Everyday AI Podcast – An AI and ChatGPT Podcast - EP 462: NVIDIA Inception Startup Spotlight: Turning ideas into products

Episode Date: February 14, 2025

How can you quickly turn ideas into products? Do you need to be overly technical to build in today's day and age of AI? And what the heck is in the NVIDIA Inception program? We'll tackle all those questions, and a lot more. Newsletter: Sign up for our free daily newsletterMore on this Episode: Episode PageJoin the discussion: Ask Jordan and Will and William questions on AIUpcoming Episodes: Check out the upcoming Everyday AI Livestream lineupWebsite: YourEverydayAI.comEmail The Show: [e-mail] with Jordan on LinkedInTopics Covered in This Episode:1. NVIDIA Inception Startup Program2. Spotlight on Lightning AI3. Rapid Changes in AI and Their Impact4. Strategies for Startups in the AI EraTimestamps:01:50 Daily AI news04:40 Intro to guests06:11 NVIDIA Startup Support Overview08:24 "AI Deployment Simplified with Prebuilt Solutions"12:40 AI Era Spurs New Startups13:43 Future of CRM and Ecommerce Revolution18:43 "Lean Methodology's Impact on Startups"20:10 Defining Core IP for Startups23:59 "Partner for AI Business Growth"27:50 Automating Expertise Documentation30:46 Risks of Rapid Product Creation34:37 Experiment with AI ToolsKeywords:Generative AI, startup landscape, software engineer, leverage AI, NVIDIA inception program, AI news, Google DeepMind, YouTube Shorts, Anthropic AI model, OpenAI CEO Sam Altman, generative AI tools, AI video model, podcast, Lightning AI, NVIDIA startup program, AI native companies, product market fit, large language models, OpenAI, ChatGPT, machine learning, deep reasoning capabilities, AI newsletter, hybrid AI model, technology invention, Lightning AI hub, developer tools, AI prompts, GPT series, computing efficiencySend Everyday AI and Jordan a text message. (We can't reply back unless you leave contact info) Start Here ▶️Not sure where to start when it comes to AI? Start with our Start Here Series. You can listen to the first drop -- Episode 691 -- or get free access to our Inner Cricle community and all episodes: StartHereSeries.com Also, here's a link to the entire series on a Spotify playlist. 

This is the Everyday AI Show, the Everyday Podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. Meet Firefly AI Assistant, now live in Adobe Firefly, the All In One Creative AI Studio. Just describe what you want to create and the assistant handles the rest, orchestrating multi-step workflows across Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome. The assistant accelerates execution. Generative AI has obviously changed how companies work,

but I think it's also quickly changing the startup landscape. You know, 10 years ago, I think when you think of a startup founder, you thought as someone maybe with a hoodie, you know, just coding like, you know, on their laptop for hours and hours. But it's not like that anymore. You don't necessarily have to be a software engineer or, have a background in machine learning to leverage AI to grow a startup or to turn an idea into a product. And that's what we're going to be talking about today on everyday AI.

What's going on, y'all? My name's Jordan Wilson and I'm the host. And this thing, it's for you. This is your daily live stream podcast and free daily newsletter, helping us all learn and leverage generative AI to grow your company and your career. So maybe it's your first time listening. Thank you for joining. If that sounds like you, what I just said, leveraging generative AI to grow your company and career, number one, you're in your right place.

Number two, the next best place for you to be is our website, your everyday AI.com. Go there, sign up for our free daily newsletter. Not only are we going to be recapping today's shows and all the best insights from our guests, but you can go listen for free like 450 other episodes where I've interviewed some of the world's brightest minds in AI all for free. Whatever you need, it's there. All right. So I'm excited today to talk about this concept of how we can just turn ideas into products and also talking a little bit about the Nvidia Inception program. But before we do, let's start off as we normally do by going over the AI news. So Google has rolled out its leading V-O-2 AI video model to YouTube Shorts. So YouTube Shorts has upgraded its dream screen feature by integrating Google DeepMind's latest video generation model V-O2. So this enhancement allows content creators to generate video clips using text prompts,

eliminating the need for a standalone text-to-video model, such as, you know, SORA, runway, something like that. So Google has improved the dream screen performance, ensuring a faster and more seamless user experience. So the feature allows users to specify styles, lenses, or cinematic effects, enhancing creative possibilities. these. So this isn't the full V-O-2, but for a lot of people, they haven't really been able to access this new video, this new AI video tool. So there is a new way to do that now. Next, Anthropics'

newest AI model hasn't been released yet, but it is set to be a different kind of model. So Anthropic is preparing to launch a new advanced AI model in the coming weeks, which is expected to offer developers greater control. So the new model is to start. as a hybrid model that can switch between fast responses and deep reasoning, potentially being on the same level as maybe OpenAIs O3 Mini in their GPT40 at the same time with one model. We'll have to see. A notable feature, and this is according to reporting from the information, a notable feature is the introduction of a sliding scale,

allowing developers to control costs by adjusting the amount of reasoning power allocated to specific tests. So Anthropic has been quietly working on this model, which aims to provide a balance between standard language model functions and deep reasoning capabilities. This news comes just hours after OpenAI CEO Sam Altman talked about a similar hybrid approach for the future development of chat GPT as he posted on Twitter that the O series and the GPT series would essentially be merging in the future. Our last piece of AI news, speaking of OpenAI, they have just released a prompting best practice. guide for their O series model. So nothing groundbreaking here, but we'll be sharing it in our newsletter. It includes tips like keeping prompts simple and direct, avoiding chain of thought prompts, since that's essentially what the O model does, using delimitators for clarity and to start with zero shot, then move to few shot if needed. All right, so we're going to have a lot more on those

stories and everything else that you need to be the smartest person in AI at your company. We're going to be recapping it in our newsletter. So make sure you go sign up for that at your everyday AI.com. All right, enough chit-chat, y'all. I'm excited to talk not just about the Nvidia Inception startup program, but also highlighting one of their company's Lightning AI and also what Lightning AI can

help anyone do and turning to the broader picture of today's conversation. So without further ado, please help me. We got three people. We got three people, y'all. So let's see if we can do this here. So please help me. Welcome to the show.

There we go. We have Will Co. who is the director of the NVIDIA startup program and William Falcon, the CEO and founder of Lightning AI. Will and William, thank you for joining the show. Pleasure.

Hey, George. Thank you for having us. Hopefully the names don't get confusing, you know. Yeah, yeah. Well, yeah, for our live stream audience, probably a little easier for our podcast audience. Yeah, we'll say Will at NVIDIA and William at Lightning AI. But let's start with you, Will.

Can you tell us a little bit for those that aren't aware? What is the Nvidia Inception startup? program. Yeah, thanks, Jordan. So Inception is a scaled program for every startup out there in the global startup ecosystem. So we're looking for anybody from two developers in a garage all the way up to amazing companies like Lightning and the kind of super unicorns. And a big part of what Inception does is say, let's take you on a journey from, you know,

towards your AI success. And we'll talk, I think, more about that today. But really, that starts with how do we educate you on AI on specifically what Nvidia does, how can you take advantage of it? Obviously, this landscape is moving so quickly that probably the biggest barrier is just trying to keep up. So a big part of what we're here to do is connect folks into all of the learning resources and all the myriad SDKs and evolving software stack that's coming out of Nvidia.

And then beyond that, as companies grows, start to find product market fits, start to get more investment, and start to adopt Nvidia technology. The big part of what we do is how can we provide more visibility? How can we get your folks on stage? How can we introduce you to customers? How can we introduce you to additional investors? So really, we are here kind of for each stage of that. And then obviously, NVIDIA is one big family in terms of supporting all of our customers,

all of our partners. So the other part of what we do is make sure that you're connected into the right resources around NVIDIA. So anyway, without making it an infomercial, NVDI.com slash startups is the right place to go for that. But happy here to have William from Lightning, who's been in our program for a while. Lightning has grown. They are kind of out of the nest and flying and doing amazing things. So happy to tell their story today too. Yeah, so William, let's just throw it straight over to you.

Tell us a little bit about Lightning AI. I think a lot of people, if you are a developer, you probably know Pytorch Lightning, but tell us a little bit about your story and what you all do. Yeah, for sure. First of all, the session program is great. So fully, fully support what Will just sent. So Lightning is a company behind Pite Torch Lightning. That's kind of what we're best known for.

And Pitech Lightning is a deep learning framework that lets you train models. at scale. So I'll give some examples of companies that use it and how they've done it. Stability AI, for example, stable diffusion, all those models are trained using Pytish Lightning. LinkedIn recently published a paper on a hundred and seventy billion parameter LLM that they train with PITU's lighting as well. And then, you know, probably the first time we started our partnership with NVIDIA back in like 2019, 2020 was the software site called NEMO today that's grown a lot. That's also built on total PITER Shlighting. So a lot of those models train using that. So it's been great.

And so as a company, what we have grown into now is really an end-to-end ML platform. And for most companies today, they're spending a lot of time going through like 20 steps to get a model or AI product into production, right? You might have to like clean data, curate, pre-trained models, fine-tune, deploy, monitoring, this and that. It's like 17 steps you have to do. Most companies don't have ML engineers on staff that can actually do those steps. And so they're kind of stuck usually in like a notebook or something like that. So ultimately what we provide as a company is like the final step in that process. And those are through our AI hub.

So if you go to Lightning.A.4 slash AI pub, you'll find those already pre-built. And they're ready to deploy with that one click. It's an API that you get and you don't have to be an AI expert to do that. You're going to get things like, I don't know, R1, for example, 670 fully, fully without the stealing, that's going to get deployed or things like fine tuning and deploying. I gave a talk at a bank a few days ago and we fine-tune a fine-tune a few days ago and we fine-tune deployed an R1 model, $8 billion for about $12. And they were shocked that they could do it for $12, right?

And I was like, it takes like an hour and a half. It's not that complicated. It was like no code. You click, upload, and it's not right. And that's pretty much for most companies. And then there's some companies that have ML people. The rest of the platform lets those people build those APIs ultimately. So all those other steps from one to 10, you can do them on the platform as well.

Yeah, William, can you talk a little bit just, you know, speaking of lightning. I mean, I think it's been lightning fast, just the landscape. right, because I remember even some of the earlier episodes of everyday AI, to do what you just talked about there would have been an extremely large endeavor, right? You would have had to have a room full of engineers. You would have had like to have a huge six, seven figure budget. But now you can get these, you know, instances up almost instantly. How quickly has this whole AI space been changing? And, you know, what are you all doing to help keep pace? with the development.

You know, ultimately the, I would say the first, first, first principle of how we built everything from scratch, even Pytosh Lightning, was with change in mind. Like when I was a Facebook guy doing research in 2019, and even then you could see that things were changing very quickly. This is pre-chat GPT. And so as a first principle, we said, look, ultimately we need to develop a platform that can, in a dying, pivot and adopt a new technology or develop something neat for it. And that's what we did.

And it took us a while. It took us about four years to actually get there. But as a result, I'll give you like a real example. So R1 came out like Wednesday or something like that, right? By Thursday, we had it. We had the whole thing already built API so you could deploy. And by Friday, we had a Fortune 100 pharma company with 67db in production, like not a prototype, not like a POC, like an actual production internally within like 48 hours, right? And for them, the value is like, as long as you have the,

lightning platform, you're going to be able to get the latest things very, very quickly, right? So that's part of it. And we also have a massive community for over 3 million developers that work with us to build a lot of these things as well. And hey, just as a quick reminder to our live stream audience, if you do have any questions, get them in now for Will or William. So, Will, I want to throw it over to you because, you know, I'm sure you talk to, you know, dozens or hundreds of startups on an ongoing basis, right? I want to ask a similar question to you that I just asked to William, how has generative AI large language models, how has it changed even the startup journey? Because, you know, like I said, to begin the show, I think a lot of times

there's probably great ideas or great products that just never made it to market maybe five or 10 years ago because they didn't have the right software engineering, developer, coders, etc. How has this change from your point of view? Yeah, I think we see two patterns. on that. One is the same thing that enterprises are tackling, which is how can a startup be more effective? There's a lot of buzz out there about, you know, the one-person unicorn. Like, what will it take in order to be so efficient with tools like Lightning or with, you know, LLMs as coding assistance, which of course is a space that's also evolving very, very rapidly? It was a great story this week, Jordan. I'm sure you saw about someone who built Wikitok, which was an infinitely

scrollable Wikipedia page as a way to kill time and learn something. And they built it all in like an hour with Claude as their back end for code generation. So I think we're seeing these incredible stories come out about efficiency. And of course, that transcends just the kind of gimmicky stuff all the way into true developer productivity of large teams. The whole part of the software development lifecycle we see being transformed by kind of AI, even as a human assist and then maybe much more agentic AI. and where that starts to take over big parts,

like doing all of your security assessments in an ongoing way, or managing your DevOps and deployment, or fixing your routing tables automatically and monitoring them in your IT infrastructure. So I think all of that falls into the category of how to startups improve their own efficiency. The really interesting part, though, is the new startups that can come into being in the AI era that could never be there before.

So these are AI-native companies. So I think this is going to be the year that we see true AI-native companies, is really popping up because we've just reached that level of abstraction and tooling, things like Lightning, all the LLMs, the coding assistance to get there. And what we're going to see, in my opinion, there are probably two groups there as well. You know, what is businesses whose user experience could never have existed before LLMs and AI? So, hey, it's a whole new user experience.

So what is the CRM of the future, for example? What is the e-commerce platform of the future that will look back and say it couldn't exist just the way we look back now and say, you know, things like Uber could not have existed without a mobile phone. There's a whole revolution of that, or HubSpot could never have existed without the cloud. And then the other side of that are new unit economics, new business models that are going to exist out of this. As we said, even the talk about the price coming down, how much you can do with how little in the AI era, we continue to see those prices plummet. And I think that's going to be incredible for adoption because it's going to allow us to create

those user experience deep personalization of products in the way that we've never seen before. So I think those are two categories, both efficiency, but also just whole new things that we have to look forward to. So, well, I do want to follow up on that. I think the Wiki talk is a great example, right? We'll share about that in the newsletter, but essentially I think this person built it in like three or four hours, right? Because I think the whole startup used to be like, oh, validate your idea in a couple of weeks.

And then it's like, oh, you know, a weekend. Now it's like a couple of hours or in real time, right? You have people building clones of actual large-scale software like on a live stream. So with this, the way I see it is we're going to be flooded, maybe in a good or a bad way, with even more and more products because it's easier and easier. So how can startups actually differentiate themselves when it is, it can be so easy now with AI? Well, I mean, this is, we're talking idea to product here, I guess. And the gotcha on that is always product can mean different.

things. So I think right now we're doing a lot of ideas to features, ideas to products are beginning to happen. I think ideas to businesses is the place that startups need to keep thinking about. At the end of the day, there still needs to be a value creation for your customers. So the idea that you can build faster, prototype faster, have novel user experience, still doesn't obviate the need to figure out how you're delivering value into the ecosystem. So one of those differentiators there is going to be, unfortunately, the same old things that we know of about startups, about understanding your market, about having good product management,

about remaining agile, being able to pivot. All of those things, I think, are gonna remain true in the AI era with a whole new set of tools. But there's certainly a trap that says, well, just because I can build it, it will be useful. I think we'll see a lot of stuff fall by the wayside or a lot of consolidation in these markets. So I would say, you know, never forget,

it's about the customers, it's about product market fit, not just, there's sort of exciting products that are emerging. So, William, I want to follow up with you on. this one because, you know, you gave an example of, you know, hey, new model comes out on a Wednesday, you know, within hours, you know, you can have a Fortune 100 company up in running on it. But I always like learning from companies and the way that they eat their own dog food, right? So, you know, I'm curious, right? So you're helping clients do this.

But how is this, you know, the speed of generative AI in large language models? How is this even helping you, you know, grow your current business or ideate on new products? products or service lines for your current clients? Yeah, so, you know, we ultimately develop a lot of developer tools, right, on the non-AI hub side. We use AI internally for things like fraud detection because we run a very expensive GPUs on different clouds, so we don't want bad activity on that. These are APIs that are deployed and they're like monitoring workloads all the time, and those are powered by LLMs in production, right?

right and um you know it's funny because like one of those was built by you know kind of like not really i mean kind of mid-level engineer i guess um in about i don't know five or six hours right like a like a what do you call it a a production system monitoring for you know bad bad activity right like fraud detection and it's the kind of things where like even in our company we'll have very junior engineers build very serious production things that like in a different time could have been a standalone company just to do that one thing. And for, for, you know, for us, it's a few hours of work to do this kind of thing. So it has changed a lot.

We also use a lot of AI in customer service. So support or, you know, people on our open source repos when they're asking us questions. We have Discord communities. We have communities on Reddit as well. And when people are asking things, like these things help them out as well. And in documentation, right? So because we have a lot of tools. So how can you find the most relevant things?

And even for customers, we have ways of having these things help them write code for Pytorch, Pytorch, Lightning, et cetera, et cetera, on the platform as well. And I think it's been hugely productive because now, you know, I'll give an example. We're only 40 people at the company. Lightning is only 40 people, including all of sales, engineering and everything. And I think in a different world, we probably would have needed like 300 people to do what we're doing today, right? Yeah, that's, that's, it is wild to hear, you know, sometimes the, you know, the, you know, the successful, startups and you just think or assume, yeah, they have to have, you know, hundreds of employees. And, you know, I think it really changes the lean methodology, right? Like what you can accomplish with

so few people, you know, Will, I do have a question, right? So I'm sure you get pitched all the time. And there's probably a lot of, you know, startup founders listening to this. And, you know, even businesses that are taking on this more startup mentality of, you know, finding new verticals to serve with different products and services. But even now, because it is much easier to at least turn an idea into a product, doesn't mean it's a good product, doesn't mean it can be successful. But what are the differentiators in today's day and age of AI and LLMs everywhere, what are the differentiators that can take it beyond a product and make it a successful product? Yeah, I think you touch on a few things there. Certainly it's all the old things that we know

in terms of making sure you're answering the questions about why this team, why, why this time, why this customer base. I think a lot of those things are still there. What we are seeing is that startups are getting farther with less money, certainly at the application tier of this. And I think investors are looking for more fully fleshed out products. So the demo is becoming more important than the pitch deck.

And so I think that's a big part of it. The other element is making sure that you know what your core IP is. And there, I think it's always a trick when the landscape is moving quickly to say, what is it that I'm building versus buying? Sort of that build versus buy question comes up. So a lot of what we do at Nvidia is remind startups of all of the frameworks that we have and all of the tools that are out there and things like Lightning and all these platforms so that you're not reinventing the things that are not your differentiator.

So what are the pillars that are the structural integrity for your house, so to speak, for your business? What are the things that if they go away, you might as well not fund your business at all, versus the things that are moving rapidly, like the latest models or the APIs that anyone can go and call? If you're basically just gluing those pieces together, I think it's a question about where your core IP is. So what I would say is stay very plugged to the ecosystem. Will you mention the community stuff? Yes, we also have Discord servers come to the NVIDIA Register developer program to see what the latest is coming out of there and all of our events and GTC coming up as NVIDIA's big events.

So staying plugged into the ecosystem matters. to the degree that you can find shortcuts, that allows you to more focus on your core. But I think there are going to be a lot of startups who lose a little bit of their way on that core. And investors are going to want to see what that is. What's the IP? What's the customer knowledge?

What are the network effects maybe that you're bringing to the table? Because those are the things that are going to create durability for the new startups. I think that's some great advice there. So William, I'll throw this one over to you, a question from our live stream audience. Sandra asking, as a technology advisory, Can we leverage Lightning AI to create tech assessment tools with minimal coding expertise?

Yes, 100%. So that's a lot of what the AI Hub is, right? So I mean, they're kind of publicly there right now. So if you go to Lightning.A.4.A.4-A-HUB Hub, you'll see them there. Those are no-code solutions that have been pre-built, and you can kind of like click and deploy things. And if you need specialized once built, then that's where we can help as well. So if you're like, hey, I want an agent that does X or an expert that does whatever, or like a way to fine-tune this particular model,

then we'll build that for you as well, right? And we'll put it in the AI hub. But I definitely want to pick it back on this concept as well that you guys just chatted about on like, how do you differentiate and how do you build these products quickly? Like I think, you know, to the audience here, I think this should be very encouraging because I think the next generation of great products

are not going to be built by developers, actually. It's going to be built by subject matter experts like lawyers or doctors who know their fields really, really well. And then they're going to use no-code tools or low-code tools like Lightning and Open AI and so on to basically build a lot of these products without meeting to hire tons of engineers, right? I think that's really where it's going to come because they see the value engineers tend to want to build tools, usually. And that's not really, I think at this stage, it's not really differentiated anymore, like Will said, right? Yeah, it's, it's funny, William, you say that

because, you know, I always think like if you're looking at a startup pitch deck, you know, And then there's the page with like, you know, the advisors or those people that, you know, are there to provide guidance. I'm like, it's going to be these people probably that are building the next generation of tools, right? But, but Will, I do have a question. So I think it was almost like a year ago where, you know, your CEO, Jensen Wong, you know, came out and, you know, essentially said, hey, you know, I don't think kids necessarily need to learn to code. And, you know, at the time, I think a lot of people were taken aback by that comment, right? Like, I wasn't. But here we are fast forward a year later.

And maybe it makes a little bit more sense, right? Now that you have this ability to essentially you can talk with your voice and build something, right? What's your take on this, you know, this thing, like vibe coding, right? Where you just kind of go and you talk to an LLM and you can build your right, like this wiki talking sample. What's your thoughts on this? And is this kind of a viable way forward? Adobe just introduced an entirely new way to create, bringing the power, and precision of its creative suite into one conversational experience.

Meet Firefly AI Assistant, now live in the Adobe Firefly app, the all-in-one creative AI studio. Powered by Adobe's creative agent, Firefly AI Assistant lets you start with your vision, just describe what you want, and shape the outcome as it takes form with the Assistant. The Assistant orchestrates multi-step workflows, drawing on 60-plus pro-grade tools across Adobe Creative Cloud apps, including Photoshop, Illustrator, Premiere, Lightroom Express, and more to help bring your ideas to life. You can also get started with creative skills, a growing library of pre-built workflows for common creative tasks, like batch editing photos, creating mood boards, portrait retouching, and creating social variations. Every step the assistant takes is visible so you can refine, redirect, or take over at any time.

You stay in the driver's seat as the creative director. Adobe Firefly AI Assistant now in public beta. See it today at firefly.adobie.com. Look, I think it's certainly transformational to development, everything that we're seeing around the code gen space. And I think the same thing's been true for generations, which is maybe kids don't need to learn to code in the same way as the last generation.

They still need to learn to think. And what this whole revolution is doing is kind of separating out some of the same. of the mechanics of what we do from the creativity. So the idea to kind of curate your ideas and your focus is still going to be top of mind. And I think it's going to accelerate, I think, the creativity of kids. I mean, this doesn't just touch coding, by the way, it touches things like Jordan has a new baby. Congratulations, Jordan.

And I think this idea that you're going to have a storybook, a picture book, where a kid's going to say what they want to happen on the next page, turn that page and have the story rendered out for them immediately from their own imagination. It's such an incredibly powerful enabling tool. We're going to see the equivalent of that in the coding in the business space too, where people are able to play with those ideas and prototype them more quickly. But that doesn't account for taste. It doesn't account for understanding your product market fit.

So I think all the same stuff is going to happen when it comes to building tools. It's just going to be easier and faster. So I certainly as a long-time engineer myself, I've been a serial founder and have a computer science background, I feel enthusiastic, not intimidated by the idea that a computer can code for me because that was never the fun part of a job. Yeah, no, it's it is weird, right? It's weird now to see, you know, what large language models are capable of that a lot of humans used to hold that close to their chest and be like, this is my special thing. and it's like, oh, well, large language models are actually fantastic at this now. Another good question here for you, William, someone asking about Lightning AI. So someone, Lisa's asking, can you provide examples of the kinds of things that you can imagine,

maybe doctors or lawyers might create with Lightning AI? Yeah. Okay. So I'm going to answer that in a general way and specific as well. But I want to teach you guys the principles of how to think about this, right? So look at your work. Anything that you're doing that is repetitive or that you have some kind of more junior analysts doing that's repetitive. Or you're finding yourself giving a lot of feedback.

Like lawyers have this problem a lot, right? Like you'll have a senior partner who has these associates that ask the same questions all the time. And they're like, look, kind of the same answer no matter which associate asks. So you might want to like upload emails from that partner and their responses. And then when an associate asks to like an expert who's that partner, it'll answer with that partner's context in mind, right? So you want to be looking at these repetitive tasks that are not really, I don't know, it's a bit of like encoding your own expertise, really is I think how you should think about it into a document that then other people can tap into. I think that's one example for lawyers. For doctors, it's going to be a lot of the same thing.

like, do you find yourself kind of going through the same answers because someone said, I have this and this and that. Like, we find you in an example model, you know, I'm not a doctor, so I don't know how good it is, but it's for diagnoses, right? So we, my co-founder, my CTO Luca came from the medical imaging world, right? And he basically looked at a dataset of, I guess, patient reports and like what the diagnosis should be. And then we fine-tint this model.

And you can ask this thing like, I'm a man of this age and I have these symptoms. Like, what is the thing? And it gives you like options, right? Like, I don't know how valid they are. But that's the kind of work where you want to have a model give you a pass at it. And then a real lawyer or doctor verify what that is. Right. So I think it's for a while we should still have human and blue for a lot of this.

But you can do that V1. Like basically convert yourselves to editors instead of creators. Have something else. Do the creation and you are the editor for that creation now instead of you creating everything yourself. Those are the principles that I would think about. I think that's such great advice, right? Yeah, finding where the humans role is and the human in the loop conversation because it changes, right? So, Will, you know, I'm curious, you know, when we talk about competition, you know,

especially as tools like Lightning AI, you know, do make it easier to turn. an idea into a product, how can startups or even businesses, you know, remain competitive? Because it seems like, you know, you can have a clone or a competitor almost out of nowhere. So what's your best advice, you know, to address, you know, how quickly competition can move now because of generative AI. Yeah, I think it's going to be interesting to see how that evolves. I think we've always, for lucrative markets, have the ability to clone. and maybe we can clone or fast follow is probably the more common case here.

The companies can fast follow on markets that are going to be lucrative where the unit economics play out. So I'm not sure that it's fundamentally different, although we may find, as you say, that they pop up out of the blue more. At the end of the day, that's where it comes back to what are your customer relationships, what's the kind of integrity of your team, the trust and reliability there. I think the real risk is if you've got the ability to create kind of overnight products, then you might end up having kind of fly-by-night businesses.

And at the end of the day, every partner out there, everyone who's taking, who's spending every enterprise who's paying a vendor for a solution, is going to want to know that there's support behind that, that that's going to grow with them, that their feedback is going to be heard. So a lot of the market-facing skills are probably going to become even more important. But that's not to undermine the idea that every technology invention has allowed us to get more sophisticated in the products themselves. I think on the one hand, we can imagine today's products a lot easier than we built them in the past.

But I think we should start thinking about tomorrow's products. It might be that we have all these amazing smart people sitting around who have now these incredible tools at their disposal. And so it's not just that they're going to create the same things faster, it's they're going to create more in the same amount of time. So I think building more features, building better user experiences, building more interactivity, more personalization, I think these things that didn't make sense before from a cost perspective are going to start to be enabled in more of our products. So I think you're going to get both sides of it. But I think from a competition standpoint, it's anchoring back on moving faster, creating more value for the customers. Those are the things we've been doing forever, unfortunately.

But AI tools will make them different, but certainly not go away. That's, I think that's great advice for those in the position of trying to learn how to build. you know, maybe separate themselves from the competition. But, you know, we've covered a lot in today's episode. So today's newsletter is going to be a good one, FYI. But I'm going to ask both of you as we wrap up here. So I'll start with you, William. So what is your best piece of advice? Maybe for, you know, companies that are looking to leverage large language models, you know, what's your advice to them on how they can, you know, better or more confidently turn new products, new services from ideas into actually, you know, economically viable parts of their business.

Yeah, great, great question. I think it all starts with the data, right? So I think a lot of it's going to be text nowadays that goes into these models. So I would write all your memos and emails and internal documentation, all your wikis, with the writing, like, don't write it for you or other humans, write it for LLMs. because you are going to feed all that stuff to an LLM at some point, right? So you should really be thinking about when you describe something, describe it so that you're going more explicit in instructions or the context

so that a model can then later use it to train, right? And then I think in the long term, the LLM is a tool that we will not really be thinking about. You don't think about the engine in your car when you drive it today. You don't think about physics when you drive a car. You don't think about how the gas gets to the engine or spark plugs, any of this stuff, right? There are some dedicated experts who know how to do that and they're called mechanics and they're very far and few, but the majority of you don't think about it. And I think that's how your products are going to end up being, right? I think a lot of luck, what we're trying

to do is make that reality today, right? So my advice, you know, kind of the next point would be like once you have, if you already have that documentation, you have those emails, you have all the things written, come to a product like Lightning or all of us and we will help you build exactly what you're looking for very, very quickly within a few weeks, right? But we need that. We need you to be able to curate that stuff because that's where your knowledge is. We're not experts at that, right? So, Will, I'll ask you the same question.

In an AI everywhere world, what's your best advice for those looking to turn ideas into products? I think right now, my best advice is to continue to experiment and to play with the actual tools. I think it's very easy to get caught up in the noise of, you know, reading about all these things, kind of learning about them in the abstract. What we find is, as with many entrepreneurial motions, that bias to action is going to serve entrepreneurs as well. So the more people can play with things, can get out there, can try things like AI Hub from, you know, from Lightning,

or visit AI.in.com has a playground where people can go in and try all the latest models from us, all the latest open source, like Lama, stable diffusion, deep seek R1. getting in playing and seeing, hey, how do I learn that? That I think, to Williams' point, about learning to speak language in the era of artificial intelligence, which has become our kind of Rosetta Stone of this. We've picked our own languages as a way to interact, but I think we still need to learn how to speak in a way that the AI understands us. And getting familiar with that is going to give a leg up to entrepreneurs and, of course, to end users alike. So I think

the biggest thing there is to continue to experiment. And the caveat on that, because it's always a double-edged sword is that at the end of the day where the rubber meets the road is ultimately creating value out of your own product. So I think there is a balance. Some people can get lost so much in the experimentation that they forget to just hunker down and solve the problem they're solving. So at some point, you kind of go to battle with the weapons that you have at your disposal and you go and create value out there. But that balance of continually experimenting to find the opportunities for innovation while just buckling down and doing the hard work of starting a business, I think that balance is going to be the trick for the next few years.

Speaking of rubber hitting the road, I think both of you helped create a very good roadmap for so many of our audience out there that are going through these challenges. So Will and William, thank you so much for taking time out of your day to join the Everyday AI show. We really appreciate it. Thanks so much, Jeremy. All right. And hey, as a reminder, to y'all, that was a lot. So many great pieces of advice from multiple perspectives. I hope you enjoyed today's show.

If you did, you're going to enjoy our write-up. So if you haven't already, please go to your EverydayaI.com. Sign it for the free daily newsletter. And if you are new here, make sure to subscribe on the podcast, you know, reach out, let us know what you want to hear more of.

So thank you for tuning in. We'll see you back tomorrow and every day for more Everyday AI. Thanks, y'all. Meet Firefly AI assistant. Now live in Adobe Firefly, the Allman One Creative AI Studio. Just describe what you want to create

in your own words and the assistant handles the rest, orchestrating multi-step workflows across Adobe Creative Cloud apps, including Photoshop, Premiere Express, and more in one conversational interface. You direct the outcome while the assistant accelerates execution. Stand control with the ability to step in and refine at any time. See it today at firefly.adop.com. And that's a wrap for today's edition of Everyday AI. Thanks for joining us.

If you enjoyed this episode, please subscribe and leave us a It helps keep us going. For a little more AI magic, visit Your EverydayAI.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.