, not it. It's not a great platform. It's like very messy, it's clunky, it's hard to get up and running. It's just not, not good. And everything's in the cloud and so you can't actually like iterate on it. You can just iterate on the layouts.
But when everything is in the repo. Yeah, you can make large changes, big changes.
You can just do so much with models and we just need a nice layer where someone can connect to my GitHub, it be constrained to a folder and then people can just edit those markdown. And if we've got a tool like that, then yeah, we're basically at CMS levels now, so. Right, that's what we're missing.
[00:45:53] Speaker A: Yeah, we.
A lot of our content is done through a cloud code project.
Like, I'm really proud of it. Like, it is really good. And the thing I added this week is it creates landing pages now because we're running some ads and I want like specific ads for like podcast hosting for WordPress because, like our page that's in WordPress now, it's like a little old and it needs to get refreshed. But I'm like, fuck. Like the tool can build an amazing landing page, but then I got to figure out how to get that landing page into WordPress.
And so I think this is kind of like what, like, you know, the Cursor guys are seeing is like the speed through which we can iterate and build outside of the context of a traditional kind of bucket, which is a cms, becomes a constraint. And like we're feeling the same thing. Like, we're not going to migrate the Castus website away from WordPress.
We might.
Because, I mean, the other, the other thing is like, we pay, we're WP engine, 300 bucks a month and we have like a WordPress, like maintenance and security guy that's 100 bucks a month. So you're like, okay, like, you know, that's three grand a year or whatever. 3,600 bucks a year. No, five grand a year just on that.
Is that worth two weeks of my time?
I don't know.
I don't know. The math starts looking kind of interesting.
[00:47:19] Speaker B: Yeah. I mean, could you also do a hybrid approach? So with the WordPress site, could you vibe code a plugin that effectively mapped a folder, like a hard coded folder in your WP directory and basically goes, all right, cool, I am going to scan all these markdown files and then basically map those slugs and if someone hits, you know, post, slash, whatever, if it doesn't exist in the db, let's check the markdown file. And I think those hybrids, that hybrid setup's going to work because even with our blog, it's like, I got that in Contentful, but I wrote just like a little script that does that. So it'll check Contentful first, but then it will also check the markdown files as the fallback. And it means I can also write, I can vibe code, an article or. Yeah, so I get both those powers and so maybe do something like that where you can get the best. Yeah, I mean, I think with landing pages, like, get a hard coded page, it'll look up, you know, slash vibe pages and then render that, if it doesn't already exist for that.
[00:48:25] Speaker A: No, I mean, I think like the, the easiest way to do it in WordPress is like a custom page template which is just blank, basically. And then it loads this whole like HTML and CSS thing that we give it. So. Yeah, but I mean, it's just like, it is that kind of friction that's like, come on. Like, I've made this thing, I just need to figure out how to get this thing onto this other kind of more structured box that is my website, which is like, I think that's.
That theme will continue over and over.
[00:48:56] Speaker B: Yeah, yeah, it will continue. And I'm trying to think, well, like, what is the Bottleneck there. The, the bottleneck is in the WordPress case. It's the database effectively because that's where all the content is effectively posted. Right? Yeah, so, so yeah, and same with.
[00:49:12] Speaker A: But no, there are like the landing page providers have plugins where like you can just what you're saying, like if it's in this kind of subfolder, they reference their external stuff and load their external stuff on your, on your root domain. So yeah, it's totally possible.
[00:49:31] Speaker B: Yeah, yeah. I'm trying to think, I'm trying to think where at the moment I, yeah, like I'm probably, I'm going to keep contentful, but only, only really for the blog landing pages. You're right. Like all the landing pages, I built a lot of them either dynamics or Programmatic pages.
So like for all of our integrations, that's all has been coded with cursor.
Yeah, I did like a terms directory. So you know like slash terms slash, like a dictionary. Email marketing. Yeah, yeah. And it's, it's, it's, it's got a decent amount of traffic and has also nabbed us like a couple of leads as well, which was very surprising for me and also helped ranking of a lot of other pages because I built something that does like internal linking.
So in a blog post it'll look up all the words and if a term exists it'll auto link for it and stuff like that.
I've improved internal linking throughout the whole site and then that brought up all the blog posts. So that's, that's, that's been, that's been pretty, pretty helpful as well. But that, that trying to do like that in WordPress was very cumbersome.
It was very cumbersome. But now it's just like you run the command on my next JS thing in cursor, I go have a shower and stuff. I come back and be deployed.
[00:50:57] Speaker A: So I wanted to ask your opinion on our new products because I think we'll have several new products. I don't think that the thing I'm working on right now is the ultimate answer. So the thing I'm working on right now is called Linkberry. It's a AI writing tool for LinkedIn. Because LinkedIn needs more, you know, AI spammy content.
I think if I'm honest that that right there is part of the reason I'm not sure that this is the thing because I just don't love the problem that we're solving there.
[00:51:28] Speaker B: Yeah.
[00:51:28] Speaker A: But I guess like from a high level perspective, if you were like with Your tatami thing maybe like, okay, we, we're going to go solve another problem. How would you think about kind of being systematic about that?
[00:51:44] Speaker B: Yeah, so tatami had a couple of benefits. Tatami was successful to me, even if it failed, that was really important.
So tatami immediately replaced a tool that I was paying for month to month. And so, yeah, everything out, I ended up actually saving money by building tatami and then I made money by selling tatami to other people that I also knew had the same problem as me. Right. So that, that, that, that's how I kind of think about it. So if I'm going to build something on the side, it needs to either save me money immediately or it needs to earn more money than Bento can possibly earn. And at the moment that I don't really have too many ideas that are like that.
So that tends to be the frame that I think of it. And, and I really love working on Bento. So for me to work on something else. Yeah.
Needs to directly benefit the main business or be in a large enough category that.
Yeah. Needs to be larger than email marketing and marketing automation.
That's pretty tough.
Which is pretty tough. Right.
And I don't have. There's not too many ideas that I have the guts to do that.
[00:52:55] Speaker A: Yeah.
[00:52:55] Speaker B: I think, you know, I think pure Play transactional email is, is a much larger category which I think, you know, becoming send grid, becoming postmark is, is interesting to me. And so if I was to actually build a pure play thing, I actually probably would go after category like that.
So, yeah, that's kind of like how, how I think of side projects. I have a lot of like, mini, mini ideas that are nice to have. But I mean, kind of going back to your point with, you know, the upper limit of some of these businesses, like if you personally, if your holding company wants to be kicking off many millions, the category needs to support it. And so it's like, well, does link in LinkedIn content creation have the possibility as a category to bring you many millions of revenue?
[00:53:54] Speaker A: No.
[00:53:54] Speaker B: Maybe not. Yeah, no, no.
[00:53:56] Speaker A: Yeah.
[00:53:57] Speaker B: So, yeah.
[00:53:58] Speaker A: And like, got good advice from our mutual friend, probably Laura Roeder. She was like, any business is hard. You might as well like build the one that you want to build because, like, there's no sense in building like a stepping stone, I think is what she was getting at.
[00:54:13] Speaker B: Yeah, yeah, no, Laura's. Laura's, Yeah, I think she's right there. And she switched on. Like, I think if I was, if I was thinking about like tough, if I was giving you recommendations, because they're free. I can just give you recommendations.
[00:54:28] Speaker A: The thing I need is more ideas.
[00:54:32] Speaker B: Let's do it.
The way that I would think about it, though, is like, okay, what is a larger market than podcast hosts in and is it a Descript competitor?
Is that category actually the largest thing, actually creating the content and record, like, being, you know, being like a Riverside? Like, could you be Riverside? Could you build something like Riverside? I think you could possibly build something like Riverside. Could Custos be like Riverside maybe?
And is Riverside a bigger. What is the recording business and the editing business? That kind of broad, general category? Is that much larger than podcast hosting? And I think the answer is yeah. So if I was thinking about. If I was in your shoes and I was picking a category that was hard, it'd probably be something like that. It'd be like a Riverside or it would be a Descript competitor. And then having, like, cast us the editor and then coming to market with that.
And I think with a lot of the tools you have, it's way more possible now to build something as complex and as competitive as that.
[00:55:40] Speaker A: Yeah, yeah. I'll give it a slightly different slant just because I think, like, again, like, I like to abstract away, like, the topic to something that other folks can relate to.
We're not doing that. We're very much not doing that. Not because I don't think we can build Riverside or Descript or something like, we can totally build both of those.
I believe the podcasting market is just not that big, even if you're Riverside or even if you're Descript. I think if you look at. So I've had Jason Cohen on the podcast a couple times. Incredibly smart guy, just painfully smart. He's just.
Yeah, the way he said it is like, you can.
You can solve, like, vertically or horizontally. So vertically would be like podcast recording, podcast editing. We solve podcasts like distribution, analytics. And then you could do, like, monetization, which we kind of solved. But, like, we could be the. The service layer there that we, like, integrate with currently.
Or you go horizontally. Like, what other things do podcasters do? So they do email and they do social media and they do YouTube and they do all this kind of stuff. And so I think that, like, the ANSW us is probably, like an adjacent market that, like, you're getting at is bigger, has a willingness to pay more, which, like, if I have one gripe with podcasters, it's like, they're often cheap because they don't make any money. Like, it's a hobby. For most people, including me, like, this podcast is as direct avenue to revenue for me as, like, almost anything. And. And like, it's. It's not really even.
And so, like, I think one of the really important things for us is, like, you probably have a direct relationship to revenue with your customers. Like, you send more emails and you make more money.
If I make more podcasts, I don't necessarily make more money or a lot of our customers don't. So, like, looking at something in the creator world, because I think that's important, is, like, we want to be able to sell to our existing customers just in a better part of the market is. Is, like, probably the most important thing for me.
[00:57:47] Speaker B: So. And do you see that kind of, like, sidestepping into, like. Like, you, like, video creation types? Like, yeah. How. Yeah, when you go to category selection and product selection, like, yeah. How do you take that?
[00:58:03] Speaker A: And then, yeah, I think it's.
I think if you had to say, like, category, like, the thing I'm most excited about is 100% is YouTube. Like, it's the best. It's the best category. It's the best channel out there right now, and it probably will be for a long time.
The second one actually is, like, content, and content takes a lot of different formats, but, like, I would just leave it at the, like, at the content level because I think, like, if you're solving content problems, you might as well solve, like, social and website and all this kind of stuff kind of problems. Those are probably the two that I think, like, the TAM is big enough people are willing to spend for them.
You can build a product, you can solve a problem, and we just need to, like, slice our piece of the pie.
[00:58:51] Speaker B: Who is, like, the aspirational companies in the markets that you're looking at?
[00:58:58] Speaker A: Yeah, in the YouTube market, the big one is, like, Vidiq and Tube Buddy. They're both like YouTube analytics helper kind of tools. And there's a newer one called one of ten.
Yeah. And they're like, they basically, their thing is like, hey, we're gonna help you make better videos. Like, better ideas, better thumbnails, better titles, better tracking, better inspiration.
Yeah, we're working on something there. It's not ready for. It's not ready for the world yet. But we're working on something there because I think we can solve that problem really well.
The other part of, like, the YouTube world is, like, editing and recording and all this kind of stuff, which, like, yeah, I don't know that anybody has solved, like, AI video editing, descript will be the one to do it. But, but like, even them, like it's just not there yet.
[00:59:56] Speaker B: Yeah, a lot of YouTube stuff and I was, I was watching her actually. What was it? Editing in Capcut.
[01:00:05] Speaker A: Yeah.
[01:00:05] Speaker B: And yeah, like, I think there's a lot of personality in the edit for a lot of creators and so it's, it's pretty hard to abstract that away even into like giving it to a contractor to do. So. It's, it's, it's a hard category, I think, to, to break into.
[01:00:20] Speaker A: Yeah, yeah.
[01:00:21] Speaker B: The editing space.
Yeah.
[01:00:24] Speaker A: So I, I don't know. I think we were, we're on a walkabout if that's not, if that's not like offensive to say to an Australian. But like, yeah, I mean, you know, we, you know, Castos is very stable. We're quite profitable now. I do have the grace from like our team and investors to go explore and that's like, that's my main job right now. My main job isn't like running day to day Castos. It's like figuring out where the next 2 million is going to come from. Which is pretty cool, but like kind of daunting too.
[01:00:50] Speaker B: Yeah, it's daunting, but like, I guess like the theme of this is we can move very fast. Like if we're optimistic and we utilize all of these tools, like you can totally move fast launch stuff, see if it sticks. If it doesn't stick, move on to the next thing and a hundred percent keep, keep iterating. Whereas before it took a really long time to validate.
[01:01:10] Speaker A: Yeah. Yeah.
[01:01:11] Speaker B: Whereas now you can validate a lot faster.
[01:01:13] Speaker A: Yep. Yep.
Cool, buddy. Well, we're, we're over an hour, which is like a super long time. Thank you for, for coming on. We'll, we'll have, have to have you on in like in a couple months and, and like reconnect because this is super fun for me.
[01:01:26] Speaker B: I don't know, I don't even know it be like still working.
I, I, I, I even, I even feel with this recording, like it was, it was a little bit all it was. It was kind of, we jumped into so many different places, but I think there's just, it's just so, just so much going on.
[01:01:42] Speaker A: Yeah.
[01:01:42] Speaker B: And, and it is somewhat overwhelming and a little bit unnerving. And so I think.
[01:01:48] Speaker A: Yeah.
[01:01:49] Speaker B: As we all try and figure out what's actually going to happen.
[01:01:53] Speaker A: Yeah.
Cool, buddy. Well, I appreciate it. Thank you.
