AI Can Build It, But Can AI Ship It?

0:00If I were to have to choose one, I would say the iteration cycles have become massively compressed. We don't want to start taking for granted, which maybe are being taken for granted, things like security and resilience and reliability and availability. And again, these are some of those core fundamental underlying principles that make AWS. The way that I like to think about it is it's my little AWS CTO on my shoulder.

0:25Corey Noles:How do you even know those terms to begin with? Exactly. You're right. And that's exactly it. You know, AWS now with over 200 services, it can be a lot for even experts. And so asking the startup advisor saying, hey, here's what I'm building. Here is presumably who the customer is. Here's how they're going to access. I think they're going to access my software. Help me plan out what I should be aware of from an AWS perspective such that, you know, I am aligning myself to the best practices at AWS. Welcome, humans, to the Neuron AI Explained. I'm Corey Knowles here today with Grant Harvey. How are you today, man?

1:05I'm doing good. I'm doing good. I'm excited to have this chat.

1:08Corey Noles:So am I. That's because starting a software company once required either deep technical experience or the resources to hire someone who had that. And AI coding tools have kind of begun changing that equation. And AWS is now extending the idea beyond writing code with Startup Advisor. which is an AI-powered assistant designed to guide founders through architecture, infrastructure, security, cost management, and scaling decisions. And today, we're joined by Deep Ubi, Director of Global Head of Technology for Startups, to tell us all about AWS Startup Advisor. And before we get started, please take just a quick second to like today's video and subscribe to the channel so you never miss an interview or one of our wacky live streams because we'd love to have you there.

1:53Deep, welcome to the Neuron. It's great to have you. Yeah, it's really good to be here. And thank you for the invite. Excellent.

2:00Corey Noles:So I'm basically your target audience. I'm a non-technical founder using AI to handle the technical side for me as I've been building a bunch of coding stuff, kind of teaching myself how to code in the process. But AI has made that possible when before, you know, I would have had to raise a bunch of startup capital or, you know, hire people who knew what they were doing. And I'm trying to build tools that I wish existed, that no one else is building. So give me the pitch. How can you help me? Yeah. I mean, so thank you all for, again, the time and thank you for mentioning in your preamble the AWS Startup Advisor.

2:36That is here. That is what I am here to to explain a little bit more to your audience about, to you about Grant and to really anyone else that is just interested in, you know, taking an idea and really like exercising their creativity and feeling unfettered from, you know, some of the constraints that existed in the past. As you mentioned, you know, having a computer science background, knowing how to code and then knowing how to do that at sort of a high fidelity and high quality, thinking about things like, well, OK, if I put this out into the world, is it going to be safe for people to use?

3:07Are they going to get hacked? Are their accounts going to get hijacked? And all those kinds of things that come into consideration kind of once you're beyond the sort of proof of concept phase. And that is kind of like, you know, I would say, you know, Corey and Grant, like the, you know, the difference between sort of vibe coding something that, you know, you and friends and maybe close peers can play around with and start feeling good about and developing a thesis around. and, you know, taking something like that and then delivering it to a customer who is going to presumably pay you for that particular, you know, usage or that software at some point in time in the future.

3:44There's a difference between those, obviously. And, of course, like if you have been in and around startups, you very much understand that there's a journey that you go through. Like you're going to show your close friends and your peers and then you've got to take the steps to say, hey, you know, I think I've got something here that I'm excited about, that I'm getting good signal from my prospective customers about, and it's time to take this to the next level. That's where the startup advisor, I think, really comes into play here, where you could use a startup advisor from day one. You could use it to help you do a lot of that vibe coding in whatever vibe coding harness of your choice.

4:17If you like to use Cursor or Codex or Claude Code, the AWS startup advisor is available in every sort of coding harness that's available today. and by sort of, you know, the way that I like to think about it is like, it's my little AWS like CTO on my shoulder, right?

4:35Corey Noles:So I can hop on this little guy or gal and say, hey, look, you know, I really need help with authentication now, right? I need my users to be able to authenticate into the service I'm building and I'd like for them to basically authenticate with their Google Workspace, you know, aliases. So the startup advisor is really, really good at saying, okay, well, you know, here's what you want to consider when you're doing something like that. Here's how to do that on AWS, by the way, if that's something that you want to do. And, you know, we also have a plethora of third-party technologies and partners that are big AWS customers themselves.

5:07The AWS Startup Advisor will also make recommendations across, you know, and is fluent in those technologies as well. So you get like a really good, grounded, you know, high-fidelity production-grade guidance from, you know, a tool that is effectively codified like all of the resident knowledge that, you know, internal experts at AWS have on building software for the broader, you know, sort of, you know, enterprise ecosystem, if that makes sense. Yeah. Yeah. Yeah. Well, you've been a founder yourself several times along your journey. When you compare starting a company today with starting one before generative AI, what do you think is the biggest difference?

5:53If I were to have to choose one, I would say, Corey, it is the iteration cycles have become massively compressed. Right. I think, again, you have to tip your hat and give all the credit to the coding agents. You know, Grant mentioned that that's something that he's been really like toying around with and playing with and feeling empowered by. But because you're able to take an idea to POC and that POC into the hands of a prospective customer, get that like rapid feedback, incorporate that feedback into the next iteration of your software. Like those cycles now are, you know, hours and days, maybe not months and quarters and years anymore.

6:38That I would say is the biggest difference. And so what does that allow you to do? Well, startups are all about speed and what we call sort of time to value. So like how quickly can I deliver a wow experience to my customer? Like that's a big thing in the startup world, right? Like what they call time to wow. You now can have many more at-bats, right? Before you had three or four at-bats. You're always thinking about your runway. You're always thinking about, well, is this – how many more months do I have where I can do this? Where you could swing twice a week now. Exactly. Yeah, exactly.

7:09Corey Noles:So I've had sessions where I'll be playing a game with my friends that I created and they'll be giving me real-time feedback on it. And I'll literally feed it to Claude while I'm working on it. And then it will update and the game will crash for a second and then it'll update. And it's like, oh, there it is. Yeah. Yeah. It's wild. And that is – I mean feedback in any form, whether it's professional or personal, it's such a gift. And so to be able to take feedback and to do something about it, it makes it more actionable. So that's the biggest difference, Corey, that I see. If I were to start a company today, I would likely do it with fewer people, and I'd be very confident I have many more at-bats for the same amount of capital that I had available to me in a previous iteration.

7:53Most people are only using about 5 % of what AI tools can actually do. They open chat GPT, ask a few questions, maybe write an email, and that's it. Meanwhile, other people are out there building workflows, automations, research systems, custom agents, and saving hours every single week.

8:11Corey Noles:AI is moving so fast right now that most people don't even know which tools are worth learning anymore. ChatGPT, Plaid, Gemini, Codex, AI agents, automations, FIVE coding, it's a lot. And that's exactly why we built the Neuron Academy. We built a practical, self-paced AI learning platform that helps working professionals like you understand the tools that actually matter and how to use them in real work. Built by us, the team behind Benarone newsletter and the podcast, designed to help busy professionals actually level up their AI skills. Inside, you'll learn prompting, AI workflows, research systems, automations, productivity strategies, and practical use cases across the entire AI ecosystem.

8:52Everything is hands-on, beginner-friendly, and built around real-world use cases you can apply immediately.

8:59Corey Noles:These are real skills that help you work faster, think better, and stay ahead in AI. Right now, you can join the Neuron Academy for just$499 a year, unlocking 50-plus lessons from full AI courses to quick-hit micro-trainings you can apply immediately. So if you're serious about leveling up your AI skills this year, this is where you need to start. Join the Neuron Academy today. You can literally now go from, you can have an idea, dummy out a quick plan, have it in a prototype that's hosted, and be getting that feedback you're talking about, you know, before bed. It's the craziest thing. Exactly.

9:37That's exactly right. And I think that, you know, this is kind of where I think there will continue to be ongoing debates and like healthy dialogues around this, you know, well, what does that mean, you know, for, you know, the shape of workforces in the future? And, you know, like our engineering team is going to shrink and are we going to, you know, And again, like I don't have a crystal ball, but I can tell you is that, you know, there are two things. One is, and I'm sure you guys have had maybe conversations around this on the pod or maybe other folks have mentioned this to you, but there's a sort of like this like new skill or capability that's being highly valued and touted in the Valley, which we're calling taste.

10:19right which i think is a new moniker for like hey do you have a sense of like how to create a really really good end user or developer experience and that's again now moving away from the world of like purely thinking about it in the frame of a gui to like can i have like a really good conversational experience with an agent right and is it going to say the right things you can have the right tone of voice is it going to have the is it going to understand sort of like you know the the right terms in the conversation like all those things are like now like new experiences that you You have to have good taste around.

10:50I think that's still going to very much be a human endeavor. And I think the second thing is like, you know, Corey, you were mentioning how, you know, in a day before you hit the sack, you could get multiple at-bats and basically try multiple iterations. Yes, and I think that there's going to come a point where, you know, you're going to feel really, really good about the like ongoing commercial viability of what you're building. You're like, okay, wait a minute. I need to turn this into a business now. Like this is actually becoming a thing. that's when you really need to like be very very thoughtful and intentional about maybe some of the things that you know we don't want to start taking for granted which maybe are being taken for things like security and resilience and reliability and availability and and again these are some of those like core fundamental like underlying principles that make aws aws right yeah startup advisor for that reason, just for that reason, like, because it's got like all of that as like requirements, right?

11:51Like I'm going to help Grant and Corey build what they're going to build. And at the right time, I'm also going to make sure that those underlying key, like foundational principles that make software durable and usable and commercially viable, like it is an expert at that. Right. And those things that, again, like vibe coding platforms are amazing and empowering and enabling. And at the same time, like you want to make sure at the right time that you have that, you know, the guidance from what we call like the neck beards. Like we used to call it the SREs that are in the basement. Like they know where all the bodies are buried.

12:25They know precisely how to make sure that these applications stay up, that they are, you know, secure. They're not going to get, you know, so anyways, like that's, that's a very important aspect of like going from.

12:34Corey Noles:And that's the stuff that you don't know. Like you might have your own product taste or your own opinions about like the actual user experience, maybe even the agent experience. But as a, you know, agentic engineer who's just doing this for the first time, a vibe coder, you know, you don't necessarily have the priors to know that information. Totally. Yeah. Well, you know, and what's nice now is that it sort of flips the paradigm a little because previously what you would have done is you'd have had an idea. You'd have gotten with some people. You'd have talked through it with people who knew things you didn't know.

13:05And then you'd have been like, we need to hire a developer. You go and you find a developer. You find an engineer. And then you get on their queue and you give them a lot of money and you wait. Only then to find out your product sucks. But now what you can do is have an idea, dummy it out, and find out if anyone likes your product and then go find an engineer. And then go find a security person and invest after you have an idea of whether people are into this. This is a thing. 100%, Corey. Absolutely right. Because I think anyone who is thinking about building a business and thinking about building software, I'm sure we can rally the support of a couple of dozen friends and acquaintances that kind of mimic the ICP of the customer base you're going after.

13:52Right. Like that, that is probably not the constraint and, and that not being the constraint to your point, Corey, like hugely enables this early iterative cycle. Like it's just, I mean, I, I can't imagine a better time to be starting up than right now. Yeah. I mean, I think the things that obviously you have to think of are the things that we don't talk about enough, which are, you know, well, you know, can you float your rent or your bills or whatever for a lot of time? And like, you know, are you going to put your family through undue strife? Those are things that I think are really important.

14:26We don't talk about those enough. I wish we did. But if you are in a place where you can take a risk for some finite period of time, boy, you're really going to have a lot of opportunity and attempts, many, many attempts at getting something right and pivoting. I think there's a quote. I think they attributed to Louis Pasteur, but they say that fortune favors the prepared. And it basically means that you have to put yourself in a position to get lucky. Like talk about how crazy it is right now. We're like how many times can you put yourself in a position to win the lottery? You know, there's like dozens and dozens of times a month where you're just like trying new things.

15:05And it's just – it can be a bit addictive too, you know? It is. We were just joking. Grant between us, how many things have we built this month?

15:14Corey Noles:Oh, well, I mean I've been working on stuff for a little bit longer than one month only. But I mean, between the two of us, like it's we we have 15 projects that we are developing at various stages. But I was going to say to your earlier point, like the iteration timeline has condensed. But also I feel like the urgency of focus has never been more important because it is so easy to get sidetracked on side quests. And and like, oh, we can develop this. We can develop that feature. Like that's the new challenge, I think, is like, how do you stay focused? And right now is a hot time too because right now not everyone realizes they can do this yet.

15:55Like a couple of years from now, everyone you work with is going to be doing things like this. And it's not everybody.

16:02Corey Noles:Maybe not starting a business but definitely making their own software to solve their own problems. Yeah, 100%. Totally. Yeah, and Grant, I think you hit on a really, really important point there, which again is like the very necessary human input to this like creative process, which is now an exercise in whittling things down. And like, how do you eliminate some of the ideas, right? Like you've got all this stuff. You guys have 15 projects that you're working on. Like which ones are worth your like extremely valuable time going forward, you know? And not being in a place where you are putting out a bunch of above average things, but like how do you get to a place where you're maybe one or two products and projects or software that, you know, what we call here, like, you know, you might have heard the phrase like that have a right to win, you know, like, yeah, you are uniquely positioned to bring to market, you are uniquely positioned to evangelize, you know, to your customer base, right?

16:59Like that then becomes a very human judgment, intentionality input into this process. And that's, I'm so glad you brought

17:06Corey Noles:that up. Totally. No, totally. I do have one, one question, which is, okay, so imagine me, I'm using Cloud Code. I'm building, like we said, between Corey and I, we're both building 15 things. How does Startup Advisor fit in? Do I go to it before I start coding a session? Do I go to it, hey, I have this code base on GitHub or whatever. Can you take a look at it? Or right when I'm ready to deploy on AWS or wherever else, when do I actually bring Startup Advisor into the process? Yeah. Yes. Again, we'll go through that hypothetical example, Corey and Grant. So you're using Cloud Code as your coding harness.

17:43Awesome. So do I. It's amazing. And, you know, you can go to the AWS Startup Advisor, you know, web page on AWS. I'm happy to share that or maybe put that in the pod notes at some point. Yeah, we'll do that. Yeah. And again, you're going to be redirected to, you know, an Anthropic page where you can download the AWS Startup Advisor agent plug-in. So an agent plug-in in Cloud Code Parlance is just like a collection of underlying skills, right? And the skills are in sort of these acute capabilities that the startup advisor has, right, across a number of different sort of vectors. Security, authentication, observability, migration, if you're migrating from like an existing thing that you've built it in and you want to come to AWS and so on and so forth.

18:34So it's got a number of these different skills. And again, the skills are all, again, sort of codified from expert humans at AWS that are in the field with customers and have done that precise work and have understood the best practices and have basically built that into the AWS Startup Advisor agent plugin. Awesome. So once you're in Cloud Code and you've installed the agent plugin, as I'm sure you're aware, like you can invoke that plugin at any given point in time, right? Like within like just the natural language prompt, you'll be like typing stuff in and you might say, hey, startup advisor, like within the context of this current thing I'm trying to do, you know, can you tell me if, you know, there are any, you know, security implications that I need to be aware of from an AWS perspective?

19:21And so invoking it is just like invoking any other agent plugin or, you know, skill within. Even a skill, yeah. Whatever coding harness you're using. So it's very much like just, you know, invoke it when you need it. But it's not something that you need to be in constant contact with, like in the sort of course of like your coding session, right, Grant? So like you'll be sitting there writing and you're doing stuff and you're talking in natural language. cloud code is outputting either code or terraform or whatever it is you know application code or infrastructure as code um and then within any of those like sort of conversations that you're having with cloud you can just invoke the startup advisor to give you that sort of like aws specific context yeah and and that's how that's how i use it that's how you know most of the folks that i've downloaded use it um yeah so it's really able to have it help you like set up like if you want to

20:13Corey Noles:do like S3 buckets or stuff like that that's AWA specific. A hundred percent. Yeah. Yeah, absolutely. And again, like I think where it's more useful, yes, for those acute things as well, like, hey, can you please help set up, you know, this S3 bucket with this IAM policy? I think where it's more useful though is even left of that, like helping you plan, right? So like think about the planning phase, which is like, hey, this is - How do you even know those terms to begin with, right? Exactly. Yeah, exactly right. And that's exactly it. You know, AWS now with over 200 services, it can be a lot for even experts.

20:49And so asking the startup advisor saying, hey, here's what I'm building. Here is presumably who the customer is. Here's how they're going to access. I think they're going to access my software. You know, help me plan out what I should be aware of from an AWS perspective such that, you know, I am aligning myself to the best practices at AWS. And so it'll do the planning for you and then really up to your, you know, how you work and how you kind of like vibe code. Like you can do it in one shot. You could like do it over many days. You could have it like execute, you know, autonomously. So there are just different ways and like that's very much a personal preference.

21:29But, yeah, I mean, I actually think that it's very helpful in the planning phase as well. At its core, we're talking about having your own little AWS consultant and engineer teched right into your harness, which is great. And it works with harnesses other than Cloud Code, correct? It does. It works with Cursor. It works with Codex, VS Code plugin. It works with Kiro as well, which is the AWS. The AWS one, yeah. Yeah, coding harness. With a great logo, by the way. I just want to high-five you on the branding of Zero. The branding of Zero is, you know, it's another 30-minute pod. I think they did an amazing job with the branding of Zero.