Ep. #3, Building Platforms in the AI Era with Ben Rubin of Verify

Christine Spang: Welcome to the show, everybody.

We are really excited to have our special guest today to be Ben Rubin. Ben Rubin has had a multi-decade year in the product trenches.

Currently he's the Chief Product Officer at Verify and before that was chief product officer for a number of years at a company called Funnel Leasing, and we're super excited to learn from his experience and to share some of his insights today.

Welcome to the show, Ben.

Ben Rubin: Thanks for having me, really excited.

Christine: So I always like to kick things off with a little bit of like an origin story.

You know, I think that everybody that like spends this much time building software had some moment where it really just sparked to them that software is magic.

So Ben, was there a moment that you discovered the magical power of software?

Ben: For me, it goes a little bit back. I am old now, and when I was 13 years old, my mom worked in IT like, you know, this is way back in the late nineties.

My mom worked in IT, and so we got an internet connection before a lot of other folks did, and it blew me away.

We had a faster interconnection than anybody else. And as soon as I got online, I knew I wanted to do something with that.

And so me and my best friend at the time, we started a small website mostly around kind of old computer games.

We'll talk about the legality of that another time, and we became successful. We generated a following.

For anybody who's interested, the name of the website was gangsters.org. Really, really good URL.

And with the success from building this kind of website that was about a passion of ours, you know, playing computer games, we realized that we had learned things that a lot of other people our age weren't aware of. And so we started trying to help others.

We started a small web-hosting business and we would assist other folks by hosting their websites, buy them a URL, which at the time being able to get a credit card and spend $70 with network solutions was a differentiator.

And we'd set up other people our age, maybe a little bit older with their own kind of hosted website. We'd moved them off of GeoCities or Tripod or whatever they were on to that.

And we were successful. Like we started, you know, right before the explosion in '98, '99, we were successful with a little small business.

So that's what, ever since that I knew, so at a very young age, 13, 14, 15, I knew I'd be doing this for the rest of my life.

Christine: Wow, that is super cool.

Did you ever get in trouble with your parents for like hogging the dial up internet all the time? They couldn't make phone calls?

Ben: So I was really lucky my mom had better internet installed our house than most people had.

So we had like an ISDN connection, I don't even know what that is anymore. And then we had cable before anybody, I think her company paid for it.

And so I didn't get in trouble with that. I did get in trouble based on the amount of, we got a number of, we were young, we would copy logos from people, we hosted some games we probably shouldn't have.

So we got a number of cease and desist orders, and only, we were only involved in one major lawsuit, but we'll talk about that another time, and that's what my parents got frustrated by.

Christine: Yeah, that's super funny. We definitely got in different kinds of trouble.

You know, my mom would pick up the phone, and when hear that, you know, (mimics beeping), and she'd be like, "Christine, get off the computer!"

It sounds like you have like a bit of a just entrepreneurial roots as well with, you know, not just like being excited about the software but also like, you know, wanting to turn that into something that you could make money with, help people.

I don't know what the strict motivation was, but clearly led you kind of straight into the business world as well.

Ben: Yeah, I mean if you look behind me, there's a ton of Legos. I've always been one of those people who, when I was young I built fantasy worlds.

When I got older, I tried to build businesses. I always was hustling something or another, mostly very legal.

And then now my, what I do for fun is assemble Lego sets because it, you know, does something that calms me down.

So it's been, you know, since a young age, I was super entrepreneurial, always trying something.

I feel like I might've started a little bit too early in some ways because, you know, when I was young actually getting a business bootstrapped was pretty hard. I think there's a lot more resources now than there were then, but we had a hell of a time doing fun stuff when I was younger.

Christine: So one of my motivations for starting this show is that, you know, we've been running Nylas for like 11 plus years at this point, and I've just seen like a lot of history of kind of the evolution of business apps.

Like a lot of people build stuff with us that, you know, is used for running various different types of businesses.

Some of 'em call 'em CRM, some of 'em call 'em platforms, whatever you want to call it.

I'm really keen and interested to just like hear from your point of view, like, you know, you've been through a couple companies and I know that like Funnel is definitely in this space, but like how has like just the experience of like building these types of apps changed over time?

Like when you got started, what was it like to just like go and build something that helped other people get stuff done?

Ben: Oh man. So if you go back to, I'll talk about my professional career rather than my hijinks as a young person.

But I started working professionally in 2005, and the first company I worked for was a enterprise company, was in the news business tremendously dealing with changing technology, and I remember how difficult it was to get anything done.

We were a super waterfall shop, you know, at that time, business stakeholders, which I was on the business side, didn't talk directly to engineers.

You would talk to somebody, project manager in the middle, and then they would speak to the engineers, and it was just a very antiquated culture.

And over the years, it just, things have changed so much, moving from a world where there was a big separation between kind of the business and engineering at a time when really product was just, I would say getting its start from a technology perspective. You can't talk about how different and how much more efficient things were.

We would go, I remember back at that time at that job and others, we would do a quarterly or a half year release, and if you didn't get things out the door at that time, you'd often wait another couple months to actually deliver your software.

So the world has changed, you know, now working in continuous integration environments, we're deploying new code to features on a daily, you know, every other day basis.

So things have really, really changed.

Christine: So you're telling me that the Office Space world was real?

Ben: Yes, I do enjoy. In addition to that, I do enjoy working remotely. My first job I had to wear a shirt and tie every day to the office, and I haven't worn shoes in probably a couple years.

I'm now a shorts and sandals guy. It's my deterrent from going back to the office.

Christine: Did you take the customer requirements down the hall and tell them to the engineers so that they could build what the customer needed without the engineers having to talk to the customer?

Ben: The TPS reports are a thing of the past, but besides that, it's very much, it was very much a world that.

The one thing that out of that great reference is, I do feel more a need to keep closer to the customer than I've ever, you know, felt before, with things changing as quickly as they are, you know, listening to the customer and being a translation point for the engineers is super important.

Isaac Nassimi: So you've basically talked about how building software, shipping software has changed over the years, and what you seem to be saying here is that we're also in kind of another, another variant of that change.

Can you talk a little bit more about that?

Ben: Yeah, I mean, I'll take a minute back just to, you know, I'll get to the middle, and then I'll get to the end.

I think the middle of my career, you know, things started getting much faster.

So, you know, every kind of old company, whether they did it well or didn't do it well, moved to kind of an agile development environment.

And then, you know, the other big change, the other big shift was, all of a sudden with, you know, the era of restful APIs and everybody kind of opening up their ecosystems or some people opening up their ecosystems, it became a lot easier to build stuff.

So, you know, one of the reasons I'm talking to you folks right now is I worked at companies that wanted to add communication capabilities, and the idea of trying to, you know, build a system that would interact with Microsoft or Gmail's email systems was insane, you know, in the past.

And then all of a sudden with the rise of folks like Nylas, with the rise of Twilio, with the rise of, you know, even going further back to Google Maps, there's a lot more possible to do.

You didn't have to be a giant enterprise in order to deliver really exciting software.

And so that was kind of like the first major milestone that I think, you know, really shifted the world and made product a lot more engaging because you could do things much quicker.

I'd say now with the kind of rise of artificial intelligence, things are moving even faster.

Getting to the point where, you know, in the past, I always felt able to have an advantage when it comes to my customers in a build versus buy situation.

You know, we had dedicated engineers working on trying to solve problems across multiple customers.

I think we're now entering a world that may shake up a little bit B2B SaaS where customers have some of those same kind of tooling than than we did. And, you know, performing, you know, some things that people build entire businesses on you can now do in a scalable way.

And I'm not talking about AI coding revolution, I'm just talking about like, hey, there's companies that all they did was extract information from documents or extracted information from spreadsheets.

And you can do that now in a pretty automated fashion at a pretty high level of accuracy.

So the world is changing quickly, and I feel like as a product person you really have to be plugged in and trying to get a little bit of ahead of, you know, the adoption of this technology.

Isaac: How do you feel that you can like maintain an edge or even dig a moat that you feel comfortable with that will still exist in a year?

Ben: Yeah, that's a really, really good question.

I feel like a lot of what I did in my last position and I do now is trying to create the rails that allow you to leverage software, and, you know, Funnel, which is a company that I was most recently with is a good example of that.

You know, we built CRM, we built communication capabilities all around the apartment rental space.

So basically we were helping folks who owned, you know, 50,000, 100,000 apartment units, lease their apartments to consumers.

And for any specific piece of what we did, you know, helping somebody with a phone system, helping digitize their phone system, helping them get access to communications across a large portfolio, helping leasing agents know what they needed to do, you could build a solution for that pretty quickly for any of those specific points.

What we were trying to do was be the rails for the entire system. So being able to solve leasing agent problems, being able to solve consumer problems, being able to solve the business problems.

And I think by having more of the entire process, not only could we provide better data insights, we were also a really good well positioned to provide automation.

So like, you know, Nylas is a great example. We were building the ability to sync with a leasing agent's email.

Once we were able to do that, we could then leverage AI to be able to automate some of those emails going back and forth.

And we were doing that a while ago, you know, even before the LLM revolution.

And so I think that kind of having a platform is super, super important now as compared to maybe before, because building a point solution, the moat as you mentioned, is super, super light, and who knows when another capability from one of these companies will come out that will erase that solution. So I think you really have to be of high value, especially in a B2B SaaS environment.

Isaac: Absolutely. And I think we've seen a lot of that play out very quickly in the the ChatGPT wrapper businesses, right?

Where there are all this slew of businesses that essentially they're just a pre-made prompt with the UI on top of ChatGPT's API, and what happened either, you know, OpenAI started shipping this stuff directly inside of, you know, their end product, right?

The actual ChatGPT product, or you saw some of the larger players that you would normally have to use these ChatGPT wrappers in the tech stack with the larger players, just build those things in and kind of just nullify those startups overnight.

So if I take that to a termination point, what you're saying, it seems like a lot of software, especially the big players, are just going to have a ton of features forming a really large platform.

How do you see that playing out?

Ben: Yeah, I mean it's a really interesting question, and you know, there is a lot of chatter online about is this the end of software development and will your 8-year-old be able to build a major enterprise or build the capabilities of a major enterprise after dinner?

I'm incredibly bullish on AI. I was, I was a big skeptic of the Web3 and the Metaverse, and I hope nobody hates me for that, but I am a really, really bullish on AI and its capabilities.

That being said, it's going to take a while for this all to figure itself out.

And so, you know, my belief is that I do think especially, you know, my area of expertise is B2B SaaS, and I think at a B2B SaaS environment, it's very hard for the companies that I serve, and you know, I've been in the real estate business now for eight years.

It's very hard for them to create cohesive experiences either for their internal stakeholders or their external ones.

And so to your point, you can, you know, this is kind of same thing I was saying before, you can put together a lot of small pieces, but to make that into something that's meaningful, it is helpful to have an organization that is stitched that together.

And so it is, it's a little weird, you know, if I talked about that API revolution, you know, we were stitching together all different types of software and providing value out of it, and now you're able to do that but kind of superpowered.

And so I think what you have to keep up with is, there's no ability to lean back, you know?

Anybody in this environment, I think, you know, there's been a lot of chatter about like PE firms and the offer, the what, the value they offer.

And I'll tell you, if you're trying to milk a company and you're not focusing on R&D or incremental development, it's the worst thing you could possibly be doing right now.

And, you know, maybe you need to use your resources a little differently. Like, I don't know if a bunch of front end developers are going to do as much for you as they might've done in the past, although user experience is still important.

But getting into this new stuff and seeing where you can find value, you know, I don't think there's a product leader out there that's not spending a lot of time on that.

Isaac: I've actually got a question on that front, which is, you talk about basically how you can never kind of sit back and, you know, quote-unquote milk the business, right? You constantly have to lean forward and be pushing R&D.

I remember, maybe this was like a year ago, someone was talking about like Uber versus Lyft and how they have to sink so much of their cash flow into R&D just to keep them ahead. Or even in that two player race that is Uber and Lyft, and, you know, that can obviously completely affect, you know, economics of a business if you're never able to go into that profitability mode that a lot of businesses say that they're eventually going to be able to go into.

Do you think that changes the way even startups are looked at long term?

Ben: Absolutely. I mean, as a senior product leader, managing that expense, managing that competitive expense, I would say almost, is incredibly challenging.

I think you're going to see startups and you're seeing startups who are not able to make it right now.

There's companies that can't keep up, and if you can't keep up, if you can't have a go-to market strategy that allows you to pay off that investment, you're going to be in trouble.

And I think what's kind of, I guess, weird right now is you have a lot of small companies coming out of this AI revolution, but you also have a lot of medium-sized to almost large companies who haven't been able to figure out how to keep up and haven't been able to figure out that R&D budgeting, and they're going out of business.

You know, the layoffs that you see, chat out layoffs.fyi, by the way, don't know who created that website, but incredible resource.

It's crazy how many people are losing, you know, folks and then investing in different kind of folks, the whole, you know, Facebook, we're going to get rid of our low performers, but then invest all the money in buying, I guess paying for really expensive ML engineers.

It's just a wild time to be in technology right now. It's really hard to predict what's happening in 3, 6, 12 months.

Christine: Yeah, so the way that I kind of see this in my own mind just to like frame how things have been changing is that like--

People constantly make things do more and add more complexity until they cease being able to manage it, and it's software, the way that we've sort of dealt with that is that we continuously add these like new layers of abstraction that allow you to like do more while still keeping the complexity something that can be managed.

And you know, it went from like, you know, we're programing in Assembly to now we got a high level language C to, you know, interpreted languages, and then we had shared libraries, and then we're like, oh, well what if we had shared libraries that somebody else manages all of the operational complexity of with, you know, APIs and sort of SaaS businesses.

And now all of that is sort of being supercharged by this AI that actually works.

Everybody's talking about AI for decades, but like it actually hasn't been useful, and now we actually are in a point where people can make products that the user experience is way better by using actual technology that exists today.

And I just feel like this like sense of being on this treadmill is like kind of core to like human nature.

Ben: Yeah, constant revolution. I think I feel like that all over the place right now.

You know, just being pushed in all different directions, not just, you know, with the greater situation that's going on in the world, and then also at the office on a regular basis.

There's so much change happening. I recently have been feeling bad for not using AI enough, like even in my personal life, you hear people talk about what they did with Grok or what they did with Deep Seeker, what they did with ChatGPT, and I'm like, oh man, could I have been more efficient in my day if I was using a tool like that?

I'm having to like train my mind to use that as a resource.

The other day I was putting together a proposal for something and I, you know, looked at it, and I was, you know, reasonably happy with it, that I remembered to run it through one of these solutions that it was a lot better.

You know, there's the AI gobbledygook, you know, crap, but then there's an ability to like enhance an idea that I have and use these tools in order to do that. And it's pretty amazing.

Isaac: I'm going to ask a bit of a contrarian question here and kind of mind this out.

I really love that aphorism of, it's never as good as you think it is. It's never as bad as you think it is.

And man, AI has definitely changed a lot of the way, like we've even show up at work, right?

I'll say it's changed the way I write emails, it's changed the way that I run a lot of my daily life personally.

But I do think there's a bit of smoke and mirrors in some places, and, you know, I'd love your opinion on a few of these things.

So for example, one of them being, I've seen a lot of, you know, theories that some of the layoffs that you see that companies are laying people off and saying, oh, we've replaced them with AI.

They're actually just trying to mask over, you know, business fundamental issues and saying that it's progress so that they don't lose esteem or value in the markets as one.

And another one being that, you know, you see a lot of these AI features come out and adoption of them, and some of them are extremely low, right?

But of course have huge cost implications for the business. So I'd just love to mind that out for you and just see what you're thinking there.

Ben: Yeah, I see those as two very different things.

I completely agree with you on number one, like I think that a lot of these companies became tremendously bloated, especially the marquee ones.

The, you know, Googles and Facebooks and Amazons of the world became incredibly bloated.

They still were profitable, so the street wasn't putting that much pressure on them to get efficient.

And then, you know, post Covid, you know, economic changes and then also, you know, for good or for bad, and I mainly think for bad, you know, what happened with X and how public it was, I think there's a lot of people who are like, okay, we can be like that too.

It's hard for me to tell, and I'm probably not smart enough to know what that means for those businesses, you know, did they actually shove away talent that would be valuable for them to have, or were they just truly bloated, and this makes them more financially efficient?

Or for some of the ones that are, you know, mid-market, you know, do they have, to your point, do they have no choice? Were they just using this as an excuse to shed employees?

So that I a hundred percent agree. I'm not at the point where the world has been, you know, completely automated.

That being said, you know, for number two, you know, every tool now has an AI feature, and you know, do we need it or not, how much value?

And I completely agree with that. I'd say I look at it in two ways, one less cynical and one more cynical.

The less cynical one is, I was recently in the CRM business and communication was probably the highest value offering that we had with our customers.

And there was a lot of pain in that communication. You know, we were helping apartment owners communicate with renters.

And one thing that we found is that, you know, for as many inquiries as an apartment owner got, very few of those were actually really qualified and serious about renting an apartment.

It's a lot of people, and there's also some bad actors in the space that would, you know, not really be a real lead, or you would get leads said to you that were not really real and AI really provided capabilities that would help not only the business with trying to identify who really is a qualified lead before they connected them with a human being, but also the consumer.

You know, one of our big problems was for our customers is that they didn't have 24 hour service.

And so being able to help somebody, you know, when they're looking for an apartment after work at eight o'clock, you know, in front of the computer, that was a huge advantage that AI was able to provide.

So in that way, you know, depending on the area that you're in and what you're doing, there are some really, really good use cases.

On the other hand, you know, every piece of software now that I use has an AI capability. And a lot of them, you know, and a lot of them are overlapping.

It's like, do I need AI in Notion and AI in Slack and AI in whatever other document repository that I use? And the answer is probably not.

And a lot of times I'd rather just navigate and browse instead of having to ask a question in language. I can find that tedious at times.

I actually find, you know, a lot of the interfaces that we've created are actually pretty efficient.

Do you really need to ask, you know, have a conversation with something in order to get the data?

So I think it really, really depends on the use case. Maybe if I just go on a quick tangent.

I've been thinking about something a lot recently, and if anything, the idea that I had, I mean this is too vague, but the idea I had was actually, there are AI solutions that are in place that actually are in place because the system itself is inefficient.

Like, let's have you talk to a bot because we don't have a really good digital front end. And I think that's an area that'll go backwards as well.

It's like sometimes when you have to talk to a bot, it's because your product sucks, and if you had a better interface where people could easily navigate, it'd be much better.

Christine: Is that kind of like when at the airport, they make sure that you have to walk a long way until you get to the thing to pick up your bag?

Ben: I mean, there's a lot. I mean, think about, I mean, the airport, I'm not exactly as clear on that reference, but you know, I never want to talk to a bot. I want to do everything on my app.

And the problem, the reason the bot exists is because the app fails when I'm on Delta and I want to change something, and they don't allow me to change my seat for some business logic that I would never understand.

And then I've got to ask a bot to change my seat. That's incredibly inefficient and doesn't make any sense.

Christine: Yeah, maybe that was a little too oblique.

I wasn't referencing bots, more about the user experience of like people don't like to get to the baggage carousel and have their luggage not be there yet, so people make sure that it takes a while to get there.

Ben: I never even thought about that so I just learned something myself.

But there are tools that are coming out now that are trying to replace shitty business process, shoddy interactions with a bot, and at the end of the day, like, no, just build your software to work and you won't need a bot.

Isaac: So yeah, I really actually agree with you.

I think that you're going to see kind of maybe a little bit of Darwinism happen on the application side because I think people have kind of forgotten when you're building software that UIs are nothing but very glorified API request builders.

I mean, there's some exceptions to that, but I mean, technically you could use a lot of these systems that you use using your command line and just making manual curl requests.

It would be horrible, but it's technically possible. And I think from like a UI and a UX perspective, we often get away from that.

I am curious actually what you see, or if you have any ideas around what user experiences will look like in like five years, because I know I'm not going to be chatting with a bot on my phone for all these kind of micro things I do.

But they're going to look different than they do today as far as like, you know, here are the buttons that are here every time.

Ben: That is a really interesting question. I'd almost turn the question, it seems like you may have some ideas about this.

I think that, you know, sometimes my ability to see in the future is limited.

I think that there will be a lot more pressure on folks to have better experiences because if your experience sucks, it will be replaced by a bot.

I think that is something that is a real challenge, but man, the technology is moving so quickly, and it's taking one, you know, step forward, one step back.

You look at the, you know, the Vision Pro and Meta Quest and the, you know, potential productivity utilizations, and that's not going anywhere anytime fast.

But I will say like, you know, with the maturity of people's phones, there's a lack of innovation in some of that space.

You know, like I didn't give a crap for my iPhone 16 versus my 15 versus my 14.

Nothing has really tremendously changed about those kind of consumer experiences.

And I don't know if I could predict what's coming next, but I'll tell you that I am waiting for something, you know, I want to see a change.

And the change for me is not talking to Siri. That doesn't really get me excited about life, but being able to get, you know, better answers quicker and be able to--

I'd love, you know, there's been a dream for years of this kind of like contextualized, I dunno if I would say interfaces, but experiences.

Like, you know, I'm going on a plane today so I'll get checked in on time, or you know, I need an Uber so there'll be one waiting for me when I get up because it knows that I'm, you know, there's people discussing that, none of that actually, you know, has been built.

So maybe there's still opportunities to use technology to kind of wrap that all up. But what do you think?

Isaac: I mean, I actually really agree with you. Zillow's CEO, I think at some point, called it the race to intimacy, right?

The phone is a great example of it. It has all the context, it has all of the data of what's going on in your life.

We're kind of at the point where technologically there's nothing holding us back from being able to provide those perfect contextual things.

And I mean, you said, you know, it's not going to be talking to Siri. I really agree with that.

There are little inklings of it though, right, where you open up your phone, and there's like this suggested, hey, you should turn on your do not disturb 'cause you're at a dinner reservation or something.

That's so primitive, but even that feels like magic compared to manually navigating the UI. And you're right, these experiences haven't changed at all in 15, 20 years.

And I really think that when you move from a place that's very optimized, right, like our current user experiences are insanely optimized, into a new place, it becomes very unoptimized, because you haven't had that time to dial it in.

I would guess that a lot of these companies have really solid stuff prototyped internally, but they don't know how to get it to that level of polish that wouldn't feel like a massive, massive step back from user experience point of view when they first roll it out.

Ben: And I guess to the point that you're making with, you know, so much conversations about privacy and data usage, that a lot of the best solutions would be the best solutions if all these apps could share information with each other.

Yet that goes against the grain of, you know, like the risk of everybody's data getting exposed.

Like, you know, if my, if Zillow could talk to my bank account, and you know, Zillow could look at my Amazon for my purchases, I bet they could find the best house for me in the entire world, but I'd be afraid of that.

I mean, maybe there's an opportunity for an agentic bot or whatever you want to call it, a bot agent, that could look across those things that would be, you know, in some future date on device as opposed to having all my data in the cloud, and it would just assemble that for me.

Like, I would love that future. I just don't know how close or far we are from that capability.

Christine: Do you think that's another sort of pressure that leads to there being sort of more unified platforms that just end up having to do a lot, because not everybody has access to all the data or wants to connect and gather all the data, and maybe, you know, you got to be a certain amount of functionality in a user's life in order for them to be like I trust you to be mediating all of this context about my experience?

Ben: I don't know if I've heard anybody say this recently, but if you think about ecosystems, there's really like two and a half at this point.

There's Google, there's Apple, and maybe a little bit of Microsoft from like a productivity perspective. Aside from that, they don't really exist.

So there is obviously, at least from a consumer perspective and to a large degree a business perspective, we're already kind of trapped in those worlds.

Those are the folks who have the majority of our data. There's a social component, but I think that a lot of that is more ephemeral.

You know, there'll be another TikTok one day, and it'll come along and I think take people, but from like, you know, where you're doing your work and how you learn about things, I think there is already a tremendous advantage to platforms.

I do think, you know, just from having this conversation, maybe this is where agents will change the world because they'll be able to open up your phone, they'll be able to open up your, you know, across Google and Apple and all that sort of stuff.

I just, I don't know how far or close we are to that lifecycle. I've tried using some of the agent, I had a couple of use cases to use with some of the new agents that are available from folks like OpenAI.

It didn't really, it wasn't as any more efficient than I was for some of the things I needed to do. I think it's going to take a lot more work to get there.

Isaac: Talking about AI agents.

I feel like there is a bit of a like a Rorschach going on where people, whether it's, you know, varying levels of technical knowledge or just exposure to different markets kind of mean different things when they say AI agents.

What are you talking about when you say AI agents?

Ben: So that's a really, really good point.

I was maybe discussing it in the most bullish or most extreme version of the idea that I could have like an assistant that sat by me, whether it was in my personal life or my business life and be able to handle tasks for me.

You know, without a lot of input, learn who I am and then, you know, do what I need. I think that's a very idealistic state that we're pretty far away from.

You know, I am aware though, you know, in a much more narrower context, the idea of almost like a, instead of sending something to a person to do a task, I can now ask an agent to do it for me, especially something that's very specific like extract this data from this spreadsheet, get really good at it.

I'll work with this robot over time, and over time it'll be able to do it, and then I'll just treat them like another part of my workflow.

I think that's a more realistic, you know, near term goal than the idea of something, you know, that solves the problem of finding me a house on Zillow by looking at all my data.

So I keep on wanting to ask you what your opinion, you know, maybe you're thinking about them in a more unique way.

Isaac: You know, I think that the lines are a lot more blurred for me when someone says AI agent, because I think some people are thinking of like, almost like you said the equivalent of having another employee in seat, and there's a certain attraction to that because that's infinitely replicable, right?

If you can have one employee in seat, then you can have a hundred employees in seats, and then you start to get to those dreams, which feel a little bit too romanticized of, you know, the one person unicorns or whatever.

But for me it's more just about giving the AI the keys to the car, right?

Letting them actually take the wheel and drive rather than using them as kind of a consultant, which is what we're doing right now, where, okay, I'm writing an email, boom, pop, pop open a ChatGPT, paste the whole thread in, let it create a response.

I'm skeptical of it, I go back and forth, I edit it on my own, and then I come back in.

What I view is when we're at agent status, that is where I can send it off to the AI, and I'm feel very trusting and comfortable that the outcome will be tantamount to if I handed it to an employee.

So maybe it's closer to the seat thing, but sort of just getting back into that element of trust, which personally I've lost a little bit over the past even year, year and a half.

I remember in December of 2023, if I was doing some programming, I could paste a million things into ChatGPT of like, hey, here's my database schema, here's how I do things.

And I could say, here's what I'm trying to do. And I could get up and walk away, and I could come back like 10 minutes later, and it was just finishing up the response.

And now what I'm seeing is you've got things like, you know, Cursor and Claude or like all of these different other options available to you that really have a lot more technical ability, right?

They're better in the LeetCode scores, and they have your whole code base ragged, but they don't feel as high effort.

So I actually feel like I trust them less to just get the job done, somehow.

Ben: But is that, it seems like that might be a perception problem as much as a reality problem, I don't know.

Isaac: Perhaps, you know, it's very tricky because you don't really know how these companies that are providing LLMs as a service are changing them under the hood as they go or reorienting them even little bits.

Let's call it like the minor versions of the patches you don't really know about.

There have been clear changes in the behavior, but what I think it is, is I really am a big fan of Goodhart's law. Spang knows that I cite this internally maybe once.

Christine: Once a week, at least.

Isaac: Once a week, okay. Which is as soon as you announce something as the measuring stick that you hold something against, you kind of invalidate it as a measuring stick.

And when we're saying, hey, we wanted to get higher scores on the SAT, we're judging our LLMs based off how they do on LeetCode, these tiny little problems.

Now you've incentivized optimizing around those specifically rather than the grander problems.

Ben: That's a really good way of looking at the world. I think that my use cases are probably sometimes less technical, you know, as a more pure product person who doesn't have an engineering background.

A lot of times my debating in my head is, how long is it going to take me to communicate to the bot well enough to create like a specification for me versus just writing the specification?

You know, I've kind of gotten to a point where if I want to communicate something to an engineer, gotten pretty good at it, and you know, putting pen to paper is not a huge pain point.

There's some value, I mean, like where I've seen value is, if I, you know, now working for an even smaller company than previously used to be a real pain in the ass to write acceptance criteria for a piece of software, and now a bot can do that in a pretty reasonable way for me.

But the actual, like if I really am trying to be creative, I don't find that much value of trying to explain over and over again to a solution exactly what I want, you know, this specification to try.

Actually I was using Bubble, which is that, you know, no code software solution.

I spent like 45 minutes trying to get it to do the simplest thing, and then I was like, this is really dumb, I'm going to just do it myself.

You know, I was trying to build like a little widget for a blog for a personal project I was looking to, and it was completely useless.

I could just go into WordPress and figure out how to do it on my own.

Christine: Do you ever prototype stuff using coding tools to like share requirements or help communicate with the engineering team?

Ben: Yeah, I mean, that's kind of what I was describing a little bit.

Like are you talking about actually, let me clarify. Are you talking about actually like creating a coded prototype off of like my?

Christine: Yeah, like here's what I want it to do roughly, you know?

Ben: I feel like I'm not sophisticated enough to potentially want to spend the time to figure out how to.

Christine: It's not faster?

Ben: Yeah, I mean, maybe it is, but I'm not, maybe I'm too old to, there is a certain element where I do feel like I am the proverbial dog that's not learning any new tricks.

But I think what I have done in that world, I've tried like the interface design stuff, and I've been like mixed about how actually good it is.

I mean, I'm still one of those people who uses Balsamic, and it takes me 10 minutes to put together something that usually really well demonstrates what I want to do, a lot of downsides for that.

A lot of people make fun of that product, but its simplicity is what I am trying to go for. And so that's enough for me right now.

I haven't gotten to the point where I feel like the bot is better at putting together a wire frame than I am at putting together a wire frame.

I'm sure as it gets more and more trained on interfaces, it'll mature, but you know, that's kind of the fundamental question that we're all thinking about is, when does it actually start replacing somebody that does something creative?

Like it's one thing to replace a mechanical process that needs to be repeated over and over again and can be 98% accurate. It's a very different thing to say, you know, we don't need product management anymore. We'll just rely on AI to do it.

Isaac: Very much agreed. Okay. I think we're actually at about time.

But before we wrap, we have a section called "Picks," which is where everyone just brings something that they're excited about, maybe that you've found over the last week or two that, you know, you just can't stop thinking about.

And it can be anything, it can be an application, it can be a piece of media content, it could be a something for your home kitchen or something that you found recently.

So Spang, do you have a pick this week?

Christine: Yeah, so I recently started watching the show "Vikings." This isn't a new show and launched in 2019, but I'm really not a big TV person.

I basically only watch TV when I finish the day too tired to read a book.

But I've been really enjoying this show, and I think one of the reasons that it appeals to me is because in like the first couple episodes, the main character basically like recruits people from his village to sail to new lands that are richer so that they can steal things from churches and bring them home and feel good about themselves.

And he is got like all the people in, you know, a room in someone's house somewhere.

He's trying to get them to join on the voyage, and they're like, "Why are we here?"

And he's like, "Because you have nothing better to do, and your lives are meaningless, but we should go west."

And I really just feel a deep connection to that desire to have more, you know, he just like, he really wasn't content with being on his farm and just like having a nice life.

And I think that all like startup people and builders sort of like come from this line of like, we're just like not content with what we have, and I don't know if this is going to work out.

Maybe like it's just all going to crash and burn. We're going to sail off, and there's not going to be land, and we're going to be lost at sea, there's going to be a storm.

But like, it's better than being here with my very known and very boring life that I don't feel inspired by. So psyched on "Vikings."

Isaac: Good pick. Also, I'm a big fan of "Hardcore History," if you've ever listened to that, or it's a podcast by Dan Carlin.

He does like eight hour episodes and he has a series, I think it's called "Twilight of the Acer."

And having watched "Vikings" and then listening to that, it felt like it was just the perfect amount of like base knowledge, and it really made it special.

I highly recommend it. That's not my pick though, okay?

Christine: I'll have to check that out.

Ben: Yeah, my pick is mastra.ai, M-A-S-T-R-A and that is something that Versal just released recently.

And I mean those guys are just killing it in every way, shape and form right now on the AI front.

But what I really like about it is it is an agentic AI workflow builder that's code first and native, right?

So like you, you actually run it in your code base, and I think it's pretty cool because you can sort of start to replace the, like the n8n, if any of you guys are familiar with that, of these kind of visual Zapier for AI agents and start to stitch into your actual product in the actual code base natively some of this AI agent stuff.

And I don't think it's quite where things are going to end up, but what I view this as is, you know, you watch like a march of the penguins kind of thing where all the penguins are standing on the iceberg and none of them are jumping in until one does.

And then they all jump in. This is starting to feel like the first penguin jumping in there.

And I think you're going to see from this over the next maybe six to 12 months, it just be very natural and common to pull this down like a normal package and start to stitch a little bit of AI functionality into every app, even ones that aren't really AI first.

So yeah, it's been a joy to play around with actually.

Ben: Awesome.

Isaac: Ben, do you have a pick?

Ben: Yeah, this is kind of an oldie, not a newie, but I recently rediscovered my passion for Kickstarter.

I miss, you know, really strange gadgets as something that would come out all the time.

I mean now every gadget's an AI gadget, and again, I don't want to talk to my golf ball or you know, wear a pin that'll tell me what's going on in the world, not something that motivates me.

But I do miss like really strange electronic products that, you know, weird people are making. And so I've been funding a bunch of Kickstarters.

I recently got a product called the Glorb, which is kind of like a disco ball that is color, you know, covered in different colors, and you can kind of program it with your phone, and I think it is the most fun thing in the entire world.

So I have a soon-to-be-8-year-old daughter next week, and me and her were playing with that, trying different patterns, going crazy about it, and it was the most fun.

So I really enjoy that. You know, sometimes the consumer electronics, I'm completely bored on my phone, you know, I want something new, I want something strange, I want something fun to play with.

And so, you know, my New Year's resolution was just buy a lot more weird crap and see what kind of creativity it might spawn.

Christine: I love the name, the Glorb.

Isaac: Yeah, it's good.

Christine: Just feels kind of gluey.

Ben: Yeah, it's amazing. I would highly recommend taking a look at that product.

Christine: Sweet. Well, thanks so much for sharing your pick and thanks for coming on the show.

This has been really fun, and we really appreciate you taking the time to join us.

Ben: Yeah, thanks for having me.