Ep 22 - Beyond the Code: Reimagining the Human-AI Partnership w/ Kyle Johnson

From protecting billions of people against online terrorism at Meta to founding tech startup Ara, Kyle Johnson is pioneering new ways for humans and machines to collaborate. His vision goes beyond building AI-powered tools—he's creating systems that empower everyday people to harness artificial intelligence without a technical background.

Kyle introduces us to the revolutionary concept of "vibe coding," where anyone can create sophisticated applications simply by describing what they want in natural language. Unlike previous technological waves that required specialized knowledge, today's AI interfaces through conversation, making it uniquely accessible. As Kyle demonstrates, someone with no coding experience can build everything from educational flashcards to custom games just by having a dialogue with AI tools like Claude.

What makes Kyle's perspective particularly valuable is his focus on the invisible ROI of artificial intelligence. While many companies chase flashy AI features, Kyle reveals how backend implementations often deliver far greater business value. Drawing from his experience with counterterrorism and integrity systems at Meta, he shares practical strategies for identifying repetitive business processes ripe for AI enhancement—where a single implementation can save hundreds of hours while improving employee satisfaction.

The conversation takes a fascinating turn when Kyle explores the future of human-AI collaboration. He envisions a world where the code itself becomes increasingly invisible, where interior designers might transform spaces through natural language in VR, or data analysts might investigate patterns through conversation rather than queries. His advice for aspiring founders cuts to the heart of successful innovation: fall in love with problems rather than solutions, embrace uncertainty as a sign you're on the right track, and remember that both AI and human expertise have essential roles to play.

Ready to start your own AI journey? Connect with Kyle on LinkedIn at linkedin.com/in/gkjohns or find him on Twitter @kyledotai to learn how you might transform your creative process through the power of AI collaboration.

Want to join a community of AI learners and enthusiasts? AI Ready RVA is leading the conversation and is rapidly rising as a hub for AI in the Richmond Region. Become a member and support our AI literacy initiatives.

Episode Transcript

Speaker 1 (00:00):
Welcome RVA to Inspire AI, where we spotlight
companies and individuals in theregion who are pioneering the
development and use ofartificial intelligence.
I'm Jason McGinty from AI ReadyRVA.
At AI Ready RVA, our mission isto cultivate AI literacy in the
greater Richmond region throughawareness, community engagement

(00:24):
, education and advocacy.
Today's episode is madepossible by Modern Ancients
driving innovation with purpose.
Modern Ancients uses AI andstrategic insight to help
businesses create lasting,positive change with their
unique journey consultingpractice.

(00:44):
Find out more about how yourbusiness can grow at
modernagentscom.
And thanks to our listeners fortuning in today.
If you or your company wouldlike to be featured in the
Inspire AI Richmond episode,please drop us a message.
Don't forget to like, share orfollow our content and stay up

(01:07):
to date on the latest events forAI Ready RVA.
Welcome back to Inspire AI, thepodcast where we dive into the
minds shaping the future ofartificial intelligence.
Today's guest is someone who'snot just building with AI.
He's rethinking how humans andmachines collaborate.

(01:27):
Kyle Johnson is a techentrepreneur, data scientist and
writer based in Richmond,virginia.
He spent six years at Metaworking on machine translation,
digital trust and complexsystems that shape how billions
of people experience theInternet.
That shape how billions ofpeople experience the internet.
But these days, kyle is focusedon something much more personal
helping everyday people harnessthe power of AI.

(01:54):
He's the founder of Aura, aplatform that helps teams turn
raw data into visual storiesthat actually drive action, and
the founder of TidewaterResearch, where he helps
enterprises navigate the messyreality of AI adoption.
Beyond the products he'sbuilding, kyle is emerging as
one of the most thoughtfulvoices around how AI is quietly

(02:14):
transforming the way we think,create and work.
In his recent article, theQuickening, he argues that
transhumanism isn't comingsomeday.
It's already here for those whochoose to collaborate with AI.
Today, we'll explore Kyle'sjourney from big tech startups,

(02:38):
how non-technical folks canstart experimenting with AI and
what it means to vibe code yourway into the future.
It is my pleasure to welcome.

Speaker 2 (02:51):
Kyle Johnson to Inspire AI.
Hey, thank you so much, Jason.
Really appreciate it.
Happy to be here.

Speaker 1 (02:59):
Kyle, thanks for joining us today.
Can you start out by tellingour audience a little bit about
yourself, your business and yourinterest in AI?

Speaker 2 (03:03):
Yeah, absolutely so.
I'm an ex-marine and formermetadata scientist that's kind
of where I got my start in techand I currently run an AI
consulting firm and, as youmentioned, my other company,
aura.
It's Aurasocial, and Aura iskind of like GitHub for
storytellers that's what we callit.
It provides infrastructure fordata scientists and data
analysts who need a place togenerate, store and share their

(03:26):
insights, not just data, so it'svery insight driven, as opposed
to just piping data to moreplaces, which is where a lot of
data startups are going now.
My interest in AI is very muchon the technical side, but it's
very applied, which I think hasa lot to do with just the teams
that I chose to work on on Meta,to do with just the teams that
I chose to work on on Meta.
I'm a tinkerer at heart.
I like actually playing withthings, actually trying new

(03:47):
things and finding opportunitiesto just make entities and
organizations more efficientwith this new technology.
I just love doing it.
But I don't just focus on thetechnical side and technical
implementations of things.
I'm also really interested injust how people are using AI to
make their lives easier, andusually that just means chatbots
, I think a lot of people seethem as a bit of a gimmick.

(04:08):
Right, they're like, okay, thisthing isn't really smart.
But in the article you spokeabout, I mentioned folks who
I've seen with my own eyes whileI was consulting really make
themselves smarter with chatbots.
They found these new ways ofworking with them.

Speaker 1 (04:31):
So I'm really interested in just all aspects
of how AI is changing our lives.
Absolutely, the concept of AImaking people smarter that
definitely resonates with me.
It's almost like there's thisnew digital literacy, but it's
not about learning code.
It's about learning to thinkwith AI.
Right?
It's not about learning code,it's about learning to think
with AI, right?
That's a pretty powerful shift,absolutely, yeah, all right, so
tell us a little bit about howwould you describe the journey

(04:53):
that's led you from meta tofounder life.

Speaker 2 (04:58):
That's a good question.
So when I was working at metaas a data scientist, I really
just chose to work on teams thathad high stakes or high impact
missions.
I didn't like I didn't want tobe on like the newsfeed team
right, which had just beenaround for a long time.
It had been optimized to helland back right.
There was just wasn't a lot oflike high impact work to do.
So I chose teams likecounterterrorism, child safety,

(05:20):
ai, language, research at FAIRand impersonation.
So I think that that had a hugeimpact on me in that I chose
teams where the outcome reallymattered right.
And, as you know, like when youwork at a big tech company,
there are teams where it doesn'treally matter and you can just
kind of get along by doing thework.
But I always stayed on verymission-focused teams.

(05:40):
But when Lama 2 dropped in 2023, I want to say I started almost
exclusively just incorporatingAI into my own work and some of
the tools that our teams wereusing and I just took into brand
with it.
There were a lot of datascientists at the company.
I felt very confident that datascience analytics would
continue without me.

(06:00):
I risked getting fired and justreally went all in on this AI
implementation kick.
I was on and it wentsurprisingly well.
I was working at Reality Labs atthe time on a team called Trust
and I built a system thatmonitored what users were saying
on external social mediaplatforms about the products
that launched.

(06:20):
So we launched a new headset.
I was building apps that wouldlook at all the comment sections
on Twitter, all the what else.
We looked at Reddit posts.
We didn't use Facebook posts,but I was basically ingesting
all that and using AI to makesense of it, and that was
something that a lot of peopleweren't doing.
And when people thought of AI atthe time, it was very focused

(06:43):
on some external facing chat bot, right, some external facing
use case or feature where theuser's directly interacting with
the bot or with the AI.
And I'm like man, I'm getting alot of mileage out of just
having the AI understand thingson the backend and write me a
little report.
And I think that that paradigm,that like way of seeing AI

(07:03):
adoption and AI implementation,was unique.
And at a certain point I waslike, okay, I don't see many
people doing this.
I see the market opportunityand that was all the confidence
I needed.
So I spoke to Mari, my wife andco-founder now co-founder and
she was like yeah, that soundslike a good idea and we opened
up Aura.

Speaker 1 (07:22):
There it is.
Yeah, that definitely stickswith me too.
Data is not valuable until youturn it into action, right?
Yep, so R is helping peoplebridge that last hardest mile of
telling a story in a way thatactually moves someone to do
something.

Speaker 2 (07:41):
yeah, yeah, yeah, it's funny.
You say like the hardest mile,because I call it the last mile
delivery of, like the data valuechain.
Right, you actually have to getit into someone's hands.
Who is going to make a decision?
Right, a lot of data scientistsand data analysts.
This is probably the number onemistake that data scientists
make in their first like fiveyears of their career.
They make these awesomeanalyses, but they don't think

(08:03):
about that last mile delivery.
They don't think about whothey're going to even deliver
this thing to.
They don't think about whatdecision it's going to support.
They don't think about, youknow, at the end of it, I always
tell, like, my mentees to likemake a recommendation for what
to do next, but they would a lotof the time they would make a
recommendation that was likecompletely out of touch with
where the business was, and youknow you can do a lot of work

(08:25):
and not get much for it.
So, yeah, that last miledelivery is very important and
at Aura, I think it's.
You know, I think that's where,if you think of the value chain
as a funnel, that's where a lotof the value drops off, right,
just getting that insight intothe right people's hands.
So that's what we're focused on.

Speaker 1 (08:41):
Yeah, awesome.
Yeah, it's really good to knowyour customer and to be able to
articulate the problem you'retrying to solve with the
technology before you go diveinto just experimentation mode.
I think that's one of thereasons why, like 90% of all
data science, experimentationdoesn't make it to production,
you know you know?

Speaker 2 (09:07):
yeah, totally agree.
I think that having um it'sfunny because data science it's
it's, it's firmly in like thetech, the tech world.
Like you would sit on a productteam with, like engineering
managers, ux researchers, dataengineers, but I think it really
straddles the line.
Sometimes it's like almost abusiness function in a lot of
ways and I think that knowinghow to be a translator between
the two worlds it's totally abusiness function, yeah, yeah
absolutely, and um that, knowinghow to be a translator, between
the two worlds.
It's totally a business function.
Yeah, yeah, absolutely, andjust knowing how to translate

(09:29):
between those two worlds issuper important, but it's not
really spoken about.
Everyone's worried about thenewest models or being super
stats rigorous, which you shouldright.
That's good, but you need toreally be thinking about why
we're even doing this, right?
It's?

Speaker 1 (09:41):
not tech for the sake of tech, for sure.
Yeah, absolutely All right.
So let's double click on Aura.
What problem are you solvingand what makes it different than
other data tools?

Speaker 2 (09:53):
Yeah.
So the way I like to pitch Aurais it's GitHub for data
storytellers, right?
Software engineers have thisplatform for those who don't
know called GitHub, where youcan kind of share your code with
everyone else and it acts likea central repository where you
can work on code together.
You can take someone's projectand fork it, which means kind of
creating a new project out ofthat one, and it's kind of the

(10:17):
one central place where yourcode can live and be shared
right, and in some cases, beeven like deployed from right If
you're using something likeVercel.
But there's nothing like that.
For people who are telling datastories, right.
For folks who need to createanalysis, they need a central
place to create the analysis, tostore it and to share it from.
There's nothing right now, andI'd argue that in most

(10:39):
organizations, google Docs isdoing the hard work.
People are doing an analysis ina Jupyter notebook, right, or
an R Markdown notebook.
They're screenshotting theircharts, they're pasting them
into a Google Doc and they'resaving it as a PDF right, and
that's where all the dataanalysis is happening.
There are very uniquechallenges of doing a data

(11:01):
analysis and no one's reallytackling them, no one's really
facing them.
So that's what R is doing, andI think that one part of one
concept that hasn't really thatisn't really spoken about is
again what I call like the datavalue chain.
Right, so you start with rawdata, which is just observations
of things that happened.
Right From there you have toturn that into insights.
Those insights need to be fedinto decisions.

(11:23):
The decisions need to driveaction and then from the action
there's impact, right, so in away it's a funnel, right.
There are drop-off pointsbetween each one.
Sometimes you don't have thedata to make the insights,
sometimes the decisions, forwhatever reason, don't become
actions.
We're really focused on thatinsight to decisions transition,
right, data scientists know howto turn data into insights.

(11:45):
But from going from insight todecision, having what I like to
call a storefront for yourinsights, right, if the data
science, if the data scienceteam or data analysis team, was
a business within a business,which it is right you're kind of
selling the insights, they needto be timely, they need to be
related to the context of what'sgoing on in the business and

(12:06):
they need to be actionable,right, but that's what you're
selling, that's what a data teamis selling.
So we're kind of buildinginfrastructure for those data
teams who need to sell theirwares, as it were.

Speaker 1 (12:17):
Okay, so help me visualize this a little bit more
.
You're selling infrastructure.
What does that mean?
In like something that they'retrying to deliver.
So they have this concept.
It's data driven.
They want to get their conceptand the richness of their
analysis out of a Google Doc.

(12:38):
What does it look like to themin their final product?

Speaker 2 (12:43):
Yeah, so it'll look like something close to medium
is how I would describe it.
Medium is great because theyreally put a lot of effort and I
respect them for this a lot ofeffort into what the articles
that you write look like.
Like you can tell, they put alot of design into, a lot of
design effort into it.
They made it very easy tocomment on specific parts of it
and highlight certain parts ofit, right, and I really, really,

(13:06):
really, was kind of inspired bythat experience.
So what we want to have is, forevery data team, a nice little
repository of analyses that have, you know, interactive charts
on them, right?
That's something that is thatshould be standard, you know,
for folks doing data analysis.
Like, if you make a chart, whyare we still screenshotting?
Like, why are we screenshottinganything?

(13:28):
So I really want to make theexperience of reading an
analysis and sharing an analysisgood, and I think that, like,
when it comes to sharing, too,like a lot of the you know,
there's this concept of, like,open graph information where
there's, you know, when youshare something on social media,
you see like a little preview,right, why shouldn't you, why

(13:48):
shouldn't that shared preview behighly tailored to the analysis
that it's related to, right andit's, you know, little features
like that can really go a longway and they're not like I've
even gotten to like the AIfeatures, right, but there are
just little things that wouldmake the product or make the
experience of folks doing dataanalysis and sharing it a little

(14:09):
bit sweeter.
So we're kind of just likehyper-focused on this one area
of data analysis and datascience.

Speaker 1 (14:15):
Cool, all right.
So you've worked on everythingfrom machine translation to
integrity systems at Meta, and Ijust love the fact that you've
worked on counterterrorism andall of that.
That's amazing.
I would love to hear somestories about that sometime, but
tell us, how has that shaped,how you think about AI today?

Speaker 2 (14:37):
Yeah, I actually love this question because it's one
I hadn't thought of and I'm soglad we're talking about it.
So I think the primary and thisis kind of a different answer
that a lot of folks might expectI think the primary way that it
affected you know that workingon these very unique teams
affected.
My view on AI is that I spent alot of my time working with

(14:57):
decisions made by humans, right.
So imagine content review.
You know Meta is not just goingto delete someone's account
because a machine said to do itRight, at the end of the day,
you need what we call groundtruth or a human to look at and
be like yeah, this really is anaccount impersonating Elon Musk,
right, which was a very commonthing at the time.
So, you know, when humans aremaking those decisions, it forms

(15:20):
a kind of data.
It forms a kind of data, right,like they might have.
We might have sampled 1000accounts, right, label them
whether or not they're spam,whether or not they're
impersonation, and then we havesome estimate of the prevalence
of this problem, right.
And you know that experience ofworking with humans to make

(15:41):
these decisions with semi smalldata, and you know having a
policy that the humans werereviewing against.
I didn't know it at the time,right?
But that is the perfect mentalmodel for how I like to leverage
AI on the back end, right?
So the reason we used humans onthe back end was because, at
the time, they were the onlyones who could say you know, yes

(16:03):
, this post is alluding to someterrorist organization.
It was actually really hard tobuild classifiers to do that
back in the day, but now youknow, an AI can easily do that,
and they can do it with imagestoo, right?
So the game has completelychanged, and what I realized,
you know, when AI started to getreally good, was that I can
take that exact same process anduse it for whatever we want,

(16:26):
right?
So a good example is the productor the feature that I spoke
about a little bit earlier,where you can kind of ingest all
the social media chatter aboutyour product or something that
happened at your company and useAI to evaluate it.
Say, hey, is this, is the userhere or the poster praising our

(16:46):
company or no?
Are they mentioning a specificproduct?
Are they mentioning what mightbe a bug?
What's the sentiment?
Positive or negative?
Are they mentioning a specificperson who works at the company,
because they did that a lotwith Facebook, right, and you
can take all of that, aggregateit and then just do data science
on it, right, you can justchart it over time Is it going
up, is it going down and thatwas a really, really, really

(17:09):
powerful perspective.
And I think that you know, whilea lot of folks are focused on
these again, these likeuser-facing AI features I've
always been very concerned withusing AI to become more
efficient in a lot of ways right, to do things more quickly, to
do things with less friction, todo things with a huge cost
reduction.
So, you know, the flashy AIfeature, everyone wants to do it

(17:31):
.
But the value sometimes it'shard to communicate because you
don't really know how it's goingto land, right, you don't
really.
Until you test it, you don'tknow what it's going to do to
your daily active users orwhatnot.
But with this approach, youknow, using on the backend, we
can say, hey, here are fiveprocesses that we're going to
automate and we can.
You know, if we do thiscorrectly, you know, within,
with this tolerance for accuracy, you can expect about an 80%

(17:54):
cost reduction.
And that resonates with withCEOs and leadership, because
it's just so tangible.

Speaker 1 (17:59):
Most definitely.
Yeah, it's a great reminderthat some of the highest ROI is
invisible.
Yeah, we all want to make themachine run smarter in the
background, right.

Speaker 2 (18:11):
Yeah, no, that's a great way to say it.
The best ROI I might steal that, actually the best ROI is
invisible, and I think thatthat's again.
Ai is just like any hype trainin tech.
You see the flashiest parts ofit, you see folks getting $100
million valuation off ofsomething that looks very

(18:31):
visually impressive, but at theend of the day, it's a
technology that needs to beleveraged for your own purposes,
and I see no shortage ofopportunity in just making
things more efficient on theback end.

Speaker 1 (18:43):
Yeah, all right.
Okay, so we have a fairly large, I would say, professional
audience that needs to hear howthey can take these concepts and
do something with them.
So what's the biggestmisconception non-technical
folks tend to have about AI?

Speaker 2 (19:03):
Yeah, I love this question.
I think the biggestmisconception non-technical
folks tend to have about AI yeah, I love this question.
I think the biggestmisconception they have is that
they can't participate in thisnew wave of tech.
So, if you think of the lastfew waves, they were a little
bit less accessible, right, Ithink of, like the big data wave
about 10 years ago.
Right, that was inherentlytechnical.

(19:25):
You needed not only like tounderstand on the data
engineering side how to evenmanage all this data, but you
also needed some stats knowledgeto process it correctly and to
extract the insights and whatnot.
This is different.
Ai is different because of themedium that it happens in.
Right, ai, you know, god blessthe engineers who created it

(19:45):
this way it operates off ofnatural language, right, the
whole point is that you can justtalk to it like a human right,
and they've kind of tweaked itto make it even more human right
, to make it like respond to youas if it were in this
conversational way.
And I think a lot ofnon-technical folks, you know,
when you hear artificialintelligence, you're like, oh my
God, this is like the finalboss of things that I should not

(20:07):
be talking about.
Right, it sounds.
It's almost scary, scarysounding, but at the end of the
day, you can kind of leverageyour intuition about how you
would interact with a human waysto use it.
But I think this thoughtpartner approach, you know, that
we kind of spoke of a littlebit earlier is by far the most

(20:28):
effective way to get startedwith it.
So, in the article we mentioned,you know there are complex
situations like real estatenegotiation or like working with
doctors to advocate for afamily member's medical care.
Medical care those aresituations where you should be
bringing the context to the AI,storing it in a project or just

(20:48):
dumping it into the window andhitting enter and then kind of
using it to reason back andforth.
And that is so effective.
And we're at the point youmight see a video from me soon.
We're at the point where, withthe same context, I would argue
that in my experience, thechatbot is giving me better
responses than I can get myself.
Right, I'm just going to say itRight, yeah, they're very good

(21:10):
at reasoning and I think youknow, being non-technical, you
might think, ok, I'm not goingto be able to like really
leverage these things.
It's like no, you can, you cango really far with the chatbot
no-transcript.

Speaker 1 (21:32):
That's pretty wild.
Yeah, I feel like maybe the gapis between technical and
non-technical people.
It's not really that.
It's more like those that useAI like Google versus those that
use AI like a collaborator, andthat's where the reasoning
comes into play.

(21:52):
Double check my know.
I saw one of your recent postsabout you know what was it.
I think you wrote it up here.

Speaker 2 (22:00):
I use this all the time.
I think I use it almost everyfive chats.
I like make a point to do it.
It's what about this situationis clear to you?
That doesn't seem clear to me.
Ah, yes, that question is sogood because it will.
It makes you take a step backand you'll usually and you're
usually kind of shocked at theresult when you first see it,
you're like, oh my God, like youare spot on, right, yep.

(22:20):
So, yeah, I like to ask thatall the time, and what I say
it's doing is it's kind of likefilling in your mental blind
spots in a lot of ways.
There are a lot of times whenyou, when we are so immersed in
a problem that we can't take astep back and be like, hey,
maybe we're focused on the wrongthing.
Or one that I got from ChatGPTrecently was I was talking about

(22:41):
my startup and it was like youknow, you're probably farther
along than you think you are,you know, when it comes to the
branding of this thing, and Iwas like, whoa, thank you,
that's a great compliment, rightFor sure.
But yeah, little things likethat.
It'll really just illuminateparts of the problem or parts of
your thinking that weren'tsuper clear to you.
Very useful.

Speaker 1 (23:02):
Yeah, I'm going to set that one to memory.
What's clear about thissituation to you that doesn't
seem clear to me?
Awesome, yeah, all right.
So if I'm a business leader whodoesn't code, what's the first
thing I should do to startleveraging AI effectively?

Speaker 2 (23:22):
That's another great question.
It's something that I end upspeaking to a lot of business
leaders about, and they asked avery similar question to this.
I'd say the first thing is totry it yourself, right, really
get used to using the chatbotyourself.
There are a lot of businessleaders who, rightly, are
skeptical, and I understand whythey'd be skeptical, because you

(23:44):
know, outsourcing any kind ofthinking to a machine is risky,
right, I can understand wherethey would see where they'd be
hesitant, but I would, I'dencourage them to view it almost
like a combination of a mentorand an employee.
Right, when it comes to youremployees, they're going to make
mistakes, right, you expectthem to make mistakes and you

(24:05):
have a process for managing, forcatching and managing the
mistakes that they make.
You got to view AI the same way, right, every once in a while.
Yes, it might.
It might hallucinate, as theysay.
Right, it might give you ananswer that you're like is that
really true?
Then you go to Google andsearch and you find out it's not
true.
That's fine, right, it's goingto happen, you know.
Be vigilant.
So, again, number one is toreally try it yourself, right,

(24:29):
and get in the habit of using itagain as a thought partner
right, using it to talk throughproblems and talk through
dilemmas and to just gainclarity.
That'll give you a surprisinglygood idea of AI's capabilities
too.
I think a lot of folks are kindof out of touch with how good or
bad AI is, and I think thosedaily interactions of just

(24:51):
talking through problems with itwill show you exactly where the
limit is right.
You'll start talking aboutsomething and you'll see it's
like you exactly where the limitis right.
You'll start talking aboutsomething and you'll see it's
like okay, this thing is not.
You know, understanding whatI'm saying right now.
That's the limit, right, andthat the experience of hitting
that boundary is good for youright.
It'll teach you a lot of likewhat AI is capable of.
And secondly and this issomething that I do to start out

(25:13):
, pretty much every consultinggig is have your team leaders
you know, however, your businessis structured conduct a really
quick audit of just therepetitive processes that their
departments are doing Right andthere will be many of them Right
.
So the customer service team isa great example.
This is where I'd argue thatmost of the best processes come

(25:34):
from in terms of like automation, their ability to be automated.
So you would go into thecustomer service department and
say you know what are our repsdoing every single day.
That takes a long time, right?
One example I found recentlywith a client was she was
working, she was leading a callcenter, right, and it turned out

(25:54):
that her reps, what she woulddo every single week was she
would sample some number ofcalls, read through the
transcripts and evaluate themagainst some rubric and then
write up a little report and doan evaluation.
And that is a textbook AIimplementation case, right,
because again, when it comes toAI for understanding, it's so

(26:16):
good, right?
So what we ended up doing wasbuilding a little pipeline that
took the audio transcribed, itfed that to a chat bot next to
the rubric, said you know,generate some evaluation, you
know, using structured output,and from there we had basically
a checklist of what the rep didor didn't do and then a little
report for what they could dobetter.

(26:37):
And doing that kind of unit byunit in your business.
That will take you a long way.
If they can say hey, here arethe five steps to what I do, you
can hand that to an engineerwho knows how to build AI
pipelines.
And that is the program flow ofwhat they're going to build.

(27:02):
So that process of just goingthrough finding the repetitive
processes, naming them,outlining the steps and then
just passing that to a team ofengineers who know what they're
doing with AI is powerful and Ithink that that alone can.
You'll see crazy costreductions.
Like you'll see, like you're,if you can do this in a week and
save like hundreds, hundreds ofman hours, it's ridiculous,

(27:26):
it's a crazy time to be doing AIimplementation.

Speaker 1 (27:29):
Plus employee satisfaction.
Personally, like nobody reallywants to be doing those
repetitive things.
Exactly, exactly, exactly, yeah, yeah, right.
So your advice is to personallysharpen your thinking, your AI
access and leveraging it on thedaily, and then upping your

(27:50):
business operations bysharpening your tools in
automating the repetitive thingsout of the business.
That's great, right.

Speaker 2 (28:00):
Yeah, and I think I start with the personal part,
too, where I'm like, hey, youyourself should be trying this
thing right, Because I think itdeals with the skepticism very
well.
And I've yet to meet a businessleader who actually committed
to using it to make decisionsand didn't like it.
I have yet to meet one.
I've done a bunch of consultingand I haven't met a single one
who, after giving it a real try,has been like ah, this isn't

(28:23):
useful.
And I think once they see itfor themselves, once they're
like, oh my God, this thingtaught me something, that's when
they're open to AIproliferating through the
company.

Speaker 1 (28:33):
Yeah.
So with the inverse in mind, doyou think there'll be a time
when people need to be concernedabout how much the AI is doing
for them and how little they'reusing their brain functions?
Have you considered that at all?
Crawling through the minefieldof figuring things out on our

(28:57):
own, it's now all abstractedaway.
What's the future of humanthought if AI is doing all of
that grunt work for us?

Speaker 2 (29:10):
That's a good question.
Yeah, I worry about thatsometimes, especially with
creative writing.
Oh, yeah, exactly, I used towrite.
If you go through my Mediumaccount, I used about that
sometimes, especially with likecreative writing Like oh yeah,
exactly.
I used to write.
If you go through my, like mymedium account, I used to write
a lot, you know, but now it'sall so AI assisted that like I'm
probably not as good of awriter, like I got to admit that
and I don't know, you know,maybe maybe in the future, you

(29:32):
know, everyone will adapt to AIwriting most of the code, just
like no one has to really docomputation by hand anymore.
But that's a really goodquestion.
I don't know.

Speaker 1 (29:43):
We'll keep thinking about it then.

Speaker 2 (29:45):
Yes, yeah, we definitely will.
It's going to come up again.

Speaker 1 (29:48):
All right.
So tell us in your experience,how can individuals without a
coding background leverage vibecoding to bring their ideas to
life?

Speaker 2 (30:00):
I love this question too.
So start with a chat bot, right, and start small.
That's definitely the approachyou want to is like daunting,
right, like a lot of like.
There's just so much and you'renever going to feel like you're
really like good at it becauseproblems are always popping up.

(30:22):
So I would say start reallysmall and start with a chatbot
just making a single HTML, cssJavaScript file right, a html
file, so you don't have a codingbackground.
Right.
Get used to building littletools.
Right, a dot HTML file, so youdon't have a coding background.
Right, get used to buildinglittle tools, right.
So a good example is flashcardsfor your kids.

(30:45):
You know, your kid has a testcoming up on something simple
like I don't even know, likemultiplication tables right, go
to Claude.
Claude is good, because it letsyou see, lets you build things
and see the actual output, likethe web page that you built in
the UI, which is amazing.
Go to Claude and say, hey, mychild is studying this.
Maybe even take a picture ofone of their homework

(31:06):
assignments.
Here's what my child isstudying.
Can you make me some flashcardsthat they could use in an HTML
file and it'll just do it my god, that's a.

Speaker 1 (31:16):
That is such a great suggestion, man, I'm gonna check
that out sorry.

Speaker 2 (31:20):
Yeah, keep going no, no, no, yeah, I do this all the
time.
Uh, so, like, that's an examplethat, um, I used with my
brother, right, but it took.
No, I, I didn't look at thecode that it wrote, right, I
don't even know.
I don't know if I use anyframeworks.
I don't know what the code thatit wrote, right, I don't even
know.
I don't know if it used anyframeworks.
I don't know what the codelooked like.
I just said, hey, create theflashcards.
And the flashcards appeared,right, you save it as an HTML

(31:42):
file.
It's magic, it's magic, it'scrazy.
You save it as an HTML file.
And I was like, hey, justdouble click this, right, put
this on your desktop, doubleclick this, and it appears and
it works, right.
Another example is my wife and I, right, we were trying to like
what were we playing?
We were playing some likerelationship card game, right,

(32:03):
that was like asking deepquestions and I was like I bet I
can make one of these.
It's a little more tailored tous, right?
So I just gave it some context,I gave Claude some context
about me, some context about her, and I said, hey, generate like
a relationship card game wherewe can like.
It was another, yeah, it wasanother card game Demonstrate

(32:24):
excuse me, generate arelationship card game that's
tailored to us, that'll likebring us closer or something
like that, and it did it, and itdid it so well and the game was
actually fun, right, but butagain, for new people without a
coding background, that'ssomething you can do without
even having to look at the code.
Right, it's like thisinteresting new way of doing

(32:45):
things called vibe coding right,where you almost forget the
code is there.
Right, there's a carpathy tweetabout that and, uh, he was the
head I believe he's the head ofai andrew carpathy.
Yeah, right, right, and thisguy's like he's been doing, you
know, like language-based AI fora long time.
I remember reading his articlesin like 2014.
But, yeah, you almost forget.

(33:07):
The code is there, right?
You're literally just speakingor typing and the app appears
and then you say, hey, changethis thing right.
And then it tweaks it right,and if you're a non-technical
person, you can get pretty farwith just this, like html, uh,
with just html files andtweaking them through natural
language.
From there, you just copy it,save it in a text file, uh, and

(33:30):
we can, like I'll maybe leavesome instructions in the
comments, right, for how to dothat save it in a text file and
then just double click it andthe app you built appears in
your browser.
So I would say, start there,start with Cloud, because you
can even see it as you're typing, and just use language to kind
of sculpt this app Like the wayI described.
It is build by reacting.
I think that's a great way toview.

(33:51):
You know how to get into vibecoding if you don't have a
background.

Speaker 1 (33:57):
Totally, you, you know how to get into vibe coding
if you don't have a background.
Totally, this was.
This is a little off the cuffand I've never done this with
any of my guests, but I I didtake your suggestion and I
created this app this morningwith claude and I'd never used
claude before and I wanted toshare it with you real quick
yeah, absolutely so so this isthis is the game.
Uh, I just died.

(34:18):
Wow, it's good.
This is the game that I createdthis morning with one prompt.
The prompt is on the left handside and basically it's space
invaders with blocks of cheese,like I said, and I'm just
blasting them away.
And I do have lives.
I just lost one.

(34:40):
You know this.
I shaped in two prompts.
Right, I said I would like youto create a video game set in
the year 2121.
Make it like the game SpaceInvaders, but with one
noticeable difference theinvaders are blocks of cheese.
And he came up with thisconcept of Space Gutas Dairy

(35:00):
Invasion of 2121.
And it spit out all this codeand I was like I wonder if this
thing's going to work.
And then all of a sudden it'spresenting this UI and gives me
the storyline gameplay, in caseI wasn't familiar with how Space
Invaders works.
It's got all of these types ofinvaders with descriptions

(35:26):
Mind-boggling, honestly, to seethis could come out and be
perfect.
I mean, not like idealnecessarily, but it worked with
one shot.
And then, you know, I was in myfirst pass, I was losing a lot,
so I said, okay, what was mysecond prompt?

(35:50):
I said let's make it two timesas hard to kill the player.
I think the cheese shoots toooften and too accurately, and so
it slowed it down and itallowed me to win my levels.
Incredible.
That's where we are and,honestly, man, I published it

(36:11):
and I didn't know I couldpublish this and I copied the
link to the Discord channel forAI Ready RVA and I made it
available to anybody to play.
I don't know that anybody'sreacted to it just yet, but I'm
definitely looking forward totheir reactions.

Speaker 2 (36:26):
Yeah, it's so crazy Because, like, I think of like
how long this would take, like,take me to code up, right, if I
was trying my hardest, if I wastotally focused, had nothing
else to do, it would take awhile and I would get the
physics wrong, right, I wouldlike, I'd be like thinking of
like there'd be so much tomanage, right, you'd have to
design it.
And the fact that an AI couldjust spit this out, you know,

(36:47):
literally word by like, token bytoken, is it blows my mind.
Like things like this nevercease to amaze me, yeah.

Speaker 1 (36:53):
Yeah, Crazy.
It's awesome.
I'm going to continue workingwith Claude and see what else it
can do.
I love that you made vibecoding approachable with just a
few statements of you know,start small.
You know, break down your tasksand see what it can do.
I think that's, that's totallyapproachable advice.

Speaker 2 (37:15):
Yeah, yeah, absolutely.
I think it's also a good way toget into coding, like.
One thing that I never likedwas how everyone's saying like,
oh, it's not a good time tomajor in computer science or
learning to code is dead right,even though I use it to generate
a lot of code.
I totally disagree.
Right, because code is themedium between natural language

(37:38):
and everything.
Now, right, if something can becontrolled with code, now you
can control it with your voiceor by typing words in natural
language.
If you have any kind ofunderstanding of code, you're so
much more empowered to buildthings with AI.
Now, anything that you can hookup to any kind of code now is

(38:01):
AI controllable.
So, yeah, I think there's neverbeen a better time and I think
it's a great way to get started,absolutely.

Speaker 1 (38:09):
So what tools or platforms would you recommend
for non-technical folks eager toexperiment with vibe coding?

Speaker 2 (38:17):
Yeah, I would say definitely, Claude, just because
, again, it has that UI that wejust talked about and you can
kind of talk through things withthe webpage next to you, right.
So it's almost like you and thechatbot are looking at.
I think of it as sculpting in alot of ways, right.
It's like you and thisassistant are just looking at
the statue and you're saying, oh, the arm looks a little funny,

(38:38):
can you like take a little bitoff the shoulder right, and it
just chisels a little bit offthe shoulder.
I really think of it like that.
So definitely, claude.
If you want to get a little bitmore down and dirty, I would say
Cursor.
You know, i't know, cursor iswhat's called a text editor,

(38:59):
right, which is where youactually write the code out, and
it has AI integrated in really,really, really nice ways.
So everything is kind of rightthere.
And you know, if you want toextend you know the game you
built or the flashcards youbuilt you would dump that text
into cursor and then just chatover it there and it can kind of
make, make more edits.
So it's really good and thingsare.

(39:21):
You know the cursor team.
I think they went from like ahundred million annual recurring
revenue to 200 in like a monthor something like that,
something crazy.
And these guys are one of theone of my favorite companies
like they're.
They're really good.
So I would say follow what'shappening in that space with
Cursor, but those two tools willget you really far.

Speaker 1 (39:41):
Yeah, that was.
My first step into vibe codingwas with an IDE like Cursor, and
I use VS Code at work and I'veused various other IDEs.
But cursor was like VS code onsteroids, and when I started

(40:02):
mucking around with that thing Iran into the occasional error
Maybe I was asking for some codeupdates that I didn't have some
of the packages built into mysystem already and so I would
have to troubleshoot my waythrough that.
But it was a profound shift forme.

(40:23):
Just experimenting with that, Idid get some of the most basic
apps working.
I don't know that it's readyfor something as easily
accessible as what I just didwith Claude this morning.
Right, but you're right, if youwant to do some enhancements

(40:44):
and get your hands in the codeand really leverage different
libraries and be moreintentional as a software
developer, you'd have to getinto the IDE, and that's where
the real power can happen.
I think, yeah, absolutely.
So what common hurdles do youthink non-technical individuals

(41:06):
face when diving into vibecoding and how can they overcome
them?

Speaker 2 (41:11):
Yeah, so I think you know there are a few different
kinds of ways that you can build.
You know, when it comes to vibecoding, one is just building
like kind of like one off appsthat you might send to a few
people and that are useful,right?
I think, honestly, there'sgoing to be a lot of folks in
the future who are buildinglittle tools for their teams.
They just need to share them.
I think that's going to be alot of code now and you, as an

(41:32):
EM, know it would be insane tobe like, hey, can you build a
tool for us that we need latertoday?
Like that would never happennowadays.
But I think with the way thingsare going, it's going to right.
So, in terms of hurdles, Ithink that the hurdles that
you'll start seeing really occurwhen you start building things

(41:52):
that are closer to being usedfor like mass adoption.
Right, and obviously you wouldnever vibe code something like
into production, right, you wantto get like real engineers to
handle like the just thenecessary, like plumbing that
comes with, like buildingsomething that's in production.
But I think there is a middleground, right, and I find myself
building a lot of likeprototypes and MVPs, both for

(42:14):
companies or just to like showfolks, or just for like to show
my, like co founder, like andlike my team, and sometimes they
incorporate expectedfunctionality, right, it's kind
of what I call it right, wherepeople want to be able to log in
, right, people want to be ableto change their profile.
People need you might want tohave billing enabled, right, you

(42:37):
might want to have you knowauthorization or with a data
like database with authorizationright, and those things are not
something that you want to becoding yourself, right, like you
do not want to code up and likeI learned this the hard way you
don't want to code up like anemail registration flow right,
it sucks, no one wants to do itright?
So I think a huge hurdle fornon-technical folks is when they

(43:00):
encounter that, when theyencounter having to build some
kind of expected functionalityand a lot of them will be like,
okay, I'm already vibe coding,why don't I just vibe code this?
Right?
And I would say, don't, do notvibe code these like standard
things, right, yeah, so I wouldsay, like, the way to overcome
that hurdle of these expectedthings that everyone expects to
be fast and smooth is to useframeworks and templates as much

(43:24):
as you can when you get to likethe bigger stuff.
So there are JavaScriptframeworks.
One is nextjs.
I like viewjs, so thumbs up,nice, I like viewjs, which is I
think yeah, it's called Nuxtwith a framework, and I would
say leverage those as much asyou can, because they have a lot

(43:45):
of that stuff built in Rightand on top of that, the AI knows
how to incorporate it Right.
The AI knows Nextjs, it knowsNuxtjs.
Going another level higher, canwe just like?

Speaker 1 (43:55):
yeah, go ahead going another level higher.
Use a framework.
Can we just talk through thatone a little bit?
So giving a prompt, a naturallanguage prompt, to the AI and
say go build this Space Invadersclone is one thing, but how do
you incorporate Nextjs into aprompting system like that?
Can you give me an example?

Speaker 2 (44:15):
Yeah, and this takes a little bit more experience and
it might be a little bit downthe road for a lot of
non-technical folks, buteventually you want to learn how
to kind of spin up like thesefull stack apps, right.
So this is where I think thevibe coding gets fun right,
Because you can build suchcomplete systems really quickly.

(44:36):
So I would encourage anyone togo to like just Google Nextjs
and you'll see what I'm talkingabout, right.
But it's a framework with kindof like that's kind of batteries
built in.
So what you would essentiallydo is you would start a project
with Next and then you'd open itup in Cursor right and then you

(44:57):
talk to the chat in Cursorright and then you talk to the
chat in cursor right and it hasaccess to all the files in there
.
It knows how to use theframework and you can get a lot
built out really quickly.
So I'd even take it a stepfurther than that and say you
know what?
Don't start from scratch withthese frameworks.
Use a template, right andtemplates.

(45:18):
You know, if you'd asked me,like three years ago, I would
have said I would have been liketoo proud to use a template.
I'd be like, no, I'm gonna codeit up myself, right, but, uh,
the templates have a lot offunctionality already built for
you and you can test them right.
As an example, like a lot ofapps like have like a sidebar
right where you can click likethe different views you want to

(45:39):
view yep, uh, you need that tobe smooth, you need the routing
to work, you need the URL tochange correctly, Right, and
that needs to feel smooth ifyou're going to have other
people test it out, Right.
And I would argue that if youstart with a template where it
already feels smooth and thenjust use, you know, kind of open

(46:00):
that in cursor and kind of makeyour changes to make it your
app, that's a lot easier and alot.
You'll get a much more stableproduct than if you were to just
vibe code the whole thing fromscratch.
So if I had to kind of like,take a step back and, you know,
give a high level view of whatI'm saying, I'm basically saying
you know, stand on theshoulders of giants, right,

(46:21):
Don't try to vibe codeeverything from scratch, just so
you can say it's your own right.
I used to have this complex inmy head about that.
Even if you're non-technical,try to find projects that are
close to what you're doing, butjust need a little bit of
customization and start thereand that's where these templates
come in.
So, yeah, I would say againstand on the shoulders of giants
, leverage the work that'salready been done and use

(46:44):
templates and kind of edit those.
So it's a little bit down theroad, right.
A lot of you can get a lot donejust vibe coding from scratch
in Claude, right.
But if you start building likeprototypes of your apps which
you probably will if you getgood at vibe coding you apps
which you probably will if youget good- at vibe coding, you're

(47:04):
going to start thinking hey, Ican build this app right.

Speaker 1 (47:05):
Start with templates, where a lot of the flows
already optimized for you.
Yeah, totally so.
Don't fight the complexity,right?
Yeah, learn from others.
At my office, we always saydon't reinvent the wheel, right?
So if the scaffolding alreadyexists, use it so that you can
focus on creativity, right?
Exactly, make your ideadifferent than what the template

(47:26):
was and better, right?

Speaker 2 (47:29):
I think that makes a lot of sense.

Speaker 1 (47:30):
Great advice, okay.
So looking ahead, how do yousee vibe coding evolving,
especially in making technologycreation more inclusive?

Speaker 2 (47:43):
Yeah, I love this question because this is
something I wrote another Mediumarticle about earlier, where I
predicted that data analysiswould happen in natural language
at some point.
Right, I didn't know, it wouldbe like 2025.
I thought it was going to belike 2035, to be honest, but it
happened so quickly.
So, to answer your question, Ithink at some point the AI will

(48:07):
become so reliable that evenpeople with a technical
background will just stoplooking at the code, kind of
like, jason, like the app thatyou built, right.
Like you say, hey, build thisfeature for me or build this
thing for me, and you're not.
Maybe you looked at the code, Idon't know, right, but you're
looking at the final product,right.
You're typing out what you wantand you're looking at the

(48:28):
product and you're reacting tothat, and then you're going back
to the natural language prompt,right, and you're tweaking it,
then looking back at the finalproduct.
In the middle of that is code,but you're confident that that
code is correct, so you don'thave to look at it that much.
I think that that will be howvibe coding evolves for the

(48:49):
better, and that's going to makeit more inclusive, where a
person who doesn't have atechnical background.
Who doesn't necessarily speakJavaScript?
They can say, hey.
Who doesn't necessarily, youknow, speak JavaScript, they can
say, hey, you know, I need toreview a bunch of tax forms.
Right, just build me a UI whereI can quickly skim through this
section.
For each one, and I hit theright arrow key and the next one
pops up, and if I hit the spacebar then it deletes something.

(49:12):
Right?
That kind of useful vibe codingsoftware, I think, is going to
be able to be built reallyquickly and I think that you
won't have to worry about thecode, right?
So that's going to reallyreally make adoption much faster
.
When I say adoption, I don'tjust mean of chatbots, I mean
the adoption of peopleinteracting with code, people
building things, right.

(49:33):
So, yeah, I think you know, thefirst thing that's going to
change is that the actual codewill just kind of fall away,
just like a lot of thecomputation that happens on our
like.
Right now we're both usinglaptops, right, but you don't
have to think about anythingmemory related, you don't think
about any computation, right?
It's just happening, I think.
Secondly, there will be a lotmore and I don't know the best

(49:54):
way to say this, but there'll bea lot more forms of creating
that will also start to occur innatural language.
So data analysis is an example Ijust used where, let's say,
you're investigating bank fraudright, you have some system
that's hooked up to a hugedatabase, right, and maybe
you're just talking, you know,maybe you're saying, okay, I'm

(50:15):
suspicious about you, know TimSmith's transactions last month,
show me the last 10transactions.
Then on a screen, thetransactions show up, right.
Or maybe you're wearing a VRheadset, I don't know, make this
even more futuristic.
And then you're like, okay, ofthese transactions, how many
were outside the US?
And then those pop up.
Or the answer pops up and yousay, okay, show me who they were
to, right, and then you see,like a few different names, how

(50:36):
many of them were over $100,000?
And then you know, you see abunch of transactions and one
name pops up and the analyst islike, ah, found it right, I
found the fraud.
I think that kind of thing isgoing to become much more common
, as right now, again, there'sthis intermediate code step
right, where I have to takemyself out of my investigation

(50:56):
mode and write code, right,that's what my job was when I
was working on integrity, right,I would be.
What I wanted to do was kind ofbe an investigator, right, was
to be asking questions, kind ofinterrogating the data, right.
But I would have to take myselfout of that loop and use my
brain to write code which kindof completely snaps you out of

(51:16):
anything else you're doing.
So I think that you know, forthis data analysis example, you
can stay in the investigativemode, right, the medium is just
natural language which we'reused to.
So the code will kind of fallaway and it'll become much more,
much more easy to just do thething you want to do, right?
I think that that's that kind ofthing, is going to extend to
many different domains that Iprobably can't even think of now

(51:38):
.
Right, I'd imagine like wearinga VR headset.
Let's say I'm an interiordesigner, you walk into an empty
house, you're wearing a headsetand you say you know, style
this with like art deco.
Give this an art deco theme,put the couch right there.
Okay, move the couch over here.
Actually, let's imagine thatthere's a house party happening

(52:21):
right now and all the peopleappear around you Right, and
these things will be happeningagain On the back end.

Speaker 1 (52:23):
There's code being written, there's some kind of
config or something beinggenerated that's changing the
thing for you, but you are justthinking in natural language and
reacting to what you see, and Ithink that pattern is going to
really expand to lots ofdifferent creative endeavors and
forms of building.
Wow, I believe it.
An hour ago I probably wouldhave been way more skeptical,
but the way you described it, Itotally believe it, and it's not
about obsessing over the codeor learning to code really
anymore, although you did saysomething that I believe to be
true we need coders, we needpeople that understand the code.

(52:45):
You have to be able to elevateyour vibed output to something
extensible, somethingproduction-ready, scalable all
of that.
That's super important forrunning businesses with software
.
That has to still exist, and Idon't think AI is there yet.
I am cautiously saying thosewords, but here we're all in the

(53:12):
vibe coding setting and lettingour curiosity take the hard
parts out of the work, which isamazing.

Speaker 2 (53:22):
Yeah, it is, man.
It's a crazy time to be alive.

Speaker 1 (53:24):
Yes, it is.
I'm excited, man.
All right, so what's one pieceof advice you'd give to someone
working inside a big companytoday who dreams of launching
their own AI-powered startup?

Speaker 2 (53:38):
Yeah, that's a good question.
I'm going to give you three,because there's just so many
things I wish people had told mewhen I was starting this.
The first is to focus on yeahyeah, the first is focus on
problems, not solutions.
Ai is a solution to manyproblems, but you will get much
further by focusing deeply on aproblem, preferably one that you

(54:00):
experience yourself, right,which is kind of how I started
Aura.
It requires a mentality shift,right?
You know, before I was afounder, I would kind of shy
away from people who weretalking about their problems.
I'm like, okay, here we go,right, let's be positive.
Right, but you want to startbeing obsessed with that, like,
when someone starts talkingabout their problem and you're a
founder, you should kind ofhone in on that, you know, and

(54:21):
be like, yeah, well, actually,what's going wrong?
Right, a lot of time you'llfind out that they're like man,
I have all these customerservice interviews that happened
and I can't, I don't haveenough time to process them.
You know, my boss is pressuringme to hit these KPIs and I just
can't seem to get it done.
As a founder, you're like, boom,that's my startup idea, right,

(54:41):
you go home, you vibe code out aprototype, right, which is
exactly what I've done before.
You vibe code a prototype.
You hit that person up like,hey, you wanna check this out
really quickly.
Maybe you send them thepublished quad link, right, and
you're like, would somethinglike this help you out?
And if they say, oh my God, yes, I can't, oh, this would be a
godsend, right, you havevalidated that idea, right, you

(55:02):
validated that someone will payyou for this, right, because you
just solved their problem.
So I would say, firstly, focuson problems and kind of match
them to solutions.
It's a mentality shift that iskind of counterintuitive, but it
will get you really far.
Like, the best startup foundersare obsessed with problems.
They look for problems, theyjoin, they go networking, just

(55:24):
so they can hear people'sproblems.
So that's number one.
Number two leverage AI to thehilt, leverage it as much as you
can, but also leverage expertsin your network.
So I have a rule that ifthere's a concept or something
that I just cannot figure out orsomething I can't crack, I
reach out to somebody on Upworkor I reach out to through my

(55:44):
network, Right?
A good example is when Istarted, I was using, like the
raw Azure cloud, right, just tolike launch my prototypes.
It's a pain.
I mean, you know how, you knowhow cloud goes, right, like they
don't try to make it easy foryou, and I was like, okay, I
don't know how to do this, so Ireached out to know everything
in order to do anything.
Yeah, it's a, it's a, it's athing, man, it's like, uh,

(56:06):
there's a lot of coordinationnecessary.
And I reached out to a cloudexpert and he was like he's like
, you know, honestly, like youcan, if you want to, you can go
this route, but it seems likeyou're more product focused and
you just want to launch.
Why don't you use somethinglike Vercel with Supabase as the
backend?
Those are two products forthose who don't know, and it was
the best advice I'd ever gotten.

(56:27):
Honestly, it saved me so muchtime and the AI, for whatever
reason, couldn't pick up on that.
But this human still had someadvantage that the AI didn't and
that he could kind of see whereI was right and be like hey,
this is probably a better pathfor you and it was some of the
best advice I've ever gotten.
So always remember that AIcan't do everything and when

(56:50):
you're really stuck, reach outto humans.
It's a great way to network, ifnothing else.
And lastly, is to just get usedto uncertainty, and this is the
trickiest one, because it feelslike.
It feels like you're doingsomething wrong.
If you were good at your job,right, if you're good at your
job, you know people bring youin when they need certainty, and

(57:10):
you usually have seen theproblem before and you're like,
oh yeah, just do this right.
There are only so many thingsthat can be different and you
just have this like greatpattern recognition for it.
When you're a founder, thoseare the exact kinds of things
that you need to be payingsomeone to do.
If that makes sense, right.
If there's an established wayto do it and you feel

(57:31):
comfortable doing it, it feelsfamiliar.
That's not what you need to beworking on, right.
You need to be kind of bringingnew things into existence and
dealing with problems wherethere isn't certainty yet
because you're doing somethingnew, you're building something
new.
So I would say, getting used tothat and understanding that
it's not a bad thing,understanding that this constant
, you know, not knowing exactlyhow something's going to pan out

(57:55):
is a sign that you're doingsomething worthwhile, right.
It's not a sign that you'redumb.
It's not a sign that you're badat your job.
It's a sign that you're kind ofon the right track, right.
And once you get comfortable,once you're like, okay, I got
this, you know this, this, thispart of it makes sense, that's
when you need to hire somebodylet them handle it Right and you

(58:23):
kind of like drive on, you know, and find the, find new
opportunities.

Speaker 1 (58:24):
So I know that's a lot, but these are things that I
wish somebody would have toldme much earlier in my career.
That's beautiful man, thatthat's, that's gold.
So fall in love with the, withthe problem, not the tool.
Right, right, and I love themindset If it feels routine and
certain, maybe it's time todelegate.
Yeah, right, yeah, yeahabsolutely.
I love the uncertainty pieces.
That's where the real humaninstinct kicks in.
Yeah, and the creative side ofus can thrive and build things

(58:48):
that no one's ever thought ofbefore.

Speaker 2 (58:51):
Absolutely and also like it kind of brings out a
side of you that you might notknow was in there.
This is getting a little wooright, but like it's um, you
know when you're again, whenyou're doing work that feels
that you're, that you're used to, right, you don't have to be
like daring, really, you don'treally have to be.
You're gonna be creative, butnot to the extent where, like,

(59:15):
you would be creative if yougenuinely don't know something's
gonna work out right.
If you want something to workout and you're not sure, you
turn on a part of your brainthat's almost like a bit of a
maverick right, a bit of acowboy.
You know where you're, likeokay, I got to figure this out
right.
And you approach it confidently.
You really don't know if it'sgoing to work out and it turns
something on inside of you and Ithink that a lot of people who

(59:37):
experience that like who itturns them into.
And I think that's what keeps alot of people as founders when
you know, as you know, like ifyou're, if you have skills to
work in tech, you can make a lotof money Right, but to to let
that go to do something that'suncertain, I think the reason a
lot of founders do it is becauseof who it, of what it brings

(59:58):
out of them.
I think the reason a lot offounders do it is because of who
it, of what it brings out ofthem.
Right, it really lets you.
It lets you see what you'relike when you're facing a real
uncertainty, when you reallyhave to, just like when there's
a lot on the line and you likeyou got to roll a six, you know,
and you got to figure it out.
So for me, that's kind ofwhat's keeping me keeping me
here, cause, like again, as youknow, like tech is tech pays
really well.

(01:00:18):
But I think that that, uh, abit that adventurous side that
it brings out of you is actuallyit's a lot of fun, love it so,
kyle, where can people followyour journey or get involved
with what you're building next?
yeah, so I think the best placeis linkedin.
I do a lot of posting there.
I try to post a little morevideo now.
Uh, I have a tiktok too.

(01:00:38):
I'll tell you about like it's alinkedincom.
Slash in slash.
Gkjohns.
Like GK Johns.
My Twitter is xcom slash KyleD-O-T-A-I.
Like Kyleai, but spell out thedot.
And on TikTok, I'm Kyleai likeK-Y-L-E-D-O-T-A-I.

Speaker 1 (01:00:59):
Awesome, All right man Surprise question, but if
you follow my podcast you'llknow it's coming.
If you had any superpower, Kyle, what would it be and why?

Speaker 2 (01:01:14):
That's a good question.
I would say complete control.
This could get very deep.
It's funny because I was goingto say mind reading.
I actually thought of this.
I was going to say mind reading, right, but I was like I don't

(01:01:35):
know if I could.
You know, these like AI as athought partner.
With using AI as a thoughtpartner, I've really, really,
really had a lot of fun tryingto understand like complex
social situations that I'm in,right.
So an example might be you know,I'm working with a company and
you know something that I thinksomething fishy is happening.

(01:01:57):
You know, and here's an emailthread from the CEO and here's a
post that this person made andhere's some context about what's
going on with the stock priceand here's this weird thing that
my report said to me, right,and I have a lot of fun like
dumping it all into chat, gpt ina project and just talking
through it and trying to figureout what's going on, trying to

(01:02:17):
get clarity on the situation, ifnothing else, just because I
think it's a fun thing to do.
So I think if I could have asuperpower, some kind of like
perfect social understanding,where I can immediately kind of
clock what's happening and knowwhat people are thinking.
I think that'd be a lot of fun,because I've definitely had a
blast, you know, using ChatGPTas my like gossip partner, yeah,
so yeah, I think it woulddefinitely be some kind of like

(01:02:40):
instant mind reading, socialunderstanding.
I walk into a room and I knowexactly what's going on between
everybody.

Speaker 1 (01:02:47):
So emotional intelligence basically you know
beyond the self awareness, butsocial awareness and being able
to read the room without gettingtripped up.

Speaker 2 (01:02:58):
Yeah, that's really cool.
I think I'd have the lamestsuperhero costume, like
emotional intelligence man.
Probably isn't the coolestsuperhero, but I think it'd be
fun.

Speaker 1 (01:03:10):
Yes, awesome, awesome .
This has been so much fun, kyle.
I know the audience is going tolove what you've put down here
and there's so much to dig into,so many nuggets of wisdom here.
I really appreciate your timeand hope you have a wonderful
rest of your day and hope to seeyou on the podcast again soon.
Yes, sir.

Speaker 2 (01:03:30):
Absolutely, man.
Thanks for having me on.
I really this was fun.
I really appreciate it.
I like this format.
This is good, awesome, well,thank you, awesome Cheers, man
format.

Speaker 1 (01:03:42):
This is good, awesome , well, thank you Awesome.
Cheers, man, and thanks to ourlisteners for tuning in today.
If you or your company wouldlike to be featured in the
Inspire AI Richmond episode,please drop us a message.
Don't forget to like, share orfollow our content and stay up
to date on the latest events forAI Ready RVA.
Thank you again and see younext time.