Introduction

Hey guys, how many of you said you were developers? This talk is not for you guys. This is for the rest of you, okay?

From Design to Generative AI Reality Check

How fast the ground started shifting

So I lead design in a company called Avail and I think it was about two years ago when the generative AI models started producing chats and we were like, okay, this is pretty cool, it's a good toy.

But when they started producing images, we were like, at least as designers, we were like what the hell is going on here so but we were like okay let's take this in stride we are going to use it in our work and you know use it as best as we can but then just about six months ago it

started producing interfaces and google i think it's about three months ago google's well gemini 3 is producing interfaces on the fly and ux designers like us we are like what the hell is going on you know this this job of ours is not going to last very long and we are seeing the writing on the wall so instead of fretting about it we I

started figuring out trying to figure out whether what I could do you know how can I improve my skills and get ahead of this thing so I started working just you know like going across YouTube looking at all kinds of things trying to learn as much as i can but uh i ran into cursor and windsurf and that's when things changed completely

for me so this talk today is about uh that uh my journey so far how i was able to go from that idea to actually creating an application and the things i've learned uh in that process

When Building Gets Cheap: Finding Your Edge

i just want to put this in front of you guys when the cost of production the tools for producing code or applications you know plummet as they do the things that matter are going to be your your perspective the way you are going to approach the problem how you want to do the same thing is going to

matter a lot and this is this is something that became really apparent so if the things like tools like lovable cursor all these guys are making the job of creating apps easy what uh what is there what's your edge that's that became the question

that i started asking so i'll just take you through what uh i've actually built and i've the title of the presentation is eight insights but i think we'll do four uh given 15 minutes and then we'll have a discussion so before i start with the uh

Demo: An App That Turns Notes Into Multiple Formats

insights let me show you a quick demo of the application this is the world premiere of my application so far I've been the only user of it right so this is so my idea so I started looking

for what ideas do can I build with right and one of the things that one of the videos that I ran into spoke about look into your life and try and automate as much as you can whatever you can right Right. So use AI to automate all these things, whatever you can possibly do.

I found myself, you know, making notes, taking notes like I have ideas while walking, while showering, whatever. Right. And I picked on notes, but it's all disparate and it's not like compiled together. Then I have to spend some time, put the entire set of ideas together into like an article or a document and things like that.

and because I also produce videos on YouTube I have to then convert it into a script and so on and so on but it's all coming from the core set of ideas so what if I built a tool that would take these ideas and use AI to deliver all the formats that I want it in and that's what I built so this is move 37 I'm just

going to log in so something like this is uh so let me just so as you can see here i've just got a whole bunch of things that i just take down notes over time it could be about a person it could be about uh you know personal website projects whatever i want to do it could be just about you know something that i that i heard about economics i put it all together it's all all it's all dumped over here and then I can actually go about here over here and

From notes to articles—and editable drafts

produce an article from it it would take the same things that I stored as notes and convert it into an article for me but it does not create the article and publish it it's it's converted as a draft for me and I can actually you know look at it over here and edit it and do whatever else that I need but that's not

Audio, podcasts, and scripts from the same source

not all so if I go here and I actually click this button it'll produce the audio version of the article for me I don't know if you can hear it here no I don't think it's gonna play here oh yes the Apple Newton was a revolutionary product that was braked with cutting -edge technology like touch screens handwriting recognition and wireless data trunk sorry so I'll stop

it there and but i could also go ahead and create a podcast from it right i click this button and it will create the the equivalent podcast so over here you can see let's say let's just use this as an example so it went about creating the podcast or have you guys heard of notebook llm yeah so you've heard the podcast feature there so i built that so uh it it uh creates the script for me based on the article that I've created which is based on the notes that I had right and this is then just delivered as and if I want to create the audio version of it I just click this button and it'll create the audio version of

Social posts generated from your personal knowledge base

this podcast so that's not all you can I do a lot of social media posting right so I could come here and I can say I need to post something about all the stuff that I've worked on recently. So I'll come here and I'll ask it to suggest a post.

It's using the LLM to look through my notes and come up with something that I've said or written about, and it'll suggest it as a short blurb over here, right? So it could be this, it could be something else. I can hit again and it'll go ahead and create new versions, right? Right.

So this is this is a quick walkthrough of the application that I've done. It's not an advertising of the application. I just want to tell you what I what I've been able to achieve.

As you can see, it's rather complex and it took quite a bit of effort. But this was something that I was able to do. I feel this is something that would have taken a team of four or five people in the traditional in earlier times. it would have taken four or five developers to put together maybe about a month or two months

or something like that but i the fact that i was able to do it without the the back uh background encoding uh is is the fact that i wanted to talk about all right so just in terms

How It Was Built: Tools, Providers, and Deployment

of uh the process uh i wanted to talk about uh the things that i learned so uh with regard to the

Choosing the right builders: Cursor and Windsurf

the tools, I fiddled around with a lot of other things, but I think, you know, WinSurf and Cursor are still what I would recommend.

One specific feature that WinSurf has released recently, I just want to kind of show that. Sorry, Cursor has released recently, I just want to show that.

It's this, right? So it allows you to do this, but it's got this feature where if you're when you're coding you or sorry when you're prompting you can actually point at things and say I want this button to look like this button and

this was incredibly hard to do before it's really funny but this is the part that I got stuck on and I had to go through naming the things in the code and this this one little feature actually you know makes cursor so completely worth it for me right so I just want to kind of like suggest that

just off the bat so this is one of the reasons I recommend cursor I tried lovable and it has it didn't give me the kind of control that I wanted right but cursor and winds of give me that that level of control with the AI providers I

Models, privacy preferences, and hosted inference

I used ulama I definitely used ulama but on a machine like this you can't run big big models so I use a hosted service and my preference is Venice AI this is

they've got privacy built into it they try to take care of privacy as well as non censorship so that was very important to me I don't want my thoughts being censored so I wanted to use some AI provider that didn't censor that

MCP and “deployment without knowing deployment”

right alright so but with respect to the context I'm sorry MCP I use context 7 but I also used render render is a hosting service which also has published its own MCP server which is crazy right so without me knowing how to host or what are the things to do to on the server to get it deployed the deploy my application, I was able to use the MCP server, MCP connection.

Forgive my language, but I was able to use MCP to use Cursor to make my application get deployed. It's incredible, right?

For each of these things, I would have had to rely on somebody else before. It just took away all of that for me.

What Your Role Becomes: Orchestrator, Memory, and QA

What is my role? So one of the things that I started noticing that what I was doing is that I became the orchestrator.

I wasn't worried about the code. 1In fact, if you look at the stats, 97 % of the code was written by the chat. I'm sorry, the LLMs themselves.

And I stand no guarantee for the validity of the code. Most developers over here would be able to tear it apart. That's fine.

But for me, being able to see my idea come alive is more important.

Sweating the UI details (because you can)

and that that was incredible right so my role over here was to orchestrate the thing so what did I want all I had to do was imagine what I wanted why did my application need to behave for example just just going back to my

application see that little red heart over there so when it's got a connection to the server it turns red in full and when it's not connected it turns just an outline I wanted to do that and in most of our most of my conversations with other developers they're like really is that feature really required I think I

want that so and I got that over here right and look at the the styling on the the the thing on the right side right I wanted the titles to come like this there's you know it's just what I imagine should happen has happened and that's incredible that's incredible and so that's that's just one thing that I

Being the long-term memory for the LLM

wanted to kind of highlight over there yeah so at the same time I became the long -term memory so I had to actually be the person that would keep track of what all has been developed because every time you speak to the LLM it's like speaking to somebody who's seeing you for the first time so i had to say this is what was done before

this is what you got to do next so take care of it and in terms of maintaining consistency within the application so that the you know like the four different modules felt felt like they were the same thing that uh was something that i was able to do because i said use this as a reference when

you're building something here right so i was able to maintain that consistency by giving it the long -term memory saying this is already done go ahead and use that and create it all over again

Role compression: designer + developer + tester

my job was also to figure out how to break this application so you know like i was a qa guy already so this is kind of like highlighting another aspect altogether which is all the jobs

the roles are getting compressed we as individuals are going to be the designers we are going to be the developers we're going to do the qa all together and i see this is a pattern that goes on

and it doesn't go on just for these functional roles I think at some point the entrepreneur is going to be the person who is going to do all of this themselves with agents so that is something that I kind of like have to keep in mind so that that comes to the last part I'll talk about that in the

Iteration at (Almost) Zero Cost

end one other aspect that became very clear for me is that the cost of trying

multiple things is zero and that's that is another power for me as a creative person you want to see does this work does this work does this work it's not

two plus two is four there's no definite answers there might be multiple solutions that create the right solution in the end so you get to try out all these things at almost no cost and that's that's another incredible value in my opinion.

What AI Is Good For—and What It Isn’t

Another aspect that I had to kind of keep in mind is what is AI good for and what is AI not good for.

So AI is good for the adaptive fuzzy creative stuff and it's definitely not good for the you know doing math.

I know the earlier example was about math but I run into a lot of problems when I try to to do math with the AI so I'm gonna stop here I'm sorry I wouldn't be able to

finish the rest of it but I urge you guys to try to take a look at the application it's online go ahead few minutes more okay so I'll just cover

this one point I think I wanted to end on this I want to hear your thoughts

Conclusion: The Rise of the Solopreneur and Solution-First Thinking

also I think we are in the world of solopreneurs I think the idea the the traditional structure of companies with these crazy hierarchies are not going to last very long.

And I feel more and more that with all these capabilities coming together, single individuals can be turning themselves into companies.

I don't see a lot of applications as the future. Just because I showed you tools that built applications, I don't believe apps are the future.

I think solutions are the future and building the apps is just gonna get us to that place where we are thinking in terms of individuals that are companies themselves just wanted to leave leave leave you guys with that this is the

address if you guys want to check out the app and try it out for yourselves you guys will be the second third fourth persons all right