Agentic AI Is Here: How ATOMS Turns Ideas into Revenue with Ethan Ouyang

 Instead of helping people ride co-faster, we help them make decisions,
 execute and monetize more on the end-to-end side.
 You can imagine in a single prompt,
 items can research market, design a product,
 the build system, launch it,
 and they can even optimize revenue for you.
 We have SEO agents as well.
 With all these kinds of multi-agents,
 they coordinate, we off-shade,
 and they run you very good efficiency,
 and they deliver end-to-end.
 This is right about now with Ryan Alford,
 a Radcast Network production.
 We are the number one business show on the planet,
 with over 1 million downloads a month.
 Taking the BS out of business for over six years,
 in over 400 episodes.
 You ready to start snapping necks and caching checks?
 Well, it starts right about now.
 What's up guys, welcome to right about now.
 We're always talking about what's here, what's now,
 and what's more now than AI.
 Two letters that you shouldn't be scared of,
 but you should be maximizing to get the most out of your business,
 out of your life.
 It isn't going away.
 That genie isn't going back in the bottle,
 but that's why we bring the best, the brightest,
 the coolest companies doing all kinds of innovative things today.
 We're talking about splitting things.
 We're not splitting items.
 We're talking about how you split up,
 and do a million different things with one tool.
 You can tell you more.
 His name is Ethan.
 Oh, Yang, he is the head of U.S. Department of Adams.
 It's the deep wisdom.
 It's the parent company.
 What's up, Ethan?
 Hi, Ryan.
 How are you?
 I'm great, man.
 Thanks for coming on.
 I always like talking AI.
 I like demystifying it a little bit.
 I think we're getting past it.
 A lot of people are using it.
 I don't even think we've scratched the surface
 of how capable it truly can be.
 I know that's a lot of what you guys are working on.
 What says you about the landscape of AI
 in business right now, Ethan?
 I can give you a brief introduction
 about our product items first,
 and then we can talk more about
 the general about AI and all these related businesses.
 The first item is a multi-agent system,
 with building revenue-ready products
 with our autonomous AI team.
 So instead of helping people ride co-faster,
 we help them make decisions,
 execute and monetize more on the end-to-end side.
 You can imagine in a single prompt,
 items can research market,
 design a product,
 the build system,
 launch it,
 and they can even optimize revenue for you.
 We have a few agents as well.
 With all these kinds of multi-agents,
 they coordinate,
 we off-shade,
 and they run you very good efficiency,
 and they deliver end-to-end.
 Really fascinating.
 Essentially, I'd call it a business in a box.
 Like it's turned key all done by AI in a way.
 Am I describing that right?
 Ethan, is that essentially what this is?
 Exactly, yes.
 We have an affluent audience.
 They understand business.
 They understand AI at a high level.
 I think agentic AI, though,
 is a little bit misunderstood
 and not completely leveraged the way it can be.
 Talk to me about the way deep wisdom and atoms
 leverages these agents within the platform.
 Most AI tools today are still assistance.
 They're way for instructions
 and optimize isolated tasks,
 coding or copywriting.
 I think items is fundamentally different.
 This is not just code,
 or just implementations.
 It's decisions.
 Atoms around the full decision group,
 upon many autonomous needs, research,
 planning, execution, and iteration.
 We don't help people just build or work faster.
 Atoms work on their behalf,
 or with prompts.
 And on the technical level,
 it isn't a single model,
 or just prompt.
 It's a system problem, right?
 What's priority for us is how agents coordinate,
 plan over known horizons,
 and actually ask you in real environments,
 not just reason in isolation.
 On the other hand,
 our company and our team
 have spent years publishing
 and open sourcing the foundations.
 We have a website called
 Foundation Agents Arc,
 actually published a lot of top researchers
 of the world.
 Our team try to gather everybody together
 and try to focus on the same thing
 is called Foundation Agents.
 And our system is built
 on top of that research
 and on top of those theories.
 Ethan, so if I need to train an agent,
 I need to build an agent.
 I need to call Ethan.
 Is that what you're telling me?
 Yeah.
 You can always call me.
 Yeah, or you can use atoms
 to build an agent
 or own SaaS plan for as well.
 I'm going to ask for some specifics, Ethan.
 Not like proprietary specifics,
 but just specifics of capability.
 Because I think people hear these things
 about agents in decision making.
 And I don't think they quite understand
 the level to what you're talking about.
 Because if you said most of it to now,
 you can have these agents,
 but you're kind of still always prompting them.
 It's like prompt and prompt and prompt.
 Versus truly training.
 And then real business decisions take place
 based on that training.
 Give some examples of how deep that can go
 with the decision making of an agent
 and activities they can actually do
 based on their own reasoning.
 We have already seen a lot of use cases
 that or a lot of products built from atoms.
 One example could be like a DTC brand
 direct to consumer brand.
 So maybe you are a designer.
 You have your own taste of designs
 and you only have a rough idea and a few sketches.
 And then you probably upload to atoms
 and you ask atoms according to what I have
 how to build a product I can sell.
 And then atoms will, the multi-agency system
 just ramp up right at the start
 and then the first start building first
 because they don't even know what to build
 with this limited information.
 Our deep research agent will start
 to do a deep research first
 and try to explore the market
 and see what actually have the opportunities
 here in the market.
 And then they will give you some recommendations
 and solid data for you.
 And you can actually, that's the phase
 that actually you can learn.
 You better understand what you actually want to do
 because most of the time when you prompt
 maybe you don't even have the full picture
 of what the product will look like.
 Maybe you haven't saw through yet
 but this will help you think through.
 And then you approve or say,
 hey, this is not what I want.
 You want more than you can iterate.
 You can keep prompting.
 And after you made the decisions
 you align with agents
 and they will start building.
 And when you build, there's a two feature called race mode.
 You can use the system.
 You can use different models
 or foundation models
 to actually give you the first
 MVP version of the product.
 And you can choose why you like most.
 And then you can continue with that version.
 We start a model,
 an action navigation model.
 And then it starts with the execution phase.
 In the execution phase,
 we keep human in the loop.
 You human can make the creative decisions
 like most of time.
 The agents will just run and implement testing for you.
 And then eventually you can publish
 and then our SEO agents can also help
 with optimizing the revenues.
 This is an example that we build things.
 We communicate with people.
 And everything is delivered to end.
 People don't have to have a very clear idea.
 They don't have to control everything.
 They just need to make key decisions.
 Yeah, so they become the manager,
 but not necessarily at a level
 where they know everything
 that how it's getting done.
 They're just controlling what gets done.
 We used to live in a world
 where the how really mattered
 because to get it done,
 you needed to know how.
 Now it's more what do you want?
 In a lot of ways.
 Yeah, or you can find some people.
 You can hire some people.
 They know how,
 but that's way more expensive
 or takes more time.
 And then it helps turn time.
 Are we replacing ourselves, Ethan?
 Is that what's happening?
 No, no, no.
 This just focus is different now
 because originally when you have idea,
 you don't even know it's a good idea, not.
 You don't even know it's going to make revenues or not.
 You have to get some resources first.
 You don't need to hire people to actually implement for you.
 Then you go to the testing phase.
 But now the execution is new.
 Instant, the judgment,
 the taste become more important.
 That really changes how
 who gets to build a company
 or who gets to build a product.
 You have your own resources.
 You have your own judgment,
 your own taste, your own preference.
 You can go ahead and try and test.
 And then you probably find something that's better.
 You are also growing.
 People are also growing from this iteration.
 Yeah, you get knowledge.
 I came up in a time working with brands
 and doing marketing.
 Spent hundreds of thousands of dollars
 in months and months.
 Many big brands had that.
 But now it's more accessible
 for this research and knowledge
 that used to be only attainable
 by large corporations.
 It's now attainable to guide small business decisions.
 And that's where the power of this comes from
 for the entrepreneurs that are willing to sort of put there.
 Oh, I got an idea to the side and go,
 I got an idea and it can actually generate revenue.
 Talk to me, Ethan, about what we ultimately output here.
 Because I get a lot of different places.
 Ecom and D to C makes a lot of sense.
 Are you familiar with like base 44?
 Yeah, I've heard that.
 App building is prompt to app.
 It is all of that capability sort of built into
 atoms as well that it can literally
 give you from prompt to visualization.
 I know that your tool is does more than that.
 But does it have that capability
 if you want to do a SaaS based
 or develop a tool that's used internally
 in a company or something?
 Is all of that here as well?
 Yes, actually that's one of the reasons
 we call product atoms.
 A product is sealed on top of a lot of unit features
 or like functions.
 There's so many features or functions
 living in the software world, right?
 About database, about storage,
 about payments you need to be able to receive money
 and pay money to buy stuff.
 Also about recommendations, about deployment
 after the code is built, you need to have a container
 or deploy your web or your application to the cloud.
 Everything ends to end.
 And those are the core features we support.
 So those like you can preview your product,
 you can basically store your data.
 We can support like logging and logout.
 And there's a time-trick fact.
 If we use one ID for users,
 we can also like implement,
 we can also support the recommendations feature, right?
 If you build e-commerce website,
 we have a building like recommendation engine
 for using logging.
 And then say, hey, this product looks fine.
 I probably want to buy that.
 Actually, that's because we have some building features
 inside, we have all these features.
 That's the very core capabilities for our product.
 I'm very familiar with base 44.
 I've used it to build several apps.
 It's visualizing the app on the screen to the right.
 You got to write a left prompt, give me a database
 and log in for admin and users on app platform
 that looks like this example that does these things.
 Building it in web app environment
 that is usable right then.
 Exactly, that's our core capability.
 That's only part of the end to end flow.
 It's more on the execution phase.
 That's also very important.
 Execution is very important.
 Ethan, I know that the tool 80% less cost
 than a lot of other tools.
 So Ethan, talk to me about cost here.
 What can people expect?
 We have our own foundation agents department
 or this group we have spent years publishing.
 And that really give us the cost efficiency
 from our agents and how we orchestrate our multi agents
 and how we design our system.
 Everything is more on the technical side.
 Those researchers really help a lot.
 And also on the other hand, we model agnostic on the backhand.
 So basically we use different foundation models.
 Sometimes we use open source foundation models
 which is way cheaper than those closed source models.
 So it depends on the task, right?
 We have a good way to try to deliver the same impact,
 deliver the same performance with all costs.
 That's our advantage and that's pure technology.
 It's a little meta, to be honest.
 You're using AI, I bet, to pick what AI you use from model.
 Let LLM in a way, that's what it sounds like.
 I'm a hearing crack.
 Yeah, we are AI-related company.
 Everybody in the company uses AI,
 not just like engineers, in the classic software company.
 You may see designers and task engineers
 backhand from the engineers.
 Now we are going to AI native and designers
 can also use AI to create the prototypes or docs
 and our engineers are more end-to-end.
 They use AI to write better performance code
 and they use AS help to actually co-design the system.
 I'd say from personal experience,
 back to this change of how to do it versus what you get.
 I find you have to be really good at debugging.
 That's a skill set when I've been doing apps
 that's getting underneath the right questions to ask,
 not how it gets done, but asking in a way
 that you sort of sort out the things that inevitably come up.
 I'm just speaking from experience with base 44,
 developing tools and apps and things.
 Inevitably, you run into these mismatch of code
 that an activity you expect to happen does not happen
 and they have self-correction in a way,
 but it's not always perfect.
 Help me understand how atoms works through those types
 of challenges and things when sort of building out tools.
 Yeah, there are two aspects.
 One is from our product side,
 we pay polishing and improving our product from internal.
 We've been like building bugs in your system
 and that will help the system to really less
 but upgrade more reliable or more higher performed outputs.
 And that's the thing that we are iterating quickly.
 We're also having a lot of talents joining our company
 and try to optimize our upgrade and optimize our product.
 That's one thing.
 And the only other hand for the user experience,
 we are posting blogs, we are posting documents and Q&As
 to majority of users because most of the time
 our users don't know how code work.
 They don't have an engineering background
 but that's fine.
 Actually, they are our target audiences.
 And so we just try to help them on board
 and they'll try to help them feel more better
 when they see about they should know it's not the end of the world.
 You have a way to make it work
 but just need to be patient
 and they just need to probably use the correct way.
 We try to give them support
 as many supports as possible, two aspects.
 How sophisticated can atoms go
 and who is the ideal customer for atoms?
 Our product is a global product.
 We call it Atoms.
 We launched in the US but actually launched worldwide.
 It's talking on solo funders, indie hackers
 or small business or small teams
 who doesn't have that many resources or domain knowledge
 which means most of the time you need a big team
 to have all this knowledge in the house in the room.
 That's our target audiences.
 And in terms of what we can build,
 I can give you some examples.
 I already gave you a DTC consumer brand example
 and there we have some more reuse cases
 we collected from our existing users like
 a businessman who runs window cleaning business
 and they use to rely on multiple apps to get things done.
 And now they build a single application
 that brings together booting, booking, estimate,
 scheduling and customer documents in one place
 and they also take in also that app
 can also handle payments.
 Everything you can, so that's why we call atoms.
 So the business depends on what kind of features
 or what the actual requirements you need
 and then we just provide those features
 and our AI agents try to select and try to query to select
 and to based on your request or your request
 and we can build with this combination
 you can build whatever you want to build almost, right?
 Because whilst we are not saying we're supporting
 all these kind of features you can imagine
 but we are iterating, right?
 We'll keep adding the recommendation feature
 maybe in the future.
 And so it's not currently not now
 because it's more on the data side
 we probably need more data
 when it's actually getting top priority.
 That's actually one way example.
 Also like we've seen the Florida based insurance company
 use AT&T to build their nending payers
 and also all these queries on their features
 inside their company to brand their products.
 Ethan, working everyone learn more
 about the software website details, social media,
 give any of those details for our audience.
 We have AT&T.f, that's our official website
 and you can just visit that website
 and you know you can sign up or you can try free
 and try to build your own stuff.
 We have all these social media and live.
 We post on X, it's also called AT&T.f
 and then we have linking for the recent talk with me
 just prefer to go to linking and X
 and all these social media try to search for our AT&T.
 Thank you for coming on the show, Ethan.
 Appreciate it having you.
 Thank you Ryan.
 Thank you for having me.
 Hey guys, you're gonna find us, Ryan is right.com.
 You'll find the full episode here with Ethan and Adams
 and deep wisdom.
 They're doing some cool stuff.
 We'll have links to all of the stuff that Ethan talked about
 and ways to get in touch with them on social media
 and learn more.
 Look, it's not time to fear.
 Time to get your ass on it.
 It's time to do it.
 That's why we're bringing these guests.
 We're trying to give you the knowledge
 to put you ahead right now.
 We'll see you next time.
 We'll write about now.
 This has been right about now with Ryan Alford,
 a Radcast Network production.
 Visit RyanisWrite.com for full audio
 and video versions of the show
 or to inquire about sponsorship opportunities.
 Thanks for listening.