Jam dev
===

Matt: [00:00:00] Hello, everyone, and welcome to PodRocket, a web development podcast brought to you by LogRocket. I'm Matt, CEO here at LogRocket, and today we have a very special episode. We're joined by Danny Grant, CEO of Jam. dev. Ricky Bourbonette, VP of Developer Relationships at Cloudflare, and Bi Yung Liu, the CTO and co founder of Sourcegraph, to talk about the DevStarterPack stack, an alliance of like minded dev tools for the next generation developers building the future of technology.

The stack not only includes tools like Jam. dev, Cloudflare, Sourcegraph, Prisma, but it also includes a log rocket to provide AI first session replay and analytics. So excited to get into this conversation with you all and explain what the stack does and why developers should look into it. Before we get into the discussion, I'd love if each of you could give us a brief overview of who you are and what tools you represent.

Dani: I'm Dani. I'm the CEO of Jam. We help 150, 000 developers fix bugs faster.

Beyang: Hey, I'm Bian Liu. I'm the CTO, co founder of Sourcegraph. , we're creators of [00:01:00] Kodi, the context aware AI coding assistant.

Ricky: What's up everybody. I am Ricky and I work on the DevRel team at CloudFlare. And we have a bunch of rad stuff for developers that we'll probably talk about during this podcast.

Matt: So maybe Danny, you're really the mastermind behind the DevStarterPacks. I'd love to hear what Why this exists and why dev should use the stack.

Dani: My co founder and I met when we were product managers at CloudFlare. We are students of CloudFlare. And the thing that we loved working there and learned from is CloudFlare believes in giving away the tools that the giants have access to every developer for free. And I've always found this really inspiring.

10 years ago, when I first found out about CloudFlare and then started cold emailing everyone who worked there to try to get a job they launched something called universal SSL, where they gave away website encryption for the first time for free. And then one night they doubled the size of the encrypted web and their whole ethos is, Whatever the tools, the giants have access to give it to individual developers.

They can build the future even better. And so dev starter pack is [00:02:00] just that we're teaming up with CloudFlare for the 14th birthday and with all of you, and we're giving away tools that usually are very expensive, our enterprise grade, but help developers accomplish more as a small team so they can go and build startups. 

Matt: Very exciting. And what a unique opportunity for everyone. Be young maybe you could speak, I think Sourcegraph often known for enterprise wide search, maybe you could speak about the impact it can have for smaller teams and why it's included in the dev starter pack.

Beyang: Yeah, definitely. First off, I just want to say thanks, Danny, for organizing this and we're super pumped to be a part of the dev starter pack. So Sourcegraph's mission is actually to make it so everyone in the world can code and more specifically to make it so everyone can participate in the code economy.

So not just writing code for kind of side projects, but also to write code for work, for the things that you make a living off of. And we started off that journey building, great code search because that spoke to a really big pain point that we heard from [00:03:00] across our friends trying to write production code.

And lately we've been tying that to AI and plugging that code search into power the world's best coding assistant In our opinion, which is Cody. And so we're always looking for opportunities where we can make accessible that tool set to an audience that would be really well served by that.

And we've previously offered Cody pro which is our pro tier for Cody. We also have a free tier. If folks are interested, that requires no special access, but the pro tier, has. Has things like greater rate limits and access to better models. And we previously offered that to open source maintainers, because that was an audience that we felt, was really well deserving of.

Great tooling. And I think with the dev starter pack, we're excited to reach people who are early on in their careers as well. We want Cody and Sourcegraph to be this sort of enabler and accelerator. And to some extent equalizer for folks just getting started in their careers. And it gives you a leg up and it gives you sort of these superpowers that otherwise would take a long time to [00:04:00] develop.

Matt: And I think we're all waiting for that. The two person company that becomes a billion dollars because they use Cody and other AI tools. Maybe the first will be from the dev starter pack. Ricky, maybe from the CloudFlare perspective how do you see the tools and where should developers get started with CloudFlare when they're first starting their companies?

Ricky: I feel like Danny did such a good cloudflare pitch. I'm like, Oh my gosh, what do I have left to say after that? But for me I think about the saying that talent is distributed equally, but opportunity is not. And when I think about the starter kit in general, that is the problem.

I feel like we're trying to help with is that a credit card. Or not shouldn't stop you from being able to build with some of the stuff. And especially as AI is getting more involved. So stoked about Cody and , my daughter just did a thing with cursor, which I was very excited about. And I think we're going to see more and more people building that couldn't before.

So that is just [00:05:00] my overall take on why we're stoked to be part of it. For CloudFlare overall. I think a lot of people don't often realize we have a developer platform that you can build and deploy applications on. And we have a database and we have AI inference. And my hope is just go and play with all the stuff and see what helps you and see what you may get stoked about. 

Matt: , your team has built so much over the years. I think a lot of us still are, Oh, CloudFlare is a CDN, but It can do everything for you in terms of infrastructure, which is super exciting. And then, yeah, maybe Danny with more AI code, I always think there's gonna be more bugs.

So where does jam. dev play into that

space? 

Dani: interesting. So most of the major AI companies at this point use Jam, which is so cool, and they use it in a really special way. Basically, the old way of reporting bugs doesn't work if you're building an AI product. Because before AI, if you encountered a bug, you would just stop, and then you would record your screen, and you would go through the bug, and then you'd send [00:06:00] that recording to an engineer.

But if in an AI world, if you're building an AI product, the AI inference won't output the same thing again. And so you can't go and rerecord the bug, but Jam is an extension to your browser that lets you grab the last 30 seconds of what happened on the screen, played back like a video, plus all of the Console network logs and send it to developer in a link.

And so these teams building AI products can use it to improve their prompts, share feedback with developers and grab what already happened because they can't go back and rerecord. So for all the developers out there who are about to build their new AI startup with the tools in the dev starter pack if you're building AI features, it's a faster and better way to share bugs.

That happened because of AI

Matt: Yeah. And then finally for a log rocket, once you're then in production, inevitably some bugs will get through and some issues will happen to your customers. What better way to know about them ahead of time and then be able to reproduce and fix it with LogRocket, where we combine session replay, error reporting, analytics, all in one, so you can see all the [00:07:00] problems affecting your customers.

Yeah, between Kodi, JamDev, CloudFlare, LogRocket, and then all the other tools in the starter pack, you're good to go. You don't have to think about any other products.

Beyang: It seems like we were covering the entire SDLC pretty well.

Dani: so this was actually, this was the intention when we started talking about the dev starter pack, it was actually Matt, it was with a conversation with your team and we were talking about how we spend a lot of time with our users talking about what tools they should use. It seems like developers today are overwhelmed with the number of tools out there and instead of getting right into building mode, they're actually spending a lot of time up front in research mode.

And so we wanted to. Provide not another marketplace, but rather a stack. This is like a stack of tools you can use altogether and you don't have to use the stack. They all are using the same promo code. It's, we are trying to make it really easy to get started. And that way you don't have to research a bunch of tools.

You can just start building.

Matt: makes sense. It reminds me back five, ten years ago when JavaScript was exploding. Everyone would spend their time [00:08:00] thinking, do I use Angular? Do I use React? Do I use Node? Do I use Python? And then people started, like the mean stack came out of that. Or Meteor was this opinionated stack. So I, you, it's really exciting to see the sort of everything come together with an opinionated view of what tools you should use.

how did you select the tools to put into the stack? Or what was the thought process behind each category?

Dani: I am a super fan of every tool in the stack, and I'm so excited for people to use them. These are tools that prioritize a great developer experience. When you meet the people that work there, they are so positive, and they love developers, and they love the mission, they love the cause. I also am really excited about what tools will join us.

So I see these first ten tools as the beginning, but the stack should grow, and so if other people are building tools out there and want to join, there's a link on the site to, to join us and I hope the stack will grow over time.

Matt: Absolutely. Maybe, yeah, let's talk about AI. We talk about how the next wave of startups will be AI native curious. What does that look like to folks? And maybe [00:09:00] beyond you have a perspective here of do you see AI being the basis of how people are now writing their initial code and developing code or how have things evolved now with AI?

Beyang: Yeah. So I think AI is a, is very powerful enabler for whatever software you're trying to build. So for startups, I think there have been a bunch of demonstrations of the capabilities of LLMs at building the scaffolding or building the sort of like boilerplate plus of simple applications.

And that's really powerful because previously, You take any like developer off the street and ask them to build that scaffolding. Oftentimes the first couple hours, maybe even like a whole day would be spent like researching what tools to use and gaining familiarity with those tools.

And so the activation energy to just get to a point where you have something basic was really high. And as a result, I think there, there are a lot of interesting app ideas that simply just never got built because the upfront investment was too high. And I think one of the things that I can do for us is just lower the activation energy.

So if [00:10:00] you want something that's simple, that's, not too much. Unlike what's already been built out there before. Most applications are like variations on some theme that's been played before. You can go do that in a matter of minutes or maybe even seconds now. And then that gets you up and started.

And then after that, I think you need more of a human in the loop to guide the evolution of the code and the program toward your exact specifications. That's something that we've tried to implement with Cody by giving access to code search and more understanding tools of that code. But I think, at all stages of development incorporating AI in the right ways can just be a huge accelerator.

Uh, It can make things quicker and it can make things possible that before just would never have been explored because the sort of like upfront investment or at least the perceived upfront investment was simply too high.

Matt: The ability to broaden the access of code to more and more people and people who maybe took one CS class in college teach them how to code, I think is super [00:11:00] exciting. Is Cloudflare doing much with AI these days or how to Cloudflare? See the whole AI market emerging.

Ricky: yeah, so much. Such a great question, Matt. So we, I think are very excited about two sides of it. First is What is possible to build thanks to these language models to these image models, right? All the things that couldn't be created before. And we are also a big believer in the open approach to things that cloudflare.

So you see that a lot in the things we do trying to. build them in a way that is open source first. And so we have where a bunch of open source models run on our network now. So we took GPUs, we like put them in our data centers and we run like llama three and stable diffusion and open source models.

People already know and love, which to me is super exciting because as a developer, I like got super stuck in the beginning days of the open source models where it's okay, I'm running it [00:12:00] locally, but I've like never deployed a model before. I don't know how to go from Oh, llama to production.

And so that's the first area we see. And then second is definitely just more people getting access to it. To code before I think developers do their best work when they're on the creative parts of coding and not just redoing the scaffolding or crud and LLMs are incredible at the scaffolding and the crud and let developers focus on what's novel and unique.

Beyang: I think too for something like Cloudflare, it's such an incredibly versatile and powerful platform and there's so many different like elements to it where, even now, having coded for, I don't know, it's going on probably like a decade and a half professionally and even more that personally.

There's many aspects of what you do that I don't fully grok or understand, because you can get pretty deep into the weeds of networking and things like that. And before tools like Cody before the current revolution [00:13:00] AI existed there would be a lot of stuff that I just wouldn't explore, right?

It's even though the docs are really good, it's just it's going to take too much time. I will just, I'll have to delegate that to someone on the team, or more likely than not, because the team's at bandwidth, we just won't explore that. But, now any part of that is, is just way more accessible.

Because you can go and ask an AI powered tool, Hey, , how would I get started with this part of the platform? How would I use, Terraform to describe what sort of like DNS rules I need laid out. And you get a pretty decent answer I found that. So that kind of gets you roughly in, into the area of where you need to be.

And then the rest of following up with that is, is just far easier than, starting from scratch as you would before.

Dani: What you're saying reminds me of serverless. The trend used to be more and more abstraction of the underlying infrastructure in development. What if you didn't even need to know how to terraform your DNS? But now it's the serverless of the front end, maybe, or like the serverless of the application development itself, where there's more and more abstraction.

It reminds me of that movement. It feels like a continuation of the same thing.

Beyang: I, I actually I [00:14:00] love this because I think things like serverless be before when we were trying to make more of deep technology accessible to a wider audience the answer would be to introduce a new form of abstraction which would lower the activation energy, but also come with its own sort of trade offs.

And I feel like now. Especially now, if you're like a two person startup, you don't necessarily have to make those trade offs anymore. You can use like the underlying technologies. AI can teach you how to use the underlying technologies so that you don't have to take necessarily like a big proprietary dependency on some abstraction that hasn't yet been proven.

I think what we've seen with a lot of the serverless stuff it's maybe fallen a little bit short. With the exception of, I think like R3, which is really cool. But a lot of the other stuff, I think you're starting to see this like swing back towards just vanilla dock containers and things like that.

And these are technologies that are not immediately. Accessible, but they're open source. They're really powerful. And if you just get over that initial hump of how do I build an MVP on top of that? They're actually pretty easy to use.

Ricky: Yeah [00:15:00] I love that point and I feel like AI is making the time to possible so much quicker because before as a developer, I'd have to find a tutorial in the docks. That was what I wanted. And then I'd make that and I'd end up with, a to do app, which is fine, but I have to edit that to do app into what I wanted.

Now I'm able to just, you Chat with an assistant and say, Hey, I want to use these technologies to build this thing. And my first version lets me know that maybe it's possible or maybe it's not possible. 

Dani: What I'm really excited about here is with time, the barrier for someone to make a positive impact at a global scale is going down, and this is just yet another thing before remote, it was like to start a company, you need to be in the Bay area now it's great developers are all over the world building software.

The movement is global. I'm excited with dev starter pack. That it used to be. To have access to the best tools, you need to raise a round of funding. And now it's to have access [00:16:00] to the best tools, just use this one promo code. And here are some great tools you can just get started with to start building,

Matt: Absolutely. Super exciting. , one area I'm really interested in being young is where are we in this sort of wave of AI? I remember 10 years ago, everyone's saying self driving cars are coming next year. Driver's ed will be gone. But yeah, here we are. And it's still now Waymo is.

Exciting, but is that could we expect 10 years or how do you see things playing out? When will we be not be writing code anymore? Versus is it going to take 10 more years to really see the fruits of all this labor?

Beyang: Yeah. It's I remember reading about the DARPA grand challenge. I think back in, I don't know, it must've been like 2005 or something. I was like in, in the middle of high school. I'd be like, that's cool. It's just around the corner and that now here we are, Waymo is here, Cruz is here, but I still can't take to the airport or anywhere else outside of this kind of seven by seven grid in San Francisco.

Still waiting on that. Still driving my car. I think the autonomous coding stuff is going to follow a similar [00:17:00] trajectory. So I think AI in production these days, especially in large production code bases, it's still a very human in the loop endeavor. So you want a human in the loop directing it and also checking the output because you do run into hallucinations and you do run into mistakes every now and then.

I think there, there's very promising direction. We've seen a couple of great demonstrations of agents stitching together longer workflows in code. But I think , most of the impressive demos you see there have been in the context of starting from scratch which is something that LLMs are just like innately better at because as a first rule, starting from scratch, you're well within the distribution of code that has been written before.

Because if you're starting from zero, there's not too many variations you can get into. But, once you have deployed that app into production and you've taken in a lot of user feedback and had to add all these like customizations and special features and that sort of thing. Now you are like very far.

along the path of being, unique and specific , to your own code base. And for that, you still [00:18:00] need a lot of human domain knowledge. You need a lot of human creativity. AI can definitely help. And certain elements of how you build that AI can help for instance, being context aware we find is really important.

Just as it is for human developers, to have access to tools like code exploration and code search. But I think we are still very much in the kind of lane keep assist era of development, at least as far as production use is concerned.

Matt: So no 

autonomous driving. 

Beyang: not not yet. I think we'll eventually get there and we'll start to see that for, specific types of bugs or specific types of issues that are more constrained. Maybe sooner than the rest. But it's going to be a long road. I don't think there's going to be any scenario where, we all wake up one year and, Oh, we're all out of the job because the code writes itself.

Now, I don't think that's going to happen. I think it's going to be very similar to the self driving car progression where we see increasing amounts of automation and increasing amounts of help for, humans in the driver's seat. And then we'll start to see limited applications in more [00:19:00] constrained settings down there.

Dani: Maybe one way to think about it is if you look at what happened with the internet. First we attach the internet to everything, and then through attaching it to everything, the internet changed what we fundamentally meant by those things. So first we put TV on the internet, and then the internet changed what we mean by TV, right?

Or first we put groceries on the internet, then the internet changed what we mean by groceries. Or maybe a better one is first we put restaurants on the internet, And then the internet created cloud kitchens. It's like change what we even mean by restaurant. There's no actual place, right?

And it's the same with AI. First you put AI on the car. So we've got like full self driving and Tesla. And then one day AI will fundamentally change what we mean by a car. First we attach AI to coding, and then one day AI will fundamentally change what we mean by coding. But These revolutions take like 20 to 40 years.

It's like first there's 20 years of just putting stuff online, putting AI on the things, and then the next 20 years afterwards, it changes what we mean by the thing.

Https: otter. ai We just live in this Twitter, Instagram, interconnected [00:20:00] world, but it feels like things move faster now than they did 10 years ago, and so part of me hopes like, oh, capital moves more efficiently and people think more efficiently, like we can actually do it in five years instead of 10 or 10 So part of me is holding a hope that actually things.

Matt: Move faster than we're all expecting but maybe that's wishful thinking on my part

Beyang: Hopefully some startup out there uh, using the dev starter pack will create something that's truly disruptive. I think that is sort of the battle, right? In, in all of our domains where we're selling tools to other businesses we sell to businesses at every kind of like stage of development.

We, I think all of us sell to very large enterprises some kind of like mainstay institutions of our current economy. And we also build and sell tools to very small startups. And I think there's just this constant battle or tension in, in, in a market economy of, the big companies that exist now that are providing a lot of value for the world as it exists today.

And then the companies that are imagining the future and building a [00:21:00] vision of the future that might seem totally foreign to the way the world works now. And I think one of the nice things that we have the privilege of doing is we see how both sides of the things work.

So at least at Sourcegraph, we take a lot of inspiration from both sides. We want to serve our big enterprise customers. What they do is really important, but we also see adoption of new patterns and new tools and new ways of doing things at the very early stage. And we totally value that set of users and customers just as much.

Matt: Yeah Well said if we can help a two person startup be as efficient as a stripe or ramp at shipping their own products That's would just get more progress in the world, which is super exciting So with that Danny, where can folks learn more about the dev starter pack? How do they get started? What's the next step for everyone?

Dani: Go to devstarterpack. io. It will show you the list of tools that you can just get started using for free today. You use the , Promo code DevStarterPack on all of them. And there's a link on that site to show us what you're building. We want to help promote what you're [00:22:00] building, support you. Today when this episode goes live and the launch happens the DevStarterPack is going to be in Times Square on the NASDAQ tower billboard, which is so cool.

Thank you NASDAQ. And I hope that some of you who are going to go and use these tools and build your startup one day we'll be back at NASDAQ to go public.

Matt: Awesome. You heard a dev starter pack. io. Go there and you have all the tools you need. Thanks everyone. Really enjoyed the conversation today.

Dani: Thanks so much, y'all.

Beyang: Thanks for having 

us.