Transcript of YouTube Video: How AI Can Help You Turn an Idea Into the Next Great App | Amjad Masad | TED

You know, one of my earliest memories

as a kid growing up in Jordan

was watching my father,

who was a government engineer at the time,

set up a new computer.

I was just a kid, but that machine really gripped me.

It felt like it was magic.

We couldn't really afford it, but my father had a feeling

that this thing is going to be transformational

for the world and for our lives.

At first, like most kids, I started playing video games.

I figured out how to boot up video games and play them for hours.

But then I had a radical thought.

What if I could make my own video games?

The very notion that I could take a concept in my head

and make it a reality that I and other people can explore

was something that was really gripping.

Now fast-forward to my teenage years

and the internet has just arrived.

And I thought it was going to change everything.

I was bursting with business ideas.

But there was a problem.

I didn't know where to start.

Which language do you use?

How do you even share your creations with the world?

And these questions sent me down a rabbit hole.

The first thing I attempted doing at solving these problems

is putting a code editor in the browser.

Can you make it so that you can code anything really easily

and share it with the world?

So eventually I had a breakthrough.

I built the world's first online programming sandbox

with a lot of different languages,

and that went viral.

A bunch of companies in Silicon Valley started using it,

and they hired me.

I got a visa to the United States and I came to New York.

A few years later, I set eyes to go west

and I came to Silicon Valley to start a company.

And that was Replit.

Our dream is to make it so that anyone can turn an idea into software quickly.

Our vision is to empower a billion software creators.

But giving people an editor was not going to be enough

to get a billion people coding on it.

We've signed up millions of people

and today we have 25 million developers registered on our site.

But we need an order of magnitude jump.

And a couple of years ago I stumbled on this technology.

That's the reason we're all here.

Large language models.

And I thought this was the unlock

that's going to make it possible for anyone to become a software creator.

The interesting thing about large language models

is that you can transform any thought in any language,

any human language,

into software.

And this, I thought, is how we bring a billion software creators online.

So right now I want you to imagine something.

Let's say you're traveling in a new city.

And you're walking around and seeing these magnificent landmarks.

But you're interested in the stories around these landmarks.

Now it kind of takes you out of the moment to pull up your phone

and Google and find a Wikipedia page for that particular landmark.

Wouldn't it be easier if you had a map

with the Wikipedia links superimposed on the map?

Typically, an app idea like that will be just a fleeting thought,

and maybe you just forget about it and continue with your life.

Maybe you'll buy a domain.

And I know I've been guilty of buying tons of domains

that never turn into anything.

But in the world that I'm envisioning,

you could actually do something about it in that moment.

You pull out your phone and you talk to your AI assistant.

And the AI assistant is not just a passive listener.

It's going to start brainstorming with you.

It's going to make connections.

It'll find projects you've worked on in the past

that kind of look like this idea.

And let's say you want to give a clearer picture.

Well, pull out a napkin. Draw a mark.

Snap a picture

and boom, you send it to your agent.

Let's start strategizing and thinking of a road map.

Now you don't have a lot of time to chat with it, you're on the go.

So you send it a voice note.

Now you head home.

And you have a prototype.

You have a tangible thing that you could actually play with.

You have a starting point to make an application out of.

But it's not just a prototype. You also have a plan.

You have a plan of action that could turn this prototype

into a production app that you can share with the world.

Now, apps take time and they take money.

So we're going to set a budget of how much money we're going to spend

and how much time we're going to spend on this app.

To take a look behind the scenes,

it's not just magic, although it'll feel magical.

The plan editor becomes a plan source.

It recursively expands all the different tasks.

It goes out and assembles the tools.

And these are tools that, as developers, we use every day.

There's your code editor, your source control

and everything that we're used to.

And you have, of course, the code.

Now this is not just about coding.

It's about a new way of communicating with technology.

Being able to talk and draw and work with AI

to make your visions reality.

Let's go back to our plan.

We're going to run it.

It has four feature ideas.

The first one is user auth.

Every app needs that.

The second one is a way to collect feedback from users.

And the third one is a way to like places you've been to.

And the fourth one is a navigation.

The first branch is kind of easy.

The AI has seen login pages a trillion times and knows how to code it.

We'll take a look at the code. We'll merge it.

The second one is a little bit hard. Maybe it's a novel task.

But the AI here is prompting us

to add more compute.

So the easiest way to solve this problem

would be to pump in more compute.

Maybe that means a larger model,

or additional contacts or additional tokens

in order to solve this problem.

And the problem is solved.

The third branch is kind of easy.

Smooth sailing.

I look at it, it works, I merge it.

Now the fourth one is kind of difficult.

The AI seems confident

that it's not going to be able to solve this problem with more compute.

So it prompts me with an increased budget.

Now it's a hefty number,

and that's because it's not just going to recruit AIs.

It's going to go out and recruit people. It's going to post a bounty.

So in this case, we need a human in the loop.

And until we get AGI,

today, natural intelligence is going to be really, really important.

So we go out, we get a human.

That quota is going to also be assisted by AI,

and they're going to help me solve the problem.

I merge everything together

and it's just a matter of hours and 50-60 dollars or so,

I have an app I can deploy.

I deploy it, I share it with the world.

Now to step back a little bit,

notice that I didn't write a single line of code.

I was basically the creative heartbeat of the project.

I manage the project,

I manage both AI and people,

mediated by AI

in order to build this application.

And that fleeting thought that you had is suddenly an application.

So, you know, I can almost hear some of you thinking, like,

"OK, Amjad, this is this is great and all.

This is a fascinating vision

but it kind of looks too fantastic, two futuristic."

Well, I'm here to tell you it's kind of already happening.

So just a couple of weeks ago, an entrepreneur on Replit, Yoeri --

he's a technical recruiter by training, he's not a software engineer --

he learned enough code to start building his startup.

And thousands of miles across the world we have Akashdeep.

Akashdeep, a student in India, he comes from a farming family

and he can't afford a PC.

He has never had a PC. All he's had was his Android phone.

He started learning to code in Replit just a few months ago,

and then decided to go and start making money on the platform.

Now he makes more money than his entire family

and he can afford a lot of different computers, not just one.

And now he's joined forces with Yoeri to build that startup.

You know, today we talk about the Fortune 500,

and it takes a ton of employees to build a big company like that.

But I think in the future,

when AI is helping everyone build their dreams,

we’re going to be talking about the Fortune 5000000.

And again, this is not just --

This is not just a fantasy.

We're already building this.

Last week we announced that we're open-sourcing our AI models,

and we're making it free for all our users

because we think that's a massive step up for all of humanity.

AI is not just a tool, it's a ladder that equalizes opportunity.

And that opens up horizons for people that are typically on the sidelines.

So I'll leave you with this.

Imagine that

the American dream,

not just --

not just in one place,

but actually in the cloud.

Thank you.

(Applause)