Reading the Currents: Themes, Apps, and Services in a Post-LLM World

Liam: Prakhar, Jamil, Brenda, this is our panel for today. We're going to be talking all about AI, how it's been shaping and changing, and allowing partners and developers to evolve. So we'll do a quick intro around the round table in a minute. So everyone can just describe a little bit about what they're doing, what they're working on. Also, do a little check on the mics.

Jamil: Hello. I think we're good.

Liam: Okay. They're working.

Brenda: Mic check.

Liam: Oh, okay. Also, apologies in advance about my voice being a little bit raspy. As you can imagine, it was a little bit of a busy week.

And I don't think I'm the only person who has a little bit of a raspy voice. So let's all be a little forgiving on people's voices today. Actually, just quickly, before we get started, I'm just gonna ask the room, like, everyone who's a member of the ShopDev Alliance, can you raise your hand?

Woo! Woo, that's amazing, that's amazing. And you should all have these beautiful gold badges, which are really, really cool.

But yeah, I just wanna say give a moment to ShopDev Alliance in general, like it's more than a Slack channel, it's more than just a group of people. It's an incredible community. Everyone in there is just so generous with helping each other out and lifting each other up.

So big shout out to ShopDev Alliance. All right, so maybe we'll kick it off like everyone just giving like a quick intro on who they are and maybe if you can like name like a recent AI discovery or experiment that you've done that has you know just had a cool impact on your work. So Prakhar if you want to

Prakhar: start off? Yes can you all hear me at the back? Oh yes very nice. Hey everyone I am I'm Prakhar from FoxSell.

I'm the co-founder of FoxSell. We basically are a bundling app for one of the bundle app and C of infinite bundle apps. We are specializing in enterprise and complex bundling problems.

In terms of AI, we do have a very small team of four people, and we do have a large team of five AI agents that work for us. So that has been pretty interesting to set it up internally in terms of support, engineering, social media, marketing, those kind of stuff. So happy to chat about that.

Jamil: Hey, everyone. My name is Jamil Bhuya. I'm co-founder of Otherhalf Studio.

We're a Shopify Plus agency essentially building beautiful, high-converting storefronts on Shopify from scratch. An interesting thing that we're attempting to do because I think Prakhar and Brenda here on my either side are going to be much better at building these AI workflows. We're trying to create a marketing orchestration layer, taking a lot of our call notes, our internal Slack communications, and just day-to-day discussions we're having to essentially pump out formatted and drafted marketing communications, whether it's for LinkedIn, Instagram, email, and just like trying to automate our agency marketing.

So that's been fun. Cool. Anyone can help, hit me up.

Brenda: Nice. Hi, I'm Brenda Storer. I am a theme developer.

I'm a theme partner. I have a theme in the theme store called WISC and I am a solopreneur, so it's all me. For that, my agency's called Coquelicot and I also do some client work through there.

And then as Trudy mentioned before, we've partnered on a new agency called Studio & The City, where we are looking for like really cool, custom, unique experiences to work on in Shopify, in the Shopify ecosystem. Top of mind, I probably could think of something more relevant, but the badges that some of you got right away and some of you got late, was a new AI experiment in data merging, like getting CSVs and made this an illustrator and then had to import the data from a CSV into this really specialized, formatted way that illustrator could understand. And like, Claude helped me so much in just like, hey, take the info here, put it here, bam, that was really awesome.

Liam: That's really cool, I love that. I think one of the most obvious and talked about use cases for all this AI and agentech technology is around everything with code generation and just being able to speed up those workflows. We'll definitely talk about that a lot later on in this chat.

I wanted to just ask what are some of the non-code use cases? people out things more around managing business. If you can describe some of those non-code, maybe it's around client management, maybe it's around support, just how you've been able to optimize your business.

Prakhar, you can start with you.

Prakhar: Cool. So we have leveraged AI a lot in non-coding things. So we do have an AI support agent now that handles our merchant facing communication.

And we have seen it almost resolve, like 80% of the support tickets without needing a human help. Apart from that, we are trying to close the loop to a complete cycle, which is merchant asks the questions on like support channel, and then the support agent basically has access to all the information. Or sometimes you don't have a help doc internally or externally that would let them answer the right, the question.

So it basically raises the escalation on Slack. It creates a linear ticket for our team to review. So our engineering team does not have to go to the support channel, it can just take a look at Slack and see what all things needs to be done.

It does a sentiment analysis and then creates the priority based on that sentiment. So if it's like very urgent or if it's frustrated merchant, then it becomes like more urgent than, of course, like in other use cases. And one other things that we have struggled a lot in the past has been always because we have been builder at heart.

So we have been building a lot, but not being able to update our help docs, which has been like a pain for us. So now using different MCPs, like Supercut has this MCP where you can just create a video and you can then query the transcript, you can get the screenshots. So our support agent basically, once you record a video for a new feature, our support agent basically get the access to the video.

It creates a help docs for the new feature with images. So that has been a very big challenge, especially when it comes to Shopify apps, because everything is in iFrame. So it's kind of difficult to get the screenshot of the app.

But with video, it's easier to get screenshots. So you can create like help docs very quickly, almost as soon as you launch. And also the change log.

So it automatically sees what is being pushed to GitHub. And if it makes sense for the merchant to know, automatically updates the change log for FoxSell as well. So it has been pretty interesting working with these kind of agents and requirements. So the bane of

Jamil: any agency's existence is writing proposals and scope of work. So that was one of the things that we wanted to automate as much as possible. So essentially it just takes our Granola call notes, looks against all of our our previous history of SOWs and it's fucking great.

One click and it generates, it sends me a Slack message and then from there obviously we'll tweak it but that's probably been one of the most time consuming things in the past that it's helped automate. So thank you. Yeah.

Brenda: (laughing) Analytics like analytics on, so in the theme partner world, We have the saddest little dashboard from Shopify that you've ever seen. It gives us like two points of information and they're both wrong. So I have built a whole like Zapier webhook that when somebody opens the theme in the theme editor for the first time, it gives me some data and it is all on a big Excel or not Excel but Google Sheet. I manually have been managing that and looking at it and adding data. So it's just me, I just have kind of looked at it for a long time, but having that aggregate in a way, like in Looker Studio that is more readable that I can really see trends has been really awesome.

Liam: Cool. Yeah, that's a really good use case. Do you find, I imagine everyone's just AI stack and suite is gonna look a little bit different.

Do you spend a lot of time tweaking that? Are you auditing the different tools that you're using? Like how do you, I guess like, what does it look like?

How do you keep up? Like how do you know when you need to change?

Jamil: often I know I'll never be able to keep up. So it's already one. It's like a classic ADHD brain where like I can get very frustrated when something doesn't go according to plan and then I kind of just go on and move on. So like anything, I always try to learn from others but like getting help implementing it from like someone who's naturally a builder. So consultants are on work, 'cause they actually do the damn work.

They don't just talk about it. But they helped us automate our Notion board. So when I move something through my sales pipeline, it'll automatically create project timelines, send it to clients, it'll generate tasks that are assigned to people.

So it's, at least for me personally, it's one of those things where it's like, I'll never be able to keep up. So it's like, what's an MVP that's gonna help solve what we need versus always like fomoing and like, I don't wanna say wasting time, but sometimes for me not being a natural builder, it does feel like I'm doing it 'cause I'm fomoing over the last Instagram video I saw. So I think like having those mental guardrails for me is personally important 'cause then I can actually do what I'm good at versus wasting time building what I think would be something subpar versus someone was a lot more AI native.

Brenda: Mamika, I'll be honest, I am not super in love with AI in general. And so my stack is probably much smaller than most people, but I will say it's like actually coming here this week and coming to, I'm knowing that like Shopify is so hardcore in AI and feeling maybe I'm left behind. Like I'm actually like really inspired by a lot of the use cases that people have been talking about and have found inspiration in ways that it feels true to me to use AI.

So I'm excited to kind of go back and play around with that. But what I do now is I use Cursor and I have it suggest while I'm coding, but I rarely use it to build new things completely. I'm in a code base that I completely made myself and that I understand very well and it's just me.

So I have found sometimes when I try and get it to do things for me, it goes too far. But I have found that it gives me inspiration to figure out how to just solve it myself. So I'm trying to look at it like that, like use it for inspiration a little more than use it for like completely solving a problem.

I get frustrated when I think it's gonna take more time for me to QA what it does, then for me to just do it myself. So often I just do it myself. What I have found really useful by myself as a solo developer is Cursor Bugbot, which is also not perfect and often recommends things that I don't need to fix, but it does find a lot of like little things that I would have missed.

And it also again, just kind of like, is somebody else looking at it and telling me, Oh, here's a thing maybe you should pay attention to. And also, I also have a process where I use GitHub. I create pull requests for everything, even though it's just me.

I have a very nice version release like GitHub, CoBase. And so I love that Cursor Bugbot creates the PR details for me of what I change. And then I'm just like, boop, change log.

That has been a real big time saver too. I also turn it off for CSS Cursor because I think its suggestions are terrible for CSS.

Liam: Cool, yeah, that's, I love that. Jamil, we were chatting a little earlier and you were talking about how you kind of, maybe we'll have a little bit of a different perspective in some ways, like as someone that's in an agency that there's many clients that actually don't want you using AI in certain scenarios. Can you like dig into that a little bit more?

Jamil: Yeah, so we have a couple interesting policies I'll talk about first. One of them we're actively trying to implement is we're not accepting AI feedback or AI briefs because especially when it comes to something, I'd say more subjective like creative. It's really like beauty's in the eye of the beholder and in the eye of AI, it's fucking everything.

It just pukes out like a brief that's so expensive or we had this one client that wanted to customize the living hell out of their buy box where the buy box on mobile was larger than the screen and it's created so many choices that the customer doesn't know where to start or stop on the subscription where it's like there's a 30-day, 60-day, 90-day module there's all these disc-ground tiers and it creates it's all these subscription stacks that you're like, you took them from one to two decisions to like four to six. And so it's just creating, it's giving people, and I'll say this for myself, it's making people who are maybe not the smartest in an area have the fallacy of being smart and experienced, right, without all the mistakes in between. So it's kind of like, hey, I know how to replace a doorknob on a house.

So therefore I think I'm qualified to build the whole damn house, including the foundation. But that's just not true. So what we're dealing with as we go up more mid-market is a lot of these mid-market brands that are more premium or lifestyle oriented that they don't compete on margins.

They're competing on like, you know, creating a feeling or a community that's commanding more of a premium. They literally tell us they do not want AI creatives. 'cause the moment the customer thinks something's not real, it's kind of like a fake Google or Amazon review, they completely lose trust.

You lose that premium factor in feel that they wanna invest in. And it's been actually very refreshing and interesting 'cause a lot of founder-based brands that are very data-focused, they're like, whatever, like just do whatever we need to launch. But the guys, the companies out there trying to build something that's going to last, they value something that's real and tangible, knowing that trust is everything.

So that's been refreshing and interesting.

Liam: That's interesting. I definitely might have experienced a little bit of that myself of like, yeah, if you see like creative, like advertising editorial from a brand and when my spidey sense is like tingling of like, oh, this is AI, isn't it? I kind of like instantly just like lose a little bit of respect.

So I could definitely see why agents, like those like bigger brands would be giving those directions to agencies. Yeah, but Prakhar, so we talked a little bit about, I guess, like some of the pitfalls around AI. Like, are there things that you don't want to use AI?

and are there things that you don't want to automate away within like the FoxSell?

Prakhar: Yeah, yeah. For me, I think I don't want AI to represent my thoughts like on my behalf. So that is something that I actively try not to do.

So all the thoughts on social media are mine. All the thoughts on LinkedIn, Twitter are mine. Also, I think there are two types of AI that we are using, like me personally are using.

One is like the asynchronous one where you just ask it and it in the background starts working and then there is the iteration one where you tell it to build something but you have to keep iterating on it because it can't do anything in one shot. So the second part is where I am actively giving a lot of inputs, a lot of planning, a lot of direction. That is where I think AI is not there yet in one-shorting things.

But the major thing is like what I think even Jamil said is the trust and authenticity is where I think I am not trusting AI with yet. But I still say yet because anything can change like in the few months.

Liam: Yeah, totally get that. So Brenda, as your background as a theme developer, I remember when I was working on themes as well, and when ChatGPT was a lot newer, and it was working well for a lot of technical things that were more general, but for Shopify specific things, it would often hallucinate. It would tell me to use liquid objects that didn't exist, like generate code where there's this filter that it's like, I mean, I would love if that filter was actually supported.

That sounds amazing. But I guess since then, the newer models have gotten so much better. Like what's the kind of like the state like the theme world with like using all these AI tech now?

Brenda: I'm not sure because I think like the way I look at it is I want AI to do the things that I don't like to do and that I don't know well and that it can do better. But like I said, I really only use it for code completion. I have structured like a new features biggest feature I added to my theme that I did with AI was a countdown timer and it was great.

I just had it like scaffold the whole thing but I had it do it outside of my code base and then I took it in my code base and it was about 80% there and it was kind of wrong and then I fixed it and that was awesome. That saved me like so much time. But I'm trying to think if I can tell if like the code completion has gotten a little more smart.

I honestly, I should go back and take a look because I think I was like, oh my gosh, what is this object? This doesn't work. And then I was like, I'm never using this.

I'm not trusting this again.

Liam: I've heard from like chatting, like I'm sure a lot of people know, um, Jan Frey, uh, you know, I was chatting to him recently about like what he's using because he does like liquid code camps and you know, he's like recommending, um, like, you know, how developers using AI in a theme context. And I was asking, are you using specific MCPs? How are you-- do you need to give the models, whichever the latest ones are, do you need to give them extra backup?

Are there special tips or tricks? And he said, actually, no. He's not noticing a huge difference even using the Dev MCP, the Shopify Dev MCP with liquid. that Claude seems to be pretty good at anything to do at Liquid now, so he's kind of just using a really default model in a lot of cases.

Brenda: I love it, actually. What I really love it for is schema. Like how often I forget a comma or something and the code completion and it also just being like, I think you wanna name it this, and fixing that I think is really awesome, Because schemas are the worst thing to have to write.

Liam: No, I guess use it for the tedious things, but the stuff that you enjoy doing. If you go like going really deep in the code, if you enjoy doing that, why not keep doing that? And then don't feel like you're forcing yourself to use AI and remove the joy from what you like building.

Brenda: Yeah. I've been thinking about something I used to tell clients when I worked in the agency like 10 years ago when they would be like, "We want to go faster, just hire more developers." And I'd be like, "Well, one woman can't make a baby in nine months." And I think there's some of that mentality that we were talking about earlier too of like, "Some things need to, like I need to figure out the problem while I'm working on the problem and I can't figure it all out ahead of time until I'm like in it and to like remember not to just build things because like Sakshi said to build things just because you can but to like also do things thoughtfully and sometimes you need to do take one step before you can take the next step and I think it's easy to get seduced by the speed of pushing out features. I agree with what you said

Jamil: like I'm not the creative on the team but listening to the creatives on our teams when they're going through UX design. It's like they've tried using cloud design. First of all, you have to keep prompting the crap out of it.

You can't just export it to Figma and kind of take it over the line. But we read this joke recently that I was like, this is kind of funny. It's like, how many designers does it take to change a light bulb?

Well the designer's going to think, do we need a light bulb or can we put in a window? So it's kind of like this iterative process. It's kind of like when you go to a Thai restaurant and they ask you, do you want the dish mild, medium, or spicy?

It's so subjective. Like what the hell is medium to you versus me? I got to taste the dish and then I got to continue iterating before I find that right spice level.

Because like, I'm brown. Medium spice to me means something very different to somebody else. So it's iterative, it's experimentation.

And it's like, you're not going to just, to your point, one shot it, right? It's like sometimes the best solutions come out of like frustration and like, you know, looking at it from different perspectives and sometimes even just taking time after you're extremely upset by it. And this is all second hand by the way, once again, right?

This is just me listening to our designers bitching about problems and you know, things like that. But yeah, I think that's where we're seeing like the value of actually trying things multiple times versus just having something generated for you one time.

Liam: Yeah, yeah, yeah, I love that. One thing I actually wanted to ask as well, so, yeah, Prakhar, we did an X space kind of like on a lot of these topics a couple of months ago. I remember Dennis had an amazing question towards the end of that of like, what is your, and like of all the stuff that like we're using that it's improving things. What is the end goal? What does the end stage look like of the biggest benefits that we're able to get?

Prakhar: Man, I just wanna talk to you. I just wanna talk to everyone and not do any work at all. I think that's my end goal.

If I can, then why not? I want AI to do all the work and maybe just talk to people, build more relationship because I would never let my AI talk to you Liam. I always want to talk to you myself. So that is one thing that is the end goal. Just meet more people, build more relationships and I think relationships are more and more important now than ever. And yeah, why do you work? Just let AI do all your work if it can and if it can do better work than you then why do you

Jamil: need to work? I don't know. Yeah I think Brenda said this best it's like you want to use it to do the things you hate so you have more time to do the things you love or actually passionate about and forever for different people that's different things you know depending on how you grew up what you value you know to Prakhar's point, I have a line where I was like, build relationships, not workflows.

Because that's how, at least for me, that's how I've built success in this ecosystem. It's through people and relationships, a bot never brought me a lead, you know? Even though ironically we're gonna try to build an outbound bot, but the point is that, you know, the point of that bot is to find signals and opportunities to kind of start a conversation, not to take over the conversation.

Because especially for what we do, it's more high-ticket sales, it's all about trust. The moment they lose trust, you're toast. So, yeah.

Brenda: I really want to use it to learn things, to help teach me things, knowing that it's probably not gonna give me all the right ways. But especially, I say every year I come here, and I'm like, I'm going to build an app. And I'm a front end developer, and I've tinkered around in back ends and done some stuff, but not really a lot.

But I sure I know enough that I could probably work out an app with Claude by my side. But it always, in the past couple years, it always felt like too big of a hill to climb while I'm trying to put food on the table. but I really think that now is the time that I could use it to help skill up in development.

Liam: It's amazing. I mean, I definitely think it is one of the best times to be able to take on something new that you want to learn that previously felt inaccessible. I didn't have the time.

Now everyone has a pretty great experience expert teacher in almost any topic that you can turn anything that's like, "Wouldn't it be great if I had the time to do X, but learn about X?" Now you can just do that, which I think is really cool. For me, I think it's about time as well.

Getting the gift of time back in a lot of ways, I think echoing what a lot of you were saying that you're able to redistribute your attention, your resources, and your time into the things that you feel are the most important. It does take a little bit of willpower though because, you know, like how do you know, first of all, like, you know, if you're directing your agents in the most valuable direction, and also, you know, just the, you know, AI can be very addictive. I think all of us probably have some experience of just going down a rabbit hole that never ends with AI, and then you look up and it could be like 7 or 8 PM, you're like, what happens over the last few hours?

Jamil: Go touch grass.

Liam: So I think we might have some time for maybe some questions. If we would like to do some questions. So, we can do some questions in the room.

But first off, maybe just a quick round of applause for the panel. Thank you so much. Like so many great insights.

I feel like we could keep talking all day, but yeah, thanks. I'm sure we'll have some great questions from the folks here too.

Audience: I have a question for Liam. Are we going to get Sidekick for partners?

Liam: Sidekick for partners? What would you -- what would you like it to do?

Audience: So, for example, if we get, like, API errors, warnings, and we can get all of the information information in one place we just asked like the AI. For example, what's the server, what's the API, what we need to do, maybe analyze our data because we get more and more data each time we come here.

Liam: And would you like that, would that be like in the DevDash, PartnerDash, or in DevDash? Yeah. - Okay. I'll add it to the list. We got the question over here.

Audience: I will just appreciate Brenda's comment that AI doesn't have to be in everything, especially she works with a lot of teams. So, teams do need personality, do need creativity, so I do appreciate her comment that AI doesn't have to be in everything. So, yeah, that's great. [APPLAUSE]

Audience: Something I've been thinking about for a while now is new developers and new people coming into this space, post-AI. You spoke about teaching. And I feel lucky in that I learned to code before AI, so I did the hard yards.

And I'm kind of terrified about new graduates coming in. And I know I'm the lazy character that would just have the AI do everything. And I wouldn't learn how to debug stuff and fight through things, just if the panel has any thoughts on that.

Brenda: Also terrified for the youth, for many reasons. But yeah, I mean, I think like from my perspective, when I say I wanna learn a new language, I wanna learn more back end, I already have 20 years of just development experience and I've worked in teams and stuff, but I really worry, I think we can't, if you have the opportunity to hire junior developers, hire junior developers, we need the future to be bright.

Jamil: Again, I'll say this as a non-developer, it's like building a business prior to this, I was in a completely different industry, restaurants and the easiest way to learn is by fucking up and when you have to do something hard and it hurts the emotions that you gain from it like you'll never make that mistake again so if you're rolling out code and again I don't know shit about coding but if something's not working and you didn't build it I guess the AI could debug it but you know product so it's not like you're like you don't know anything. Right so it's like that to me is like the thing it's like if you haven't built it and you can't be accountable for the mistakes or the issues in it you won't truly learn and you won't understand the value of what it means for it to work properly versus not work properly so that to me is like scary.

Prakhar: Yeah, and I also just like to add, I think we need to come back to fundamentals now because a lot of things if are done by AI and you don't know what's happening, then that's where you need to learn like how things are working on the backend and how things are happening at a root level because once you build it, and then there's a security issue don't know what to do, then that's not going to work. So reliability, security and a lot of those things comes back to fundamentals. So I think it's good to stress on those back again.

Brenda: I was also thinking like all the junior developers I've ever worked with in like a year, they're not a junior developer anymore, but your AI is never going to graduate, you know?

Liam: Hopefully, no.

Brenda: Not that quickly, right? No. I don't believe it.

Audience: Hey, hi. Quick question, I think, from maybe an aside, what is the cost of AI in terms of real dollars? I know like a lot of indie, you know, solo founders.

There is a budget to it. There's a money to it. There's a token cost.

There's a quality of output because you put a lot of effort into AI, a lot of loops, a lot of failures. Like how do you like manage that budget? Especially as like a recurring budget that really affects like you literally buns your pocket.

Liam: That's a really good question. I guess it's how are you budgeting AI costs? So how are you managing efficiency, things like that?

Prakhar: Oh, yeah, for us it's very important to save money because money is important. Our cost of running AI is pretty low because we have replaced a lot of AI products with our own internal layers. And it's just like we have bought a server on OVH Cloud, which is like huge server, 32 GB RAM, 8 vCPU, and it's like 100 euros per year.

So that's not very much. And then we are running on the Codex subscription, which is $200 per month. And yeah, that's pretty much it, the cost of AI.

If you just remove the development time that I spend on tweaking all of the things when I'm not focusing on the product that I should. So yeah, it's not very expensive. It becomes expensive when you are building this and then you realize how much you were already paying for the AI products that you replaced.

So essentially we replaced Bugbot. We replaced the Cursor agent because now we can just do add. In Slack, we can do add this and just can you open a PR.

And because it has access to all your code repositories, it knows exactly if I'm making an API in this backend repository, then this front end repository should be consuming it. So it has a holistic vision that way. So yeah, I think it's pretty cheap for us as of now, but as the scale would increase, I don't think it's going, the cost is going to be proportional to the scale.

So that is a good thing. And in the future, I'm assuming it's going to get cheaper and cheaper like the internet. So man can hope.

Jamil: Yeah, we don't think about our budget too much given we use it probably for more basic needs than probably Prakhar here. So for us, it's literally just a line item and as an agency, it's always just trying close deals, bring in more revenue than expenses and figuring it out month by month.

Brenda: So I mean mine is directly out of my pocket. So I have a $20 a month of pro subscription to Claude and then I have Cursor on an annual single plan which I think is like a hundred and something dollars, US dollars a year. And then I only use Cursor Bugbot on my theme code base.

I don't use it on client work. So, and I'm only, I'm not in a huge building area, in my theme, I'm updating it once a month. So I have that on like a per use basis.

So it's very minimal. I don't spend a ton on AI right now.

Liam: Good, hopefully we'll be able to keep those expenses down even more. I think we have time for one more question. Alex, did you have a question?

Audience: Sorry, one follow-up. I think just question back to Liam. I know Shopify or Tobi posted a photo with a huge compute rack, which is pretty much self-hosted.

So how would you recommend for especially the community of how much should we run local LLM? How much trust, how much focus, how much energy? Especially going forward because we don't know the scale, we don't know the cost, we don't know what firewalls, what blocks may happen because many countries, many people, different issues.

So that's something that I'd love from you.

Liam: I think it's kind of up to every partner, organization to kind of evaluate that for themselves. And it's definitely something that you need to spend a lot of time auditing, researching. Like there's so many different permutations of how you can create your own AI stack now you can use it efficiently and then that can be different depending on the size of the organisation you're in. Something that you're doing for yourself independently is going to look so different to something where there could be 30 employees that you need to manage for. Yeah, one more question.

Alex, you had a question?

Audience: I just wanted to ask you guys, how do you keep up with advancements in AI and how do you differentiate between the noise and the real game-changing things that are appearing? There's obviously lots happening all the time and a lot of people to follow on social and everywhere else. So, yeah, what are your hacks? How do you keep up with AI advancements?

Prakhar: Don't touch it if it works. So as an engineer, it's very easy to start working with AI and then forget about your product because it's super interesting. But I think I spent a lot of time while we were setting up our agents.

But now there are things coming daily and it's like Jamil said, you can't keep up with it. So we've realized it's better to just do not touch it if it works and if it breaks, take a look but focus on the product that's important. If there's something revolutionary that comes, just take a look and see if it can help you.

But if you just go and spend all of your time into tweaking small little details just to to optimize a little bit of your workflow, it's very easy to get distracted. So that's what we follow now. Adi has, my co-founder has belted me a lot for not building product.

So that's what I'm doing now.

Jamil: I think it's like the previous fallacy of like busy work. You feel busy doing this, so it feels productive, but really it's not moving the needle or it's a distraction. So again, for someone like myself, I recently joined this like sales AI monthly call where it's kind of like when you're building a business, you wanna recruit people smarter than you.

And same thing for me, it's like, I wanna just watch people smarter than me trying to build this thing, rather than me figuring it out myself. 'Cause like, again, I'm just like, even before AI, I wasn't like a super user with like smartphones and things like that. So it's like just absorbing via osmosis and then kind of making it my own if it makes sense.

And again, to Prakhar's point, it's like, if you got an MVP and it's working, like if it ain't broke, don't try to fix it.

Brenda: Totally, I think like, I get it from people. Like when a bunch of people are talking about something, then I'm like, okay, I should check it out. So that's nice from humans.

I learn about AI. But I also think like tech FOMO is, even before AI, everybody's always talking about something and you're always behind. And so I think I've learned how to try and just not get sucked into that as much as I'm tempted to anyway, and AI falls into that category.

I'm totally behind I think compared to most people, but that's okay.

Liam: I think there was a similar question he had on a panel that Brendan was on, shout out Brendan. And his answer to this was that he had like just a bunch of friends that were just kind of on the same wavelength as him, that he trusts that would share what they're working on and just kind of like, essentially like a micro community of folks that like you're passing back and forth and having ideas. And I think even that like personal touch means that like you're not just absorbing what's being posted on X or what's in a blog post you're actually building together, like you're possibly distributed, like these could be people all over the place.

I think that also happens in SDA, like I think the ShopDev Alliance is this crew of people that like again on the same wavelengths are building amazing things, sharing those stories with each other. And I think there's like the AI Stories channel. Is that the name of it?

AI Stories? Yeah, like I love hopping in there and seeing what you're all working on. And like that now will spark like inspiration for me.

And I think that's, again, like we can take these tools and build things together and inspire each other. All right, well. Again, thanks so much for everyone on the panel.

sharing these amazing insights. Thanks for having us. Thanks for your attention.

MC: I know it's been a really long week so far, but yeah, thank you so much.