Building AI Products in Public

I've loved this conversation because I think it speaks to the moment in which builders and folks with a CS background do want to learn like content and marketing which is your background and

then folks who are newer to vibe coding and coming from other spaces maybe who are great at like communicating and marketing we're bringing them into like that engineering background. So it's

it's been a really interesting bridge for me. Yeah. and um I think you're such

for me. Yeah. and um I think you're such a great example of someone who's really personified communicating out building in public really well.

>> Thank you. Yeah, I kind of see myself as the the bridge that connects technology with humans.

>> So, it's actually a two-sided relationship. So, on the one hand, how

relationship. So, on the one hand, how do you communicate the benefits of technology to um a large audience and make them feel like, oh, I I should care about this or it's relevant to me. on

the other hand, how do you take the pain points of a large number of people and then translate that into like a product or uh the techn underlying technology?

So, it's a two-way relationship and I really enjoy kind of uh being a storyteller on both sides. I think uh in the AI age um there is a lot of power to

being at the intersection of things. So

uh if you are a researcher or engineer and you also write really well that's a huge advantage and then if you're a writer artist or a marketer and but you also can vibe code then that's also a

huge advantage. So I would I kind of

huge advantage. So I would I kind of encourage everyone to step a little bit outside your comfort zone. Um because AI has made that a lot less uncomfortable than before. It's actually everyone's

than before. It's actually everyone's superpower if you want to like learn something new. Yeah. Do you have any

something new. Yeah. Do you have any mindset strategies just around consistency or people who are building the muscle to put themselves out there for the first time?

>> Yeah, I think people worry too much about problems that only become problems once you succeed >> because a lot of people say, "Oh, like what if I get all this negative

feedback? What if everyone laughs at

feedback? What if everyone laughs at me?" I'm like, "Nobody cares." Like

me?" I'm like, "Nobody cares." Like

nobody cares about you before you get attention. So the first step is to get

attention. So the first step is to get attention and then the good kind obviously. So um and then people always

obviously. So um and then people always estimate how much other people care about them. Um people always assume that

about them. Um people always assume that they will got go viral somehow and then all the consequences that come from going viral but people don't realize how hard it is to go viral these days. Um,

so >> and just to emphasize like you go viral all the time. Like if you are a builder, um, you are the person to watch in the like content builder space.

>> And I also think it it's not really about hacks or techniques. It's really

about becoming the kind of person that goes viral or doing the kind of thing that goes viral. M

>> um I think my content goes viral because um I'm a non-coder turned vibe coder sharing my personal learning experience in public and sharing my creations and

that resonates with a lot of people at this time time and age because a lot of people are trying to participate in the AI revolution but don't know how especially people who don't come from a

traditional engineering background. So

kind of they see in me um maybe a path that they can follow to participate in AI and that's the the reason that's the deeper reason. It's not about uh the

deeper reason. It's not about uh the headlines I write or like the words I use or the polish. I think people need to realize that there are two parts to content. There is the substance and then

content. There is the substance and then there's the style. And a lot of people spend too much time on the style. So

they worry about oh is are my words well written or should I use AI to like polish it that like doesn't matter.

>> I think what matters more is the substance. It's like is what you're

substance. It's like is what you're saying something people actually care about today? Does it speak to the times?

about today? Does it speak to the times?

Does it >> um is it meeting the moment?

>> Um and I think it needs to be a van diagram of like what you want to say, what your audience is, what people are already interested in and what's good

for you. Um so um I think finding that

for you. Um so um I think finding that positioning takes time. Uh which is why consistency is super important because a lot of people give up even before they

find their positioning. Um because uh positioning comes from repeated like just putting in a lot of reps and then getting feedback um getting engagement data and seeing what people are saying

and then slowly tweaking it. I didn't

find my uh positioning on day one as well. it I just in the beginning I

well. it I just in the beginning I literally my only goal was to post minimum three times a day. Um had no requirement for what to post about just like literally whatever is on my mind.

And then every time I build something I always record a screen recording and then put it out there.

>> And then another thing I noticed is a lot of people um assume that you need to build a super polished product before

releasing it to the world. Um but theity the reality is a lot of times you you can just share a work in progress and that works as well. Um especially if it's a side project like not your huge startup idea.

>> So a lot of times I literally just like screenshot a prompt or like a demo and then it like it's not even shipped. It's

it's not deployed. It's just on my computer but then it can still go viral because the idea um is interesting to people. So I think we don't have to

people. So I think we don't have to always expect perfection but like overcoming that mental barrier is all about >> um assuming that nobody cares about you

by default and that is very hard to get attention and then just u be authentic be yourself and then seeing it as a way for um self-expression.

>> Can you talk a little bit about what makes content resonate? Mhm. Um I think like I mentioned before content has two

parts the style and the substance. Um a

lot of people overfocus on the style. So

when we think about AI generated content is very polished and when AI is writing something the words are very flowery the language is very flowery. Um but that

doesn't mean that's good content. So

good content does not equal polish. M

>> um and I actually think it's going in the opposite direction these days. I

tend to feel like the content that resonate with me >> are raw >> are unpolished or imperfect because human are raw and imperfect.

>> So we resonate with uh things that resonate with us are things that we see ourselves um in them. So for example, we might see a funny video and then

laugh at it. The reason we're laughing is because it reminded me of something in my life.

>> That that's why I'm laughing. And then

if I see a super touching story, I might cry. The reason I cry is because it

cry. The reason I cry is because it reminded me of something sad happening to me before.

>> So humans love themselves.

>> So good content reminds humans of themselves. And then that's the problem

themselves. And then that's the problem with most AI generated content is they're too distant from our lives that I never see myself in that setting. even

though it's perfect, it's very polish, it's very beautiful, it's not relatable to me. So, so in that sense, how do we

to me. So, so in that sense, how do we create content that's relatable to people and remind people of themselves?

Um, and I think the answer lies in you have to put yourself out there.

>> Um, you have to inject your own soul into the content and not care too much about the polish. So, for example, when I'm creating videos, I literally just record Loom videos on my screen. um I

don't care about my background or my lighting or like whether I have makeup whatever like because that's not the core of the the core of the content is the idea I'm sharing or the product that

I'm demonstrating and so that's so always remember the most important thing is the core >> and so if that's the case then um the

issue becomes how do we uh let our authenticity let our authenticity shine through in our content how do we be ourselves and still because I I think a

lot of people feel vulnerable putting themselves out there. Um so I think it's really about you don't have to talk about everything. You can be very

about everything. You can be very strategic and selective about what part of yourself you want to put out there.

And so you can be both authentic and it's good for you >> and protective of yourself.

>> So for example, I built I post a lot on Twitter, but there are certain topics that I never touch. um I never talk about my personal life because that's not an aspect of my um life that I want

to be in public.

>> So building in public does not mean you have to share your trade secrets.

>> That's so important. I think it confuses people what it means to build in public and I think you see a lot of founders now putting everything out there.

There's the cult of personality. Um but

yeah, go on on how you like pick and choose. Um, I think uh you do have to be

choose. Um, I think uh you do have to be because there are literally a million things you could be talking about because we all have different identities. Like for example, um I have

identities. Like for example, um I have an identity in my family. I have an identity for my hobbies. I have an identity for vibe coding, for my work.

So you can actually hand like select selectively pick what you want to say and then which identity you want to be on uh in public. Um, and I I think it's

really depends on what your goals are.

Like for example, um, also what you genuinely want to share. Like for me, the past six months have just been a super

shocking experience of like be going from someone not technical to like someone who's like shipping stuff.

>> Yeah. And 7,000 stars on GitHub >> and then people are actually using my product. So that's super serial for me.

product. So that's super serial for me.

So for me that that comes from a genuine place of hey this is really cool like I can't wait I just really want to share this with you. Um so if that's something hap so if if something that happened to

you happens to be something that the world cares about then great like that's obviously something you can share and then it doesn't have to be super fancy.

It just needs to be really real.

>> Yeah.

>> Something that strikes me about you is you are so prolific with the building in public and the ideas that you put out there. What advice would you give um to

there. What advice would you give um to maybe folks that are closer to your background in marketing and content on how to build in public and do it in such a prolific way?

>> Yeah, I think the key to being prolific is your content needs to be a natural extension of what you're already doing on a daily basis. It needs to be a natural extension of your work or your

life or your hobby. Otherwise, you're

not going to do it every day. um you

can't trust yourself to carve out like an hour every day to do content. That's

just not realistic for most like working professionals. So, and it has to be

professionals. So, and it has to be something you're already doing. So,

let's say at work you're already thinking a lot about integrating AI into your job or you're trying to learn VIP coding, you're trying to uh watch all these tutorials to learn how to use

these tools, then your content can be a natural extension of that effort. So

every time you learn something new, screenshot it, share with people. Every

time you build something cool, it doesn't have to be perfect. Can just be a demo, share it. So make it super low effort and super low barrier is um key to consistency. Um I think just don't

to consistency. Um I think just don't overthink. I think everybody overthinks.

overthink. I think everybody overthinks.

Everybody overestimates how much traction it will get. Like most likely nobody will care and sometimes some people will care and then building an audience takes a very long time. A

minimum a few months. Wait, walk us through like from when you started to your audience what your path looked like.

>> Yeah, so I started posting consistently on Twitter uh mid last year. So maybe

like eight months ago. At that time I had maybe 2,000 followers on Twitter and I hadn't posted on Twitter for eight years.

>> Wow.

>> Because before that I felt like Twitter was a lot of like clickbait. like I

didn't >> just just didn't open Twitter for a long time and then I realized all the AI stuff is happening on Twitter and I want to be part of the AI conversation. I

have to be on Twitter. Um so the first thing I did which is very important is to before you even start posting um start curating your timeline.

>> Uh what I mean by that is like because I had neglect neglected my account for a number of years. my account was full of topics I didn't care about anymore or um

people who were just trying to get clicks and like it's just not very um the content quality was not very good and then I noticed that uh if I want to

actually learn about AI I should follow builders not influencers like not pure influencers but people who are builders and have an audience at the same time like people like Andre Karpazi for

example people who actually share tips and um uh sharing what they're building, the learning journeys etc. um so I started following um people who are

founders who are uh working in the research labs who are putting out good stuff as like developers or just people who are sharing genuine use useful

things instead of uh purely trying to get clicks.

>> And I actively unfollowed a lot of the people uh who uh I didn't want to see their content on my timeline. And then

when I saw that type of content like oh ch is dead or like you know that breaking >> I would um proactively click not interested.

>> So telling the algorithm hey don't don't give me that anymore. Um after a maybe two weeks of like intentional curation my timeline became very clean. Like

right now my my Twitter is B b b b b b b b b b b b b b b b b b b b bas basically a AI learning platform is literally all like super useful like product launches or promps or like new technologies

coming out is very very useful. Um and

so when I see that kind of stuff on my timeline, it also inspires me to create um similar content. It also gives me new new ideas. And then when you engage with

new ideas. And then when you engage with that kind of people um and genuinely engage with them, they're also more likely to engage back and then that's uh and then they're more likely to see your

content on their timeline going forward and that's how you grow. So for it took maybe like three to four months of consistent posting before seeing any

result. And I think in the beginning

result. And I think in the beginning it's important to not just post but also engage with other people but but like don't use those automated AI engagement bots actually genuinely engage because

people can tell if you're being um authentic or not. And then when you're actually being authentic they they return your authenticity if they if someone likes or engages with

your comment on their timeline. um the

algorithm will more likely to push your future content to their timeline and then they they have a like more a higher probability of retweeting your next tweet for example and uh that's a lot of

that's where a lot of the growth happens and then also just like >> I think on Twitter um anything that is not text does really well so videos

screen recordings um GIFs images so I post a lot of screenshots like for example if you read a super interesting artic article. Literally just like

artic article. Literally just like select the quote you like and then screenshot it. That's that's that's a

screenshot it. That's that's that's a good tweet as well. Yeah.

>> Cool. And so, uh, you mentioned earlier posting like three times a day and you started a year and a half ago and now you're at like I don't want to misstate it, but like 40,000 followers or

something. Yeah. So, you just that's an

something. Yeah. So, you just that's an amazing like growth path. Um,

>> yeah. any any other last things to share for someone who's maybe in the 2,000 follower seat that's looking at you as like okay the year and a half I want to be like where she is.

>> Yeah, I think um one just uh have just have the discipline of posting consistently even without positive feedback in the

beginning is really important. just just

stick with it and then uh >> look at the data like what what are people engaging with and then once you pick out some patterns that people engage with do more of that and then

don't be scared of repeating yourself because I think content is a huge repetition game if you said something that went viral it will go viral the second time you say it the third time

you say it will go viral over and over again >> so you can have the same the uh you can have different variations of the same idea you don't have to come up constantly come up with fresh ideas.

>> Okay. Yeah, that is not intuitive. So to

just uh hone in on what works?

>> Yeah, exactly.

>> Cool. So I know a lot of folks that come to you are maybe new to vibe coding and looking to learn AI. What have you learned or seen in that space?

>> Um yeah, the b the past six months have been kind of a wild journey for me because I had literally zero coding background. Um before I didn't study

background. Um before I didn't study computer science in school. I worked

mostly in marketing and product and uh I started trying to get my hands into on web coding uh more than a year ago. Um

at that time the models were not that good. Um and um I I tried to turn my

good. Um and um I I tried to turn my ideas into rally but often failed. Um

and at that time my conception of vibe coding is it's a prototyping tool >> for me to maybe make a prototype and then send to an engineer and then they can code it. I was I have to have an

engineer to make a good product.

>> Um and I also got a lot of public I I got a lot of uh negative feedback like a lot of frustrations from just I just ran into all these bugs that I couldn't

solve and I was you know a lot of the loops uh with AI that I couldn't solve.

So I I it often felt like a skill issue when you couldn't solve something. And

then uh my first turning point was I think end of last year when Gemini 3 came out and then I started um using Google AI studio a lot more to build a

lot of the front-end interaction demos.

Uh a lot of them integrated with uh voice chat or text to speech and the live APIs and all that. Um because a lot of time a lot of the things that tripped

me up before that was integrating the various APIs like the plumbing part.

>> It wasn't even so much about model intelligence. It's like all the under

intelligence. It's like all the under how do you connect this thing to that thing like it was very hard for non-coders. Um I think Google ad studio

non-coders. Um I think Google ad studio did a good job of bringing them like all the the whole family of Google products into one product. So they solved the plumbing and then that's when I realized

oh wow I can basically describe an idea and it'll actually work. like before it didn't work, now it actually worked. But

at that point, it only worked for me because uh going from a demo to a deployed product was still a barrier for me. Um I didn't know how to make it

me. Um I didn't know how to make it scalable to a lot of different users.

Like I could only make it work on my laptop. But even at that point, my

laptop. But even at that point, my screen recordings were doing very well >> on Twitter, but I didn't know how to ship it. And then uh came clock which I

ship it. And then uh came clock which I think was like a huge milestone in AI coding completely reshaped um how it's done

>> because before claw code I feel like all the AI coding products were more like tools like they help me um write code that's as a tool. Uh when I started

using claw code I I genuinely felt like I hired a senior engineer who is always available to help me always enthusiastic

to build my ideas uh never judging and then on top of that also teaching me a lot about how coding works how technology works. So for example you

technology works. So for example you want to uh have people pay for your product and you you know you need to use stripe for it but you have no idea like how to integrate it. literally just tell

Claw, "Hey, I want to use Stripe. Like,

tell me what to do." So, a lot of times I feel like it's like I'm I'm executing Claw's commands. It's not the other way

Claw's commands. It's not the other way around.

>> I'm literally just a copy and paste machine for API keys for cloud.

>> Um, but it's a very liberating feeling >> because when uh a lot of people are like me, they have a lot of ideas.

>> Uh, but they didn't know how to turn them into reality. In the past, you would have h had to hire an engineer or uh raise money, build a team, etc. Now, you can literally just like get a claw

max account and then that's your engineer. Um that's probably already

engineer. Um that's probably already better than most um junior engineers.

And um it's also I think also because it's not a human engineer, I feel very safe and secure when work working with it. I'm never afraid of asking stupid

it. I'm never afraid of asking stupid questions or um feeling like my idea may like getting judged, whatever. Like

there's no kind of emotional barrier to building anymore, >> which makes it it's like a pure intellectual exploration. It's like what

intellectual exploration. It's like what if I did this, what I what if I did that. So

that. So >> I'm doing a lot of fun experiments with it. A lot of times I feel like I'm just

it. A lot of times I feel like I'm just playing like it doesn't feel like super serious work. And I um there's a quote I

serious work. And I um there's a quote I really like from Peter Steinberger from open uh the creator of Open Claw uh when Lex Freeman asked him, "There's so many

people building agents. Why do you think you're succeeded?" And Peter said, "It's

you're succeeded?" And Peter said, "It's hard to compete with somebody who's just there to have fun."

>> And I I feel like I'm just there to have fun. I'm not doing it for a super

fun. I'm not doing it for a super utilitarian purpose or um I I just feel like it's a it's almost like magic to me. Like I couldn't do this thing

me. Like I couldn't do this thing before, now I can.

>> And also the feeling of being very self-sufficient. Like before you had to

self-sufficient. Like before you had to rely on other people, now you don't have to anymore. So now what what can you do?

to anymore. So now what what can you do?

Mh.

>> Um I I think a lot of people who don't come from a technical I I do think the term technical and nontechnical are slightly outdated these days because we can't just like separate people into

these binary constructs anymore.

>> I think technical is a skill you can learn. It's not like you're born

learn. It's not like you're born technical or non-technical.

>> Um everyone everyone can learn how to steer these um AI tools to get the results you want. No matter what you do, knowing code will make you better at what you do.

>> Whether you're an a marketer or product manager or like a consultant or lawyer, like whatever, a small business owner, no matter what you do, as long as your

work happens on a computer, um I think learning AI coding will benefit your work. Um, and I think the best way to

work. Um, and I think the best way to get started is to >> and sorry, it's almost like we talked about this earlier, like coding is the new form of a writer's literacy. Like

the way that we're all taught writing in school, >> that's what I think it will mean to be literate in the future.

>> Exactly. So, um, if you think about the education system, like everybody needs to learn to read and write.

>> It's like considered basic literacy. I

think coding will be something similar in the future where it's no longer a skill that only a minority of people will will have. Um kind of like writing

because in the past only >> uh so we we think of writing more as a skill than a profession. Um you don't have to be a professional writer to write well or to benefit from writing

well.

>> Uh no matter what you do, writing well will benefit your career. Um and I think coding is similar. Um before AI code coding was a profession reserved to a

small very small minor minority of people requiring years of training. Um

now I think it's a skill accessible to everybody. So uh some people will say oh

everybody. So uh some people will say oh but like the vibe coders are not going to ship anything in like the code will not work in production. But here's the thing like you don't have to build something in production. I think that

mentality is a very >> engineer ccentric way of looking at things because for engineers coding is your day job. So your job you have to ship something in production if it's not

in production it's not meaningful but for everybody else even if it's just like a local workflow that saves me hours every day even if it's only me using it is useful it's it's beneficial

to my work. So I think that requires a mindset shift of like >> we can't judge a piece of code or like um product just by the traditional uh

criteria like are is it reaching millions of people is it making a lot of money even if it's not it it can still be valuable to the individual who who built it. So um like kind of like the

built it. So um like kind of like the concept of personalized software. So

like even something as simple as renaming files on your computer >> can be useful to a lot of people and that doesn't need to be shipped. Um it

can just be like a very small workflow.

So um I think um the if coding is evolving from a prof profession into a skill then um how do we best learn it as Yeah.

>> Yeah. And the other thing I would say is I think OpenClaw demonstrated this where it is incredibly powerful to have your personal assistant. And there are also

personal assistant. And there are also like security issues and things that you do want to understand before giving it root access. And so I think that brings

root access. And so I think that brings us to our next point which is like how to continuously learn so that you are experimenting but also like pulling in people for the health and security

issues or whatever else comes up.

>> Yeah. Exactly. I think um something that really helped uh for me was to start small. like don't be too ambitious when

small. like don't be too ambitious when you're just starting. Like if you start off wanting to build like a scalable system, that's not going to happen. And

all you're going to get is negative feedback. But if you are just trying to

feedback. But if you are just trying to build say a personal website or um like a workflow to to help you with some something repetitive at work, then that's most likely totally doable with

the help of like a the coding agents.

So, and then the other thing is um I think you don't have to define the product super in the super specific way even because the agents are so smart now

like they they they're not just a good engineer they also have very good product sense.

>> So if you just tell it hey here's the goal I want to achieve. I have no idea what the UI should look like >> but like figure it out and most likely the AI will figure out a better UI than

you can. So, a lot of times I'm not even

you can. So, a lot of times I'm not even writing these super long specs or PRDs anymore. I just toss a very vague idea

anymore. I just toss a very vague idea of what I want and then let the AI exercise its own creativity and often times it's it's pleasantly surprising

how good the product sense is for for Claude. Um, and is able to

Claude. Um, and is able to >> come up with interactions that maybe I wouldn't have thought of. So I think um I actually advise I actually ask people

to write shorter prompts not longer ones. Um because you the you have to

ones. Um because you the you have to give AI space to exercise it intelligence.

>> If you control it too much then it's kind of constrained by by your specifications.

>> Yeah. So it sounds like the advice is to just get started to have a relationship with the AI system that you're using whether it's claude whether it's chat GBT that is like spacious for it to surprise you. And then I know we've

surprise you. And then I know we've talked a lot about this, but just to drill the point home, I really think the third part of this is like sharing and having a sharing muscle and you do that so well. So any other things to

so well. So any other things to emphasize around the sharing? Yeah, I

think um um actually before the sharing, yeah, I I want to talk a little bit about >> um I think people should shift their

mental model of their people should change their mental model um for their their relationship with their coding agent.

>> Um because a lot of people still think of the coding agent as like just an assistant um that helps you write code.

Um but it's a lot more than that. Uh so

for example I think my coding agent is also my teacher.

>> So I ask it very stupid questions that I would never ask a human engineer because I don't feel self-conscious when I'm asking these questions to an AI. It's

not going to judge me. It's not going to laugh at me.

>> So it's a very safe space for learning.

Um, and I think that's a huge um point that a lot of people haven't really um leveraged is like you can turn AI into your teacher for anything and it will

patiently explain every concept to you in whatever way that works for you until you understand. It will repeat it 10

you understand. It will repeat it 10 times if you want. uh that level of patient uh p um that level of um kind

kind of customized learning experience is not available to most people in the real world >> and it's extremely valuable. So a lot of times people will say, "Oh, I don't know how to install cloud code. Like I don't

even know what a terminal is. I have no idea how this works." Literally just ask the AI. Like what is a terminal? Where

the AI. Like what is a terminal? Where

do I find it? How do I install? If you,

for example, if you get stuck installing claw code and then you get some error message and literally just like copy it to claude and ask, okay, what do I do?

Um, I think people underestimate how much that will actually solve the problem. 99% of the time um people still

problem. 99% of the time um people still underestimate the intelligence of the the models and think oh if I run into some error that's the end of it like or I give up like it's it's too dumb and

but >> but like it's very common to have errors when uh creating software is like part of the experience >> right and also these coding models have changed so much I mean we'll link it but

Kapate posted like two days ago December was really the turning point for coding models that actually worked and we talked about your audience and um I know it's definitely different from like the

AI researcher background that I have and it's a lot of people don't realize that December was the inflection point and that if they go back to use claw code it's going to be completely different.

So if something didn't work for you in November, try again. It's going to work now. Most likely it will work.

now. Most likely it will work.

>> I I have this very >> strong um kind of personal experience of like things not working out before working now and it's just a matter of a

few months. um the change is very very

few months. um the change is very very um big. Um so I think we should go into

um big. Um so I think we should go into it assuming that everything will just work. If you assume everything will just

work. If you assume everything will just work then you will give AI more ambitious task. You will give it more

ambitious task. You will give it more room to exercise intell intelligence and you will build more interesting things.

Uh whereas if you just assume it's a tool that only executes simple stuff, your conception of what it can do is limited by how you're using it. Yeah,

>> I think that's such an important point.

Um, >> and and now so towards that share reflex, which again I don't think is intuitive to a lot of folks. I certainly

operated in stealth for a really long time. Um, can you talk a little bit

time. Um, can you talk a little bit more, especially for folks that are new to vibe coding, how they would then like share and post about it? Yeah, I think a

very simple way of sharing is just every time you build something can be very simple like a workflow or a project or a website or app um just record a screen

demo. That's probably the easiest way to

demo. That's probably the easiest way to start. It doesn't have to be like a

start. It doesn't have to be like a fancy startup launch video. Mhm.

>> I think people are very fed up with that kind of >> very high production like launch videos with the founders sitting on a couch and all that.

>> People are very tired of that aesthetic.

>> Um and then if you just >> show up as someone who's like genuine and authentic and hey, hey, here's what I built it, why I built it, here's what's cool about it. Um just share your screen and demo the product that that

most of that kind of videos do very well actually uh without being super fancy.

So creating that kind of screen recording usually only takes me like 20 minutes or so. Um because you're just recording your screen. Um and then I

think the key is to showcase why you built it like the opinion behind the product because I do think software is becoming a form of self-expression. So

when I'm building when I'm building a product I'm embedding a worldview into it. So for example for my say the

it. So for example for my say the product I'm launching which uses AI voice chat to where the AI is the host and asking humans the questions the

embedded worldview is human is the main character the AI is just helping you to extract your best thoughts but the AI is not the main character like the soul the

core of the content still comes from the human >> and um that op that's the opinion I want to express when I'm releasing that

product. So I think uh yeah, if you

product. So I think uh yeah, if you build a s I think a good if you want your product to be sharable and like for the maybe your post to go viral or for

people to for it to resonate with people. Um, first you need to have an

people. Um, first you need to have an interesting opinion >> and then you translate that opinion into code or product and then claude is very

good at translating your opinion into a product because it has that engineering product sense. So you just have to bring

product sense. So you just have to bring a very unique opinion and you don't have to worry about oh is it mainstream enough or will I get a lot of users. I

actually think the future of software is going to be less and less mainstream and more and more niche. Um, and then, um, a lot of people say, "Oh, but I'm worried

that like my idea is too niche or like I'm like the super I have this super unique knee that maybe not many people have." I'm like, "You're not as unique

have." I'm like, "You're not as unique as you think you are."

>> Most likely, whatever painoint, weird hobby or painoint you have, there are at least like 10,000 people out there >> who also share the same interest or

painoint or niche. So even if it may sound niche like 10,000 people is very successful for like a vibe coded side project already.

>> Yeah.

>> So and then by sharing that um idea in public you can attract so many like-minded people who who you can then form connections with and build meaningful relationships with. So I

think like building building a public is also a very great way to like make friends and like build human connections.

>> Yeah. And I think the key thing that is like different about this moment, it feels like um as opposed to like the previous SAS era is people are not spending so much money or so much time on what they're sharing on these launch

videos. It's almost like the tick

videos. It's almost like the tick tockification in a way of like tech media. It's now that AI can produce

media. It's now that AI can produce things perfectly. Actually, people want

things perfectly. Actually, people want the like quick raw messy like directly from your brain thought. So, if you're getting started, like just put things out. Know that like you and I or other

out. Know that like you and I or other people are not spending hours or like so much money on this.

>> I think I I want to speak to how to produce content quickly >> because I think people have this stereotype that um making a video takes hours. So, if I don't have like a

hours. So, if I don't have like a stretch of four hours in my calendar, I'm not going to make a video.

>> But for me, making a video takes 20 minutes. I'm sure for you it's a lot

minutes. I'm sure for you it's a lot shorter as well than what most people think and it gets shorter and shorter the more you do it.

>> Um so we can't just lump like the concept of video into the same bucket like there are hundred different ways of making a video. On the one side you can hire an

video. On the one side you can hire an entire production team have very polished back on the other side you can just record yourself on the loom. one

takes days, one takes 10 minutes and then there are a lot of uh things in between as well. So, uh making a video video is just a way of communicating

with an audience. Um and you don't have to like again the style is not the key.

The key is the substance of what you want to say. And I do think for most um busy professionals or builders uh your day is mostly spent on building the thing. Um, you should not spend most of

thing. Um, you should not spend most of your day on like content creation. That

should be that should just be a natural extension of your what are you already spending your most of your time on. Uh

that should be like oh by the way let me share this like very kind of low-key casual way. Um that's the only way you

casual way. Um that's the only way you can make it consistent. So um I think people are always shocked when I tell them how little time I spend on my

content. Um like for tags often it's

content. Um like for tags often it's just like throwaway comments like one or two minutes and then for video often it's just like 20 minutes. Um so again it speaks to the fact that like I

actually I think the the more casual the less effort sometimes you spend on the content the more likely that it will resonate with people.

>> Um because then it's going to be more raw more authentic and then it stands out among a sea of very perfect AI generated content as well.

>> Totally. The key is like how can we um sustainably learn new technology and I think we almost need to reverse the way we used to learn in school. So in school

it was very much learn first build later. So you go through maybe several

later. So you go through maybe several years of training and you take all these foundational classes and then slowly go up. Um and you learn a lot of concepts

up. Um and you learn a lot of concepts and a lot of times uh without knowing whether it will be useful or not to your work and then after graduation hopefully

you will use it. Um whereas now I do think um for folks who are not looking to pivot into a career in engineering and who are just like taking it up as a

hobby or trying to stay relevant um you can very much do the opposite which is build first learn later. So um just just start just get started building

something anything like something easy and then um make it work. So make uh the coding agent will actually code something that actually works and then

seeing that magic moment of having software that works. Um getting that positive feedback um I think that's what makes uh learning sustainable is that

you have to have that uh positive experience but don't stop there because you have to then ask the coding agent like how does it work under the hood. So

that's how I'm approaching a lot of the times like after a project is done I would just ask code I would just ask claude to uh comprehensively explain the

code base and how the technologies work how they're connected to each other in whatever ways you want like sometimes I will ask it to I used to do it as a markdown file but then I felt like it

was not very readable and ugly so I was like can you just make interactive HTML and um illustrating the different parts of the project And I think that has been

really helpful for me. So,

>> and then uh really using AI as not just an assistant but also a teacher. So, uh

whenever the AI says something you don't understand, like just always push like what do you mean by that? Can you tell me more? How does that work? And like

me more? How does that work? And like

push it until you understand everything.

>> Cool. And then the caveats I would give just as a computer scientist and I think we're seeing this a lot with OpenClaw.

Uh there was that Meta Super alignment exec who gave OpenClaw access to her full inbox even though she put it on a toy inbox for a few weeks and it went and delivered everything uh deleted everything. Oh wow.

everything. Oh wow.

>> So like this could happen to anyone. But

the key thing is be really careful with any root access to your computer. If

you're doing OpenC, please do it in a virtual machine.

>> Um and any user data, you just want to be really careful about what is deployed. Um, but similar to what you

deployed. Um, but similar to what you said, just continue to ask Claude questions and it's its own flywheel and if something gets outside of your comfort zone, like that's where community comes in or like the people you follow on X or tweeting it out. Like

I think a lot of folks will pitch in to help.

>> Yeah, that that's why I think starting small is also really important. Like

don't start building like a production level like product for real users.

>> Just build a very simple workflow for yourself and then give it a very limited set of files or um personal information.

um when you're not confident yet and then slowly go up from there. Um and

also sometimes when I get stuck on something the my first resort is always ask and that will solve it like 99.9%

of the time if Claude cannot solve it and I then just ask like a engineer friend and most likely they'll help me solve it. So um there is no unsolvable

solve it. So um there is no unsolvable problem most likely um because we're not trying to invent new things here. which

is like building it's very doable for most coding agents already and >> also I think remembering that everything is learnable is super important because

uh for example if you're trying to read like a dense paper in the past it might not truly being accessible to you if you don't have a background in this field or

you're trying to listen to a talk um or a lecture that's pretty technical in the past I would just have given up it's like nothing to do with me but how um

you can consume that information with AI as your teacher on the side.

>> So maybe for example ask it to remix into a easier version that's um personalized for you or um every time you see some term you don't understand

um just ask it to explain in a very accessible way. Uh sometimes uh so I use

accessible way. Uh sometimes uh so I use uh Gemini in Chrome. So Google has added Gemini to the Chrome sidebar.

>> So sometimes when I'm watching some technical videos, um Gemini can actually see my screen at the same time. So you

can just ask, hey, what does that mean?

Is able to have all the context and explain it.

>> So I think this is available to everybody, but not everyone is tapping into it fully. Um, if you think it's pretty crazy that you can now have a

world-class 247 tutor on whatever subject you need and can explain whatever way that works for you. So, if

we're not making full use of it, I almost feel like it's a it's a waste of this like intellectual um resource that's available to me and

I'm trying to push myself more outside my intellectual comfort zone as well. I

think as adults sometimes we just stay in our comfort zone a lot because it feels safe also because uh after you graduate from college learning is no longer your job.

>> So learning is no longer compulsory.

It's a fully voluntary activity and it takes a lot of discipline and initiative. Um, so and then before it

initiative. Um, so and then before it was kind of painful to go through these boot camps or like drag yourself into these courses or whatever. But now I

feel like the mental barrier for learning stuff has also been lowered for me because um for example if I want to learn something I can literally just tell clot code like hey this is what I

want to learn but I want to learn it in the hands-on way like as a maybe project based learning instead of the traditional wrote learning like taking classes or boot camps

>> and then I would ask claude to >> let's build a project together that can teach me this thing. Um and Claude is very good at like both teaching you to think and building the project that teaches you to think.

>> So and I think the best way to learn anything is to just like learn by doing to do it especially for CS or technical concepts.

>> So >> um I really encourage people to build like see building and learning as the same activity almost like the the way you become better is by building more

and then the way you learn uh you the way you build better is by learning more. that it's a very symbiotic

more. that it's a very symbiotic relationship and um we we should be more proactive and um curious when we are interacting with the AI and like

consciously turn it into a teacher and then maybe even get it to design a personalized curriculum.

>> Yeah. Cool. Well, thank you so much.

Thank you.