Rethinking Software Development with AI: Scott Dietzen, CEO Augment Code – Ep 2

Welcome to What the futr Podcast — where we dive into what’s next in AI, tech, sales, and the human factor behind it all.

In the second episode of What the futr, Scott Dietzen, CEO of Augment Code joins hosts Sandesh Patel and Chris Brandt for a deep dive into the future of AI and software development.

They explore how the frustrations of traditional programming, combined with a passion for machine learning, led to the creation of Augment Code: an AI-powered platform that’s already doubling developer productivity.

From eliminating repetitive tasks to understanding massive, complex codebases, Augment Code is redefining how enterprises build software and making coding fun again.

Subscribe for regular episodes.

👉 Explore the platform: https://futrconnect.io

👉 Business inquiries: sandesh@futrconnect.io

Transcript

00:00:00:07 - 00:00:22:02
Sandesh
Yo, thank you for tuning in. To What the futr. Today we're diving into one of the hottest debates in AI. AI for code. What's going to happen with Software development? Is Computer Science still going to be a sought after major? And if my kids are watching. The answer is yes. Absolutely. Computer Science, Machine Learning, AI, Data Science, all gold.

00:00:22:04 - 00:00:46:00
Sandesh
Daddy approved. So to dive into that market, we tapped into an old friend, Scott Dietzen, also known as “Dietz”. Dietz, was the CEO at Pure Storage. He's now joined with Mike Spicer, who is also like Pure Storage at Augment Code. He is the initial investor and the founding CEO. So we've seen these two guys do it before. They disrupted that market and set the standard for flash storage.

00:00:46:01 - 00:01:15:07
Sandesh
So I love the people. I love the idea. But even more I like their go to market strategy and making it completely self-service for the customer. Could this be the future of tech sales? Just might be. Thanks for tuning in. Yeah, you got to change the game.

00:01:15:09 - 00:01:30:23
Sandesh
If you're talking about AI, you're often talking about how software development is going to be impacted by AI. There's a lot of buzz everywhere, and that's exactly what we're going to be talking about today with Scott Dietzen also known as Deets Thanks for joining us Deets

00:01:31:01 - 00:01:34:08
Scott
Sandesh, Chris great to be here. Thanks for having me.

00:01:34:10 - 00:01:59:17
Sandesh
Yeah this is going to be a good one. This is going to be a fun one. So what I would love to hear from you is a little bit around. How did you end up here? I heard you retired. Didn't know you're in in the game, but maybe you can tell us a little bit about your career, your journey, and then how you ended up at, at augment.

00:01:59:19 - 00:02:30:06
Scott
Yeah. Let me go back in time first. So I actually, completed a PhD in machine learning, a very long time ago. A different era of AI. It was the era of symbolic reasoning. But believe it or not, I had Geoff Hinton as one of my, AI professors, and I followed his career. And, you know, he he took a lot of grief back then because, the symbolic reasoning camp didn't think neural networks were ever going to amount to anything.

00:02:30:07 - 00:02:54:08
Scott
So, you know, wind the clock forward. And I was incredibly excited to see, large language models emerge, and their capabilities. And now, you know, in between, of course, I spent, 20 odd years doing for system software startups, the most recent, being Pure Storage. You know, every one of them depended on really large, complex code bases.

00:02:54:10 - 00:03:27:19
Scott
And had, you know, a brilliant team of engineers to try to bring those code bases forward and evolve them. And it was so painful. That process of, you know, trying to deliver a, you know, fast, performant, reliable, secure, software base and being able to to move it forward efficiently. And so it was the these two threads of my career, the pain of software engineering, and then appreciation for machine learning, that, you know, I had this fear of missing out, that these two, streams are going to come together.

00:03:27:22 - 00:03:38:00
Scott
And I very much wanted to be part of it. And I wanted to help software engineers because I'd spent so much of my career, feeling their pain.

00:03:38:02 - 00:03:58:16
Sandesh
That's great. That's awesome. So you you lived it. So now you can actually try to fix it, which is very cool. On this podcast, what we're trying to do here to try to pick winners, Scott. And we think that Augment is a winner. So, the structure that we have is I'm kind of the sales guy, and Chris is the technical side that actually knows the smart stuff.

00:03:58:16 - 00:04:14:00
Sandesh
So what we'd like to get into a little bit more is really what we would discuss in a sales pitch. So, maybe we start off at, you know, augment. Tell us a little bit about the company, how you guys started. What do you guys do?

00:04:14:01 - 00:04:38:12
Scott
Yeah. So, we had a, a small team there, two co-founders that I work very closely with, Guy Gar-Ari, who was, lead researcher at Google did, a bunch of the work that became the Gemini, large language models. And then, his, co-founder, Igor Ostrovsky, was the, chief architect of flash, Flashblade at Pure Storage.

00:04:38:14 - 00:05:03:04
Scott
And he had spent a year plus reinventing himself around AI technology. We had, you know, some significant system software expertise. And so, you know, this was before ChatGPT, but we were, you know, we were seeing the potential of large language models, to, to help manage code. You know, there's so much detail in tens of millions of line mano repos.

00:05:03:09 - 00:05:22:12
Scott
You know, there's just no simple changes in a code base of that scale. And it's very hard for humans to keep everything in their head while they're working through a complex change. Right. You make update one file, and then you're, jumping around the code base, and then, somebody interrupts you and you're like, oh, where was I?

00:05:22:14 - 00:05:44:09
Scott
You know, there's a lot of tedium and, and, and painful drudgery associated in that, in that task. And we thought large language models could help. So we got started. And then, you know, chat GPT launched. Along came GitHub copilot. And I would say the big development was all the vibe coding. Space, especially as the models got better.

00:05:44:11 - 00:06:08:07
Scott
But, you know, we started at the opposite end of the spectrum. You know, we set out to help software engineers. So these are individuals that are looking after really large, complex code bases. You know, maybe, for networking or storage or security or database or large, SAS applications. So we wanted to be able to tackle the big, hard problems.

00:06:08:09 - 00:06:15:09
Scott
And so that was, you know, what we set out, to, to go solve was, AI for software engineering.

00:06:15:11 - 00:06:23:17
Chris
You know, you mentioned a term there. I just want to make sure everybody understands vibe coding because that's that's kind of a big deal, right?

00:06:23:19 - 00:06:43:08
Scott
The term was coined by Andre Karpathy, who's just a brilliant, leading mind, in AI. But it comes out of, you know, people that say aren't software experts at all, but have an idea that they would like to make a simple little application out of. Maybe it's a weekend project. It's often also called 0 to 1.

00:06:43:09 - 00:07:02:16
Scott
Right. So I don't have any code. I have an idea. I, you know, I talk, I have a dialog with the AI, and out of it comes a simple, product that I can use. But you know, it's on the scale of, generally a weekend to a week, a couple weeks of work, to create something.

00:07:02:19 - 00:07:25:00
Scott
And that's kind of where vibe coding stops. The other end of the spectrum is, you know, the large enterprise with tens of millions of lines of existing code, that has, you know, so much built in history and knowledge in there that you need to be able to understand before you can actually be effective at evolving, that code base.

00:07:25:02 - 00:07:28:01
Scott
And that's where augment, you know, set our sights.

00:07:28:03 - 00:07:45:00
Chris
Yeah. Well, you know, and I think, you know, one of the things that's really interesting about what you're doing, is, you know, you see, so much like you mentioned, you know, GitHub copilot and things like that. And that's a lot of, you know, folks doing, you know, like using AI to, to do even vibe coding and things like that.

00:07:45:02 - 00:08:10:03
Chris
But, you know, you're focused on, on what's the bigger picture here. And I don't think, you know, not everybody entirely realizes how many artifacts there are in a large software development project and how nearly impossible it is to keep everything in sync and produce all that, too, right? I mean, can you talk about just the incredible nature of sprawl and some of these projects?

00:08:10:05 - 00:08:32:09
Scott
Yeah. If you look at, in open source land, I mean, you've got big projects like, chromium, for example, Linux, and so on. But, those code bases are actually tiny compared to some of the, code bases you find in the enterprise where they're not, 20 million lines of code, but 200 million lines of code.

00:08:32:11 - 00:08:55:01
Scott
Where no engineer, has any understanding of how all the pieces fit together. Right? You know, the best you can hope for is to understand, your portion, your end of the world. And these code bases are vastly too large to pass this context to any AI model. It's very computationally expensive, you know, cost the square of the size of the code base to try to pass it as context.

00:08:55:03 - 00:09:17:14
Scott
And for a large code base, there's just no way for the AI model to digest it. And so that was the problem. We had to, to go off and solve. How can we take, you know, a, a, a monstrously large and complex sophisticated code base and how we can we teach the AI everything it needs to know in real time about that code base to accomplish the task at hand?

00:09:17:16 - 00:09:41:03
Chris
Yeah. I remember an anecdote from a coder that I knew who worked for a very large OS developer that I won't mention, but he said he came across a piece of code and it said, you may think you know what this does, but you don't. Do not touch this, you know, but I mean, there's there's a lot of that even in like the biggest companies, you know, you have this institutional knowledge that, you know, doesn't get transferred.

00:09:41:03 - 00:09:47:17
Chris
You have a lot of stuff that doesn't get documented properly. And it gets really hard to figure out what's going on.

00:09:47:18 - 00:10:16:12
Scott
Indeed. One of our, early, customers, web flow, the chief architect had this epiphany, because, you know, his senior leadership team spent a significant portion of their time, I think it was 25, 30% onboarding new hires. And helping them discover how to be productive inside of their complex, code base. And, you know, that's, a substantial amount of senior architect time.

00:10:16:14 - 00:10:41:12
Scott
And the ramp time is tough on the new hires, and they've got to, you know, queue up to ask their questions and so on. They found that augment, could deliver that mentoring, you know, seven by 24, and do it as effectively or better than the chief, and lead architects on the team. And so, you know, you had, somebody that could give you a tour, help you understand the things that you were missing.

00:10:41:13 - 00:11:06:04
Scott
Might even look and say, had the. Hey, this documentation is inconsistent, with, you know, how this software really works. Can I fix it for you? Or, this code looks redundant. Can I, delete it? And, you know, reuse this code instead? So that ability, to, you know, kind of get onboarded with, much more help, is is transformative.

00:11:06:06 - 00:11:38:00
Chris
Yeah. I mean, one of the criticisms of AI has been that it's been focused largely in the sort of creative human work. You know, that it was supposed to take away the drudgery, but now it's doing all the graphics and the fun stuff that everybody wanted to do. Humans are being left to do the drudgery stuff. But I think that what's cool about what you're doing is you're really focusing on that drudgery stuff that is just way too much for people to, you know, like handle, you know, or would they want to even live in that world where they have to handle all that?

00:11:38:03 - 00:12:00:10
Scott
You know, there's there's so much, redundant and tedious work associated with software development. You know, all the time, you know, we're now in this situation where, you know, a new class gets implemented, and then the engineers says to the AI, hey, go crank out a set of unit tests and run them and let me know how it goes.

00:12:00:12 - 00:12:20:16
Scott
And report back to me if there are problems or, you know, if you find an issue, can you fix it? So engineers have the freedom to stay involved in the areas that they care about. And usually those are the creative, inspirational tasks, you know? AI doesn't have a vision today for how you want to evolve software.

00:12:20:17 - 00:12:41:03
Scott
You where you want to take it, what you want it to be capable of in the future. So that sort of long, longer term vision for the software is still very much the domain of human engineers. What we're delivering in AI is, is vastly reducing the friction between where you are today and where you want to go to.

00:12:41:05 - 00:13:04:03
Scott
So, you know, I'm actually really bullish about the software engineering industry because, you know, I think when software becomes a lot more predictable and you can get all of the features you want and you can get them reliably, I think, people will invest more in software rather than less. So even as we drive up productivity, I hope we'll have, the need for ever more software engineers.

00:13:04:05 - 00:13:07:06
Chris
I've got to imagine this greatly accelerates a team. What you're.

00:13:07:06 - 00:13:30:21
Scott
Doing. We certainly see that, you know, we we see teams delivering 50% two weeks, productivity. And it's and it's only accelerating. So the models have gotten so much better, you know, even over the past year, I think things really inflected, with, sonnet. Mid-year last year in terms of it, breakthrough in code understanding.

00:13:30:23 - 00:13:54:17
Scott
And, you know, then you could give, the AI a task that might take a human engineer, you know, 3 or 4 minutes. Five minutes. And it could go off and complete it. Now we're up to an hour kind of task. And the AI is, you know, evermore sufficient. At least ours is, because it has this deep understanding of the code base so it can take on ever more complex tasks.

00:13:54:21 - 00:14:15:07
Scott
And I, you know, I think I don't see any reason that that's not going to continue to grow. But at the same time, you know, we're very much about complementing human engineers, making them more productive, making their job more fun. You know, they're trying to take away the drudgery and the pain associated with evolving these tens of billions of line code bases.

00:14:15:09 - 00:14:41:18
Chris
You know one other thing, too, that you mentioned regarding this, too, is that there's the best code is code that you've already created, that you know, that works. And you mentioned sort of like AI surfacing redundant code, but the ability to identify problem to and surface code that already exists for an individual so they don't have to go and find it or reconceptualize it to has to save an enormous amount of time.

00:14:41:18 - 00:15:06:07
Scott
Or, you know, one of the problems with these naive AIS that don't understand your existing software is they're very good at proliferating code. So you start writing in a, new class. You know, they're they're very happy to keep adding ever more software into the code base augment. But still, one of the only AIS that will detect and say, hey, there's there's code in here that we could reuse.

00:15:06:09 - 00:15:15:16
Scott
Or there's redundant code in here that we could delete. Right. And that's, software engineering Nirvana is getting rid of code, not adding more code.

00:15:15:18 - 00:15:33:20
Sandesh
Yes. Scott, I got a couple questions. 2 to 2 part question. First question. What is your sales cycle look like? Where? Who are you calling on? How are people finding out about you? How do customers experience, augment?

00:15:33:22 - 00:15:57:10
Scott
Yeah. So, you know, we we really started selling the solution, you know, third quarter last year, and, it was in that point, a lot of top down conversations, you know, through existing networks and investors and so on. We would call on, VP of engineering and CTOs, that we had familiarity with.

00:15:57:12 - 00:16:16:16
Scott
Things really changed, early this year, when we launched, self-service. So, you know, in, in general, software engineers, they can't try out a product or suspicious of it right out of the gate. Right. If you don't think your product is good enough to let anyone, give it a try, they'll be very skeptical.

00:16:16:18 - 00:16:42:11
Scott
And that self-service, engine has exploded, for us. So, we just recently launched, agent, and that keep that usage of the of the product is just grown, from nothing, to you know, six figures of developers, using the using the, the technology. And we've seen our consumption of, tokens per minute.

00:16:42:11 - 00:17:03:21
Scott
This is the inference that's used to run our agent, has is grown 15 fold, just over the last month. And so you see this huge influx of, of developers, discuss what the models are now capable of and, you know, removing themselves from a lot of the less fun aspects of software engineering.

00:17:03:23 - 00:17:20:20
Sandesh
Totally. This might be a tricky question, but we'll. Why are you finding, like, the developers become your champion or are you finding more? It's like the leaders that see the efficiencies than the, you know, just the cost savings that they can drive out of it.

00:17:20:22 - 00:17:47:14
Scott
I would say, the, the top down leadership has a mandate to, to use AI to take advantage of AI. And they see this space exploding. And so they're they're curious about it. But our bond, is with the smart architects, and the lead engineers, you know, they're often individuals that have defected from other coding AI's because those coding eyes didn't keep up with them, didn't understand the software, that they were working on.

00:17:47:20 - 00:18:08:00
Scott
And so when they, they try augment and they have this epiphany like, wow, this is a fellow expert in my code base, right? I I've spent years becoming an expert. Now I've got a code programmer who is also an expert. And so their bond with the product is really what's driven this explosive growth that we're enjoying today.

00:18:08:02 - 00:18:20:18
Sandesh
Is there is there any thing that you've experienced in the last 12 months that really surprised you, that you weren't expecting, but kind of came out of the blue.

00:18:20:20 - 00:18:23:23
Chris
In I, I would imagine just about everything.

00:18:24:01 - 00:18:26:08
Sandesh
Yeah, I know, loaded question.

00:18:26:10 - 00:18:46:21
Scott
The market, moves so insanely fast. Right. So the models are getting better, you know. Well, faster than, you know, my highest expectations, you know, in terms of, you know, what agents are now capable of. I mean, we have, engineers on the team. I think sometimes they do a little more work in order to get the agent to do everything for them.

00:18:46:21 - 00:19:05:16
Scott
But it's like a point of pride that they're. All they're doing is what I call metaprogramming, right? They're not actually writing code themselves. They're using the AI to manipulate the code, and they're telling the AI what they want. And then the AI goes off and does everything. I did not envision, you know, that we would be there.

00:19:05:21 - 00:19:25:00
Scott
You know, it's certainly not at the start of, of this year. The market is moving so much faster. You know, every quarter, I feel like we're compressing a year to 18 months of progress than what I've seen in, you know, back in the dotcom days and certainly in, the flash storage days that preceded this.

00:19:25:01 - 00:20:08:03
Scott
So, I mean, it's incredibly exciting. And one other thing about, startups, you know, especially a system software startup is when you get surprised, you know, it's almost always bad news, right? There's something something broken. Something went wrong. You know, we still have those surprises, but now we have upside surprises where the AI does something that you were just completely not expecting and and, wowed by, you know, like the story I mentioned about, deleting code, or troubleshooting, you know, an issue in the, the, the build in test environment, like a Cicd pipeline issue.

00:20:08:03 - 00:20:24:05
Scott
And the AI is able to, undiagnosed what happened and even propose a fix. You know, it just like this emergent behavior where the eyes keep on being able to deliver more insight into what's going on in your environment, or is it just really staggering?

00:20:24:07 - 00:20:37:17
Chris
There's two, two big areas I want to want to sort of talk about. One is, the so you can you talk about how you integrate with Ides, with these development environments and how all that works? Yeah.

00:20:37:17 - 00:20:49:17
Scott
So there are two broad approaches. So first of all, the IDE is where engineers spend the majority of the time. Right. It's a it's a portal that they trust in. No. Into the code and into all the tooling that surrounds the code.

00:20:49:17 - 00:20:54:00
Chris
And they very much like working in their IDE. They don't like change.

00:20:54:00 - 00:21:17:16
Scott
Yes. I certainly think that's that's true. And so, from the outset, you know, we wanted to use the IDE as our portal, as our delivery tool for the, for the AI. And we then we wanted to preserve compatibility, with the existing IDE ecosystems. So, you know, vs code, from Microsoft is one of the most popular Ides they elected to open source.

00:21:17:16 - 00:21:48:09
Scott
It, and some of our competitors have taken that code base and forked it and changed it, so that it's no longer compatible with the Microsoft ecosystem. But, you know, they have the ability to to go in and change arbitrary things inside of the IDE. We didn't want to take that step. We thought the vast majority of the customers we target, you know, the enterprise scale customers value their Microsoft relationship and want to stay compatible with all of the, plug ins that work, inside of, vs code.

00:21:48:11 - 00:22:18:01
Scott
But there's another whole community that's, you know, at least as large on scale, which is, JetBrains. In fact, you know, in my career, I've spent a lot more time with JetBrains because of, Java and then C and C plus. Plus where, that toolset is much more popular. And, and so, you know, being able to plug in to the existing IDE, ease and, and be able to just act, in them with full compatibility with everything else that works in that environment.

00:22:18:03 - 00:22:23:03
Scott
Was a big selling feature, you know, especially up into, the enterprise customers that we target.

00:22:23:03 - 00:22:27:06
Chris
Yeah, that's got to be a really, important differentiator for you, I imagine.

00:22:27:08 - 00:22:49:11
Scott
Yeah, especially, you know, so many of big companies, you know, the in fact, it's very rare to find a big company that standardized on a single IDE. Right? So, different projects, different engineers have their, you know, their own toolchains that work with them. And so compatibility with, what is already there. We support VI as well.

00:22:49:13 - 00:22:56:12
Scott
So, vim, is now a solution, and I've got a buddy that's been, putting in Emacs. It's just going to.

00:22:56:12 - 00:23:01:06
Chris
Ask, you know, if you got vim, you got to have Emacs, because that debate never ends.

00:23:01:08 - 00:23:23:23
Scott
Yeah. We when we launched vim, I took some heat because I'm an old Emacs, guy. But we had a particular customer, a couple of customers in the storage, space. This is, data direct networks and, pure storage. Fun spot for all of us that, they still use a lot of, vim, in text editors in their practice.

00:23:23:23 - 00:23:31:06
Scott
And so having an AI that works in those environments, was, was much a much easier sell, to, to win those customers over.

00:23:31:06 - 00:23:36:13
Chris
I always liked vim. The Emacs key were just mind numbing to me.

00:23:36:15 - 00:23:41:19
Scott
It was whatever you get started with, right? It's like, switching. Switching. It was rough.

00:23:41:21 - 00:24:03:22
Chris
So I want to I want to pivot to another area that I think is top of mind for so many enterprises, and that's security and you have a lot of, things to offer in that regard from many different angles. You know, one is sort of the you know, you mentioned the predictable that you can bring to the development process.

00:24:03:22 - 00:24:28:12
Chris
And certainly predictability is a big component of designing security into your platform. Right. But even beyond that, I mean, you're also, dealing with a company's code base, which is the absolute crown jewels of companies, right? So you have to keep that secure too. So, I mean, you're you are a security play from many different angles.

00:24:28:14 - 00:25:03:17
Scott
Yeah. There's there's two different levels of, thinking about security. There is, you know, the, making sure that you're building in security into customer code. And part of, you know, our special secret sauce is that we are able, to digest these large code bases, and that includes assimilating their security policies and best practices and then using those security policies and best practices as we're generating new code and modifying existing code, so that we're doing so in a way that preserves all the properties, of, of that system.

00:25:03:19 - 00:25:31:15
Scott
But then there is, of course, the securing of our cloud service. The customers are making use up. And, you know, the way that we are able to deliver that knowledge of, of customer code is we do have to build a real time semantic index of their, of their code base. And we store artifacts in our cloud service, in a highly secured way, so we can bring the right artifacts to bear, in, you know, for whatever AI task is at hand.

00:25:31:17 - 00:25:57:23
Scott
So the model is able to understand the, the customer code and do the right thing. So, you know, I think we were the first of the solutions to get, SoC, compliance, both. Level one, level two, SoC two, level two, where we currently are, we've had continuous penetration testing on the platform. You know, we have engineers from places like Databricks and Snowflake, that have, you know, very strong, security credentials.

00:25:58:01 - 00:26:06:18
Scott
And so, you know, we built this architecture, to, to make sure that we were fully hardened, to be able to protect our customers intellectual property.

00:26:06:20 - 00:26:24:06
Chris
So, Scott, I mean, you know, other we've talked about some really key differentiators I don't see, I don't see a lot of competition doing what you guys do. Exactly. I mean, what do you what would you consider the big things that, you know, like you're not going to get anywhere else?

00:26:24:08 - 00:26:54:03
Scott
Yeah. So I think the, the things that the primary thing that differentiates us is this ability to understand a large, complex, code base. And then we, we do very well winning over, you know, these enterprise, scale customers. But, you know, self-service is brought down. I have six figures of developers from all over the world, that just want an AI that's more expert, that, you know, is able to hang with them in terms of whatever, it is, that they're, they're working on.

00:26:54:05 - 00:27:15:07
Scott
And I would say the competitive landscape continues to evolve. Most of the customers we encounter have some experience with AI for code. GitHub Copilot is often been in the account ahead of us. What we at least seem to find is a lot of the senior architects have, elected to move away from it because it, you know, hasn't been able to keep pace with them.

00:27:15:07 - 00:27:33:03
Scott
And so they don't feel like it's a good partner that's, creates a big opportunity for us to go in, especially our pricing is usage based. So if you don't use the product, you don't pay for it. And so it's a really easy insert, into, you know, an existing shop that's using GitHub Copilot. It's like anyone's welcome to try it.

00:27:33:03 - 00:27:52:01
Scott
And you get a bunch of chips. And if you use those chips up, we'll, we'll give you more. I would say, you know, the the landscape has changed, though, we see, you know, significantly more cursor now in a lot of our accounts, it's also a very popular solution, probably coming more from the vibe end.

00:27:52:03 - 00:28:19:04
Scott
But, you know, we've we've definitely encountered it in some larger enterprise, customers. Now, one, fun stat is, we get 1500 cursor developers per day coming to augment, where they, replace the cursor AI with augments. I, inside of the cursor ID because their ID because they're getting better, support for whatever task that they're doing, at hand.

00:28:19:06 - 00:28:42:21
Scott
But it's going to stay a ferociously, competitive market given the scale. Right. This is a, you know, $1 trillion spent per year on software engineers around the world. And, you know, so there's I think all of them are going to be using AI, and with the material productivity improvements that come with that, you know, I think we're going to continue to see a ferocious fight.

00:28:42:23 - 00:28:49:02
Sandesh
Exactly. So, Scott, where does augment go from here?

00:28:49:04 - 00:29:15:16
Scott
You know, we are still in such early days of, of this market and in terms of how much I think the AIS are going to be capable of, of accomplishing, you know, we're continuing right now to, to to hone, our agent, platform, to be able to deliver much more capability to do ever more and richer, richer tasks on behalf of, the software engineer.

00:29:15:18 - 00:29:33:11
Scott
Yeah, I would say the that that is the chief excitement for this year. You know, every engineer wants to be a tech lead, and be able to farm out the tasks that they're not excited about. In this case. And I can go off and do these things, quickly. And, you know, we'll continue to improve the results.

00:29:33:12 - 00:30:06:05
Scott
And how long, you know, the the, what kind of sophistication the, the AI is capable of in terms of, completing those long term tasks. But I look, you know, as you look forward, there's so much more opportunity. You know, if you take something like, code reviews, there's so much insight, that's captured in there, how can I bring that to bear across a large code base and say, you know, where where is this code base deviating from, you know, policies and best practices?

00:30:06:07 - 00:30:37:18
Scott
Where are the opportunities to improve quality? You know, the US economy alone lost $2.5 trillion last year due to software failures. How can we go in and, you know, troubleshoot the software before it breaks? Or, you know, when it does break because software will continue to break. It's great if the AI wakes up first, before, you know, the human, gets woken up and starts, offering the human insight into what happened and what mitigations might make sense.

00:30:37:19 - 00:31:06:06
Scott
So, you know, there's just and and engineering management is another area, you know, we've been helping individual engineer engineers, but we haven't yet, crossed, the chasm to help management. See what's going on with their code base and find additional opportunities to improve quality or, to optimize. So, you know, I, I think we're going to have just an insanely fun next five, five years.

00:31:06:08 - 00:31:08:09
Scott
Can't imagine what we're going to see.

00:31:08:11 - 00:31:33:12
Chris
You know, another thing, you mentioned that the incredible, you know, trillions of dollars of losses from software defects. But I would I want to also mention that, you know, there's been an interesting in the, like ransomware, for example, the initial vectors for ransomware was, a credential compromised for many, many, many years. But now it's been software defects that have recently bypassed that.

00:31:33:12 - 00:31:43:22
Chris
So in order to make yourself secure, these are really important kinds of steps to take in terms of the code reviews and things like that, to make sure you're building safe stuff.

00:31:43:22 - 00:32:03:12
Scott
And there's you know, you can look at code change history in order to identify common characteristics of defects that might still be in the code base, and then proactively identify them and fix them. Before they cause, cause a problem. So we're going to get much higher quality software. It's going to be a lot more predictable.

00:32:03:14 - 00:32:24:17
Scott
And it's ultimately the, the barriers to giving us the features that we want have been vastly reduced. And I think that's even going to go back to really old, old ossified code. Right? There's a ton of code that is just never changed because it's viewed as fragile because nobody's around who really understands it. That fragility is going to go away.

00:32:24:19 - 00:32:49:03
Scott
We'll be able to feel comfortable that we can understand any piece of software. You know, we've had customers tell us they stopped, they started deleting their documentation because it's better to ask, augment real time what's happening inside the code base than reading docs that are maybe stale. Yeah. Out of date or not so well-written. And so I think that's a, you know, kind of way to keep all this software fresh all the time.

00:32:49:05 - 00:32:53:14
Chris
Yeah. You can spend less time digging through documents and more time doing stuff.

00:32:53:16 - 00:32:55:00
Scott
100%.

00:32:55:01 - 00:33:05:13
Sandesh
A great, message that I'm getting from this is you're still recommending computer science for a major in college is, It sounds like you're you're pro computer science.

00:33:05:15 - 00:33:28:15
Scott
Very much so. You know, there the the human guidance for the software process is still so essential. You know, I mentioned the agents being able to to generate nearly 100% of the code now with a lot of coaching from a human and that it's that coaching that's crucial because the human spot's the direction and, leads the evolution.

00:33:28:17 - 00:33:48:22
Scott
But, you know, the AI does, you know, the busy work of casting it into, you know, the, the concrete code that you need, to run it. I think that that is going to continue to be the case for, for a good while. You know, it's possible the advent of superintelligence, sometimes called AGI, you know, changes that.

00:33:48:22 - 00:34:03:14
Scott
You know, then maybe the software can or the AI can figure out exactly what kind of software humans need and want. But that feels like a good ways off. To me, even with this extraordinary process we're seeing in the underlying models.

00:34:03:16 - 00:34:09:02
Sandesh
So if somebody wants to hook up with augment, how do they find you?

00:34:09:05 - 00:34:32:18
Scott
So, augment code.com is the website, the the product is available for anyone to trial. If you have vs code, if you have any of the JetBrains, products like Intel J, or PyCharm, you know, you're, you're more than free to grab it and vim, Emacs. Not yet, but hopefully soon. So by all means, grab the product.

00:34:32:20 - 00:34:55:15
Scott
Give it a try. You know, there we have a contact sales for larger organizations, but small organizations can do everything they need online. And, you know, we got to the six figure, of developer customers, through self-service, where we didn't even have to, to talk to them. And that's a one of the really exciting things about how fast this market moves.

00:34:55:17 - 00:35:21:06
Sandesh
No doubt. For sure. No doubt. Well, you guys heard it here. If you want to hook up with augment, you know how to find them. I also want to point out that Chris and I do a lot of research on AI companies, specifically startup startups, and we're trying to find those winners. And some of the things that we look for is not just the tech and like the market that they're trying to go after, but the people behind it is always very, very, very critical.

00:35:21:08 - 00:35:42:09
Sandesh
And that's why when I when I saw some of the crew getting together here, I was like, okay, something, something cool is happening here. Yeah. You, you you, you know, we we follow the money, but we really also follow the people. And so, you know, to to your guys's credit, it's not just the tech, but you, I can see you building a great company.

00:35:42:11 - 00:35:49:09
Sandesh
And that is really, really cool. So congratulations. And thank you so much for, joining Chris.

00:35:49:09 - 00:36:01:16
Scott
And I very much appreciate the kind words. And Sandesh and Chris, always great to hang out. Look forward to checking back in down the road and, talking about the latest developments in this topsy turvy market.

00:36:01:18 - 00:36:06:13
Chris
Sounds great. And I'm sure you're going to have yet another home run here, so good luck with everything.

00:36:06:15 - 00:36:09:03
Scott
We'll do our best.

00:36:09:05 - 00:36:09:22
Sandesh
Thanks, Scott.

00:36:09:22 - 00:36:13:02
Scott
All right, all the best, gents. Cheers.

00:36:13:04 - 00:36:37:13
Sandesh
Well, that's the show. But before you roll sales pitch warning, we're building more than a brand here. We're building a community. So your support means everything to us. So please, like, comment, subscribe, follow, and also reach out to us directly. We want to hear from you. Thank you so much for your support. Until next time.