Full transcript
0:02Prompt to product rapid AI app creation.
0:05Today is about building actually a bunch
0:07of apps using AI. Um, so first question,
0:10how many people here have I coded an app
0:12of any kind before? All right, I've got
0:15about 30%. Cool. How many people here
0:19have used what tools have you used? Just
0:21shout them out.
0:23>> Bolt. What?
0:23>> Cursor.
0:24>> Cursor. Anything else?
0:26>> Lang chain.
0:27>> Lang chain.
0:29Hiantic. All right, cool. We've got a
0:30bunch of stuff. Great. Today, we're
0:33going to be using four different tools
0:35to build apps, and we're going to
0:36compare and contrast and see how it
0:37actually comes out. All right, first
0:39off, who am I? If we can click it,
0:42that's me. That's my face. It's I'm I'm
0:45the co-founder and CEO of Funkable. I'll
0:46tell you about Funkable in a moment. um
0:48went to MIT for undergrad, graduate
0:50degree in AI and uh our whole goal for
0:53the last 10 years at the company and for
0:55many years before that at MIT has been
0:57about democratizing technology. Myself
1:00and my good friend Paul who you'll
1:01you'll see in a moment um were part of
1:04the teams that built MIT Scratch if any
1:05of you are familiar with that. Then MIT
1:07App Inventor, drag and drop blocks to
1:09teach kids to code. Made the company
1:10thunkable to enable people to learn how
1:13to code. We now have 40% of the Fortune
1:15500 use to build apps. We've got a bunch
1:17of people who are vibe coding, building
1:19apps, and a huge uh education population
1:22as well. Kids learn to code with us. If
1:24you know anybody in the UC system,
1:25they've probably taken a class on us.
1:26It's uh it's pretty awesome. Separately
1:29conduct the San Francisco Civic Symphony
1:31in the Bay Area and uh and freestyle
1:34rap. So, that's you know, some fun stuff
1:36here. Then, who's Paul? Hey, everybody
1:38look to the back and Paul stand up and
1:39wave your hand. That's Paul. He's great.
1:41Uh Paul's uh one of the founding
1:43engineers here at Funkable. um also was
1:46at MIT undergrad and and grad school uh
1:48with me um studying working on Scratch,
1:51App Inventor uh and and making awesome
1:54things. Also a very uh very active
1:56marathon runner and one of the
1:58godfathers of visual programming. So if
2:00you guys have really thought about or
2:01know about visual blocks, he's he's
2:03pretty much the guy. Um Lego Mindstorms,
2:05App Inventor, Scratch, all that. Um Paul
2:07is is is the godfather. All right, so
2:10what are we going to do today? Today,
2:12the goal of the workshop, one, learn
2:14about app creation tools. We're going to
2:16actually learn about what tools are out
2:17in the market. Two, we're going to have
2:19firsthand app experience. We're going to
2:21have a firsthand experience building an
2:23app, meaning we are actually building
2:26real iOS, Android, and web apps in the
2:2930 minute timeline. And three, we are
2:32going to learn from other participants,
2:34meaning this will be collaborative. We
2:36will actually talk to each other, work
2:37with each other, and build some stuff.
2:39All right, first things first. Let's
2:41begin this by Paul. Can you actually
2:43build something for us real quick
2:46just to kick it off and show what I
2:47mean? All right. So, we're going to go
2:49to a website like Funkable. We'll use
2:51Funkable and a few others and we're
2:52going to go there and we're going to go
2:53to prompt our way to building an apps.
2:55So, what app are we going to build? Paul
2:58will kick us off. We are going to build
2:59an app that gives us information about
3:03Type Faster. Paul exhibits at the
3:06Computer History Museum. Yes, my
3:08guessing is so good. I've got AI in
3:10here. Uh, thanks Paul. So, this is going
3:12to actually start generating an app. So,
3:14we're going to see tools like Funkable
3:16where you can type in what do we want to
3:18do and we're actually going to give you
3:19a result. All right. Usually these
3:21things take 30 seconds to a minute to
3:23do. And to not waste any time, let's go
3:25back to the presentation. Then we'll
3:26come back and see what you got out of
3:28there. So, we'll start with Funkable.
3:29What is Funkable? It is the most
3:31powerful visual development platform to
3:33design, develop, and deploy native
3:35Android, iOS, and mobile web apps with
3:38no coding needed. And so what we're able
3:41to do is actually allow you to come in
3:44visually or prompt your way to building
3:46an iOS, Android, and web app. For eight
3:48years, we were a noode visual
3:51development tool, which means that for
3:52the past two years, we have been a Gen
3:54AI vibe coding tool because that's how
3:56it works these days. Um, and so now we
3:58are we are the most powerful tool for
3:59vibe coding mobile apps, iOS and Android
4:01apps. There are a bunch that obviously
4:03do other things, web apps, etc. Um,
4:06Paul, if you could hit it real quick.
4:07See the power of visual development and
4:09why visual development has been the way
4:11until Gen AI came about that people were
4:13building is because it took things like
4:15what you see on the left side there
4:16where you're writing a bunch of code and
4:18turned it into what you see on the
4:19right, which is drag and drop blocks
4:21that you can snap together. If they fit
4:23together, it works. It will compile. If
4:25it doesn't work, if it doesn't compile,
4:27it won't work. Right? It makes a lot of
4:29sense. As you see these things naturally
4:32fit together, it all works. And why did
4:34we approach it this way? Because at MIT,
4:36we figured out that people understand
4:38Legos intuitively, right? You give a kid
4:41Legos, they're able to snap it together.
4:42And so, the same idea we applied to
4:44programming. And that ended up leading
4:46to us building a bunch of tools that you
4:50have probably seen before that I was
4:52mentioning earlier. And so that's been
4:54the way in general non-developers,
4:56novice developers or developers who
4:58don't feel like writing code have been
5:00coding for the last I'd say 10, 15, 20
5:03years. Then of course two years ago,
5:06three years ago, everything changed. Now
5:08people are prompting their way to
5:10building apps.
5:12All right. And so for today's workshop,
5:15what Oh, yeah. There we go. Back at it.
5:17Funkable. We've got 12 million apps
5:19created by 5 million users in 184
5:21countries. So there are a lot of people
5:23who have built apps on us. Over 3% of
5:25all apps on the Google App Store, the
5:27Google Play Store, and the Apple's App
5:28Store have been built on Funkable
5:33and uh it's pretty awesome. So now that
5:37we've actually heard some about what's
5:40going on, we've heard about what people
5:42have done, we're going to get into the
5:43workshop. So what are we going to do for
5:45today's workshop? We're going to use
5:47four different tools. Raise your hand if
5:49you've heard of any of these tools.
5:51Okay, nearly everybody raise your hand
5:53if you have built with any one of these
5:55tools specifically. All right, great.
5:57So, we're actually going to build with
5:59four tools today. Lovable, Bolt, Replet,
6:02and Funkable. Now, I've already told you
6:03about Funkable. The other three are in a
6:06similar world. Lovable and Bolt, more
6:08for web apps. Replet, the CEO of
6:10Replet's actually talking right now um
6:13in the other room, and they're really uh
6:15you know, focusing on being an IDE to to
6:18to write code. All four of these are
6:20tools that allow you to prompt your way
6:22to building apps, but they all do it in
6:24different ways. All right, before we
6:27tell them what we're actually going to
6:28do, Paul, why don't you pull up the the
6:30the thing that you prompted a while ago.
6:32I'm assuming it finished since that
6:33should have taken a couple minutes.
6:34Great. All right, so this is what was
6:37built. Paul typed in an app that tells
6:40us about exhibits at the Computer
6:41History Museum. And as you can see, this
6:44is actually live running on the website.
6:46So this isn't like some weird test
6:48server that was hacked together. Paul,
6:49what actually came together? All right,
6:51we have exhibits. It shows you different
6:53computers that are actually there. You
6:55can click. You've got different tabs.
6:57You can load the different tabs. View
6:59about information for one prompt in
7:02about one minute. That was that's pretty
7:05sweet. Now you can actually click into
7:06it. You can remix it. You can add
7:08functionality. There's a little box at
7:09the bottom that you can prompt. That is
7:11awesome. Incredibly powerful. Very cool.
7:14And so that's what we've done on
7:16Funkable in about, you know, one minute
7:18or less. I don't know how long that
7:19actually took, but it was trivial. And
7:21we actually got this app built and now
7:22we can publish it, deploy it, all of
7:24that. And the other tools are also very
7:27powerful in similar ways. So, I mean,
7:30this is this is awesome. Uh, I'm
7:31actually very pleased to see this
7:33result. You never know with live demos
7:34what's going to happen. Um, all right.
7:36So, back to what we're actually doing
7:38today. Um, we are going to ask
7:42everybody, if you have a laptop open,
7:44great. If not, and you're at a table,
7:46get in groups of, call it teams of
7:48fourish, and we're going to start
7:49building. Agree on the prompt you want
7:52to build. And if you have multiple
7:53computers at the table, use different
7:55tools and actually write the same
7:57prompt. If you are uh if you only have
8:00one laptop at the table, then everybody
8:02use uh use that one laptop. You know,
8:04share with your friends and actually see
8:06what the results are. and we're going to
8:07compare and contrast the results. At the
8:10end of this, I want you to discuss the
8:12results with your team and then we're
8:13going to report back out here to what we
8:15saw. So, we're going to have about 10 to
8:1715 minutes to prompt with each one of
8:19these tools. So, I'm going to ask you to
8:21split into your teams, get at your
8:22tables. If your table is very light, go
8:24join another table and let's start
8:26building with these tools. If you have
8:27any questions, comments, concerns, raise
8:29your hand, let me know. I'll be around.
8:31Paul will be around. Um Chris as well
8:33for my team. Team Chris, if you can
8:35wave, thank you. Chris will be around
8:36and we can make sure that you're getting
8:38set up. Is there anything that I missed
8:40about this workshop, Paul? Awesome.
8:42Thank you so much, team. Let's do this.
8:44And we're going to come back in about 10
8:46to 15 minutes and I expect everybody to
8:48have built four of the same app on these
8:51four different platforms and we're going
8:53to talk about what we actually saw, what
8:54we actually got. And for those who
8:56haven't built before, this is going to
8:57be your first time actually prompt
8:58engineering your way to apps in the
9:00matter of minutes. Thank you so much.
9:02See you in a few.
9:07How has this been for you? Give me a
9:09thumbs up, thumbs down. How's the
9:10experience been? All right, I'm seeing
9:12some thumbs ups, some soos, some thumbs
9:15down. Good. Um, ideally, everyone had
9:18thumbs up, but it's actually more fun to
9:19have these these uh this varied
9:22experience. Um, okay. So, to round this
9:24out, I'm going to ask one person from
9:25every table to stand up and give me 30
9:28seconds reflection. Um, what worked,
9:31what didn't, what' you like, what
9:33didn't. Um, what was your favorite? Um,
9:36give me uh give me reactions and uh
9:38let's start. Who's willing to volunteer?
9:42>> Otherwise, I like starting in the back.
9:44Perfect. Please.
9:46>> And I'm going to need quiet because it's
9:47hard to hear.
9:48>> So, what was
9:51>> actually even better, pause a second.
9:52I'm going to give you a microphone. Turn
9:54to your right.
9:55>> There we are.
9:56>> Okay. Um, what worked well? Uh, we
10:00actually got four pretty decent or I
10:04guess our prompt was about, uh,
10:06suggesting out of off thebeaten path
10:08travel locations to go to. Uh, and we
10:12got pretty similar responses with just
10:15different uh, input
10:18processes. Uh, one of the things that we
10:21did notice from differences was the UI
10:23choices and the amount of visualization
10:25that was used. So like lovable had a lot
10:28more pictures and was more beautiful. Uh
10:33they all seemed pretty easy to use. We
10:36also did like like I think it was
10:38Lovable that had the steps of here's the
10:41plan that we're going to go make for or
10:45creating the app. Do you want to change
10:46the plan?
10:49>> Oh, and Replet had those planned too, I
10:52guess. So we we liked seeing the plan of
10:55what was going to be created. We uh
11:00and generally we're happy with all four.
11:04>> Thank you so much. All right, let's go
11:06to the next table right in front if you
11:07don't mind.
11:10>> Yeah, nice.
11:12>> Yep, that'd be great. Please.
11:13>> Oh, yeah. So, we tried three different
11:15apps all on developable.
11:17>> Nice.
11:18>> The first one was we think to general
11:20because it was trying to us bloodites
11:23that don't know how to code actually
11:25build an app
11:27with AI. Um so that didn't work so well.
11:31And then we tried two very specific
11:34apps. One to do micro donations for a
11:37university. Um and the other one was
11:39micro donations for real estate develop
11:42being developed by women in San
11:43Francisco.
11:48So that would be the big lesson learned.
11:52Fantastic. Thank you so much. And I
11:54think that's a great um a great comment.
11:56Yeah, the more specific you can be with
11:57prompting uh the better it is. With
11:59that, let's go to the table over here.
12:05>> Um our experience of trying to create an
12:09app that would go north you to the
12:12presence of
12:14uh
12:14>> we never
12:16made headway.
12:18were unable to log in despite multiple
12:21efforts. So we're never able to build
12:23one there. Um we did get results from
12:26Funko Global and Replically
12:30they're very different. Funko was by far
12:32the fastest. Um and it immediately also
12:36included both textual visual elements of
12:40how to navigate information in terms of
12:42past, present, groups that we wanted to
12:45learn, etc., etc. Um Replet and lovable
12:50both put the focus on the map itself um
12:55with the option of having you know
12:58filling in prompts on location alerts
13:02that uh would be immediately translated
13:04to the unkn
13:13have always been nice.
13:15>> Nice. All right. Um, very good. Who
13:18else? Uh, give me a hand raise. You're
13:21all going to get called on. All right.
13:22This table. Ron, please. The mic is
13:25coming to you, Ron. There's the mic.
13:28Yes. So, Lovable was only one of the
13:31four apps that was actually able to
13:32generate some results for us.
13:34>> Very good.
13:34>> Uh, designwise, colors, everything
13:37actually looked pretty good. Well, I
13:39guess given the prompt um
13:43I said replet was spinning its wheels
13:46from
13:49spinning it wheels and then bolt would
13:51not allow me to find it after I put it
13:54into
14:04So there
14:09>> all right we got a good thing that did
14:11work. Cool.
14:13Um
14:15who else table over there? Oh wait one
14:17more. Evan make a comment.
14:19>> One additional comment. Um using uh AI
14:24to help refine the prompt was useful.
14:26>> Yep. And that's Thank you so much. So
14:28we'll go to the table over there. That's
14:30also a good thing to realize. It is very
14:32hard to get it right in one shot. But if
14:34you can keep prompting your way, you
14:36realize that usually in about 15
14:39minutes, which feels like a long time,
14:40but also is, you know, different than 3
14:42months, you can actually get something
14:44pretty good. If you're able to get
14:45pretty specific with your prompts, you
14:46can do QA. You're like, this button
14:48didn't work as I needed to fix that
14:49button. And it usually will. All right.
14:50Table over there.
14:52>> All right. Great. Good. So, we want to
14:53have a college recommendation, right? We
14:56try to be as detailed as possible up
14:57front what inputs we want output we want
15:00and like this table here we weren't able
15:02to get um those to work it just the
15:04login was well it's based on um
15:06comfortable was the fastest of minutes I
15:09think got a result very good um took 10
15:14took quite a while but the result was
15:15very good also and the output from logo
15:18I think was also great very fast um the
15:21interface was quite similar for all
15:23those three that worked because really I
15:24guess we had the potential and what we
15:27found is that I think the rapid one kind
15:30of helped the design was a bit easier to
15:33fill out a lot more drop down top which
15:35we didn't say but it kind of figured it
15:36out right so overall we liked it and
15:38even though it took longer I think
15:40that's a good result but we also really
15:42like think about
15:44thank you
15:45>> thanks so much all right anybody who
15:47hasn't gone I know this front table
15:48hasn't gone here anybody else
15:52I know y'all your partner left so you
15:54guys want to otherwise will go right
15:55here, right here. And then we'll move it
15:57to the front here.
15:59>> Um, I approve
16:02resources in San Francisco. And what I
16:05like about I like I like the way it
16:08looks and it looks very usable, but I
16:12also like because it gave me the map. It
16:16created an about page without me even
16:19asking it. So those were my two
16:21favorite.
16:22>> Fantastic.
16:24Thanks so much for that feedback. And
16:26then we'll get to this front table here.
16:28And uh I think you
16:32>> please. Yeah. One of you.
16:34>> I I tried dump the bowl with quite an
16:38interesting idea that had nested logic.
16:42And yeah, the first pass didn't work,
16:44but I was impressed by the speed at
16:47which it occurred. Now I'm trying the
16:48same thing on lovable and it's much
16:50slower.
16:51Try replexing.
16:54Yeah. Um, Replet is also very slow 10
16:58plus
16:59and visually.
17:09>> Yeah, we're using the same prompts,
17:11right? and test to this floor but still
17:14not like help me to log in and yeah BS
17:18doesn't work and uh rapid it's very slow
17:23uh and also interface is very different
17:25the sable definitely it's a very mobile
17:28uh design at least on my app uh iPad uh
17:33and uh in my opinion it's very good
17:35design um and then compel between
17:39lovable and sarable which are the uh
17:42funnest to uh to uh to to to add but
17:47only their their tool right to develop
17:50the app. Uh their nable actually it's
17:54not very visual. It's not a lot of
17:56pictures. Um so yeah honestly speaking
17:59this summer is very fast and also can
18:02can help us to test the design very
18:04quickly.
18:06All right, one more right here. I think
18:07for the last table
18:10>> unless I missed anybody else.
18:15>> So, yeah, my partner took off, but I did
18:18did multiple run on my cell phone.
18:20>> Oh, nice. Which happens to be Android.
18:23So, I asked to build an Android app.
18:25It's just a messenger app.
18:27>> And
18:28uh it it ran pretty quickly, only for a
18:33couple minutes. And then I got uh two
18:35point where it says next steps export to
18:37GitHub install dependencies as Android
18:40platform update dependencies build the
18:41project sync to Android run on device.
18:44So this is not going to happen. Um so I
18:47thought I was done but then I scrolled
18:49down a little bit and there was a
18:50preview if I clicked that and it opened
18:54and it was functional. So, I couldn't
19:00obviously messaging anybody, right? But
19:02it had some fake people in there and
19:05some fake messages and I could send
19:07messages and it did look pretty good.
19:11So,
19:12yeah, not not too bad for
19:15three minutes, right? Yeah, this is this
19:18is cool. I appreciate all of you guys
19:20giving the feedback. I think this is
19:21such an exciting thing where again in uh
19:24in two three years ago, none of this
19:25existed. a year ago basically started to
19:27exist and now I mean on one hand it's
19:30like ah this could look better work
19:31better on the other hand it's like whoa
19:32in you know 3 minutes I got something
19:35that used to take two months to build
19:37right especially if you start
19:38interacting with hardware you know again
19:40taking funable which I know the best
19:42obviously you know if you do something
19:43with a camera now it spits out a camera
19:45map you're like whoa I can open my
19:46camera and take a photo dealing with
19:49asynchronous hardware calls and all that
19:51is a nightmare and now with these
19:53prompting tools it's uh it's way better
19:55and So there are a lot of different
19:57tools out there. Um obviously I have my
20:00biases to what I like but these are I
20:02would say these are the four that I hear
20:04the most about in this regard. Um
20:06obviously there are many more to prompt
20:08in different ways. Um are there any
20:11final comments you had your hand please?
20:18>> There's a lot of great questions. So the
20:20question was what are the implications
20:21of this? There's a lot of really good
20:22questions that people are trying to
20:23figure out the answer to. For
20:25enterprises, what happens if your code
20:27is all vibe coded and you don't actually
20:28know every line of code that's gone in?
20:30For security folks, what uh what are the
20:32security implications? For um for for
20:36junior engineers, if you're learning
20:37code from AI and not reading code that
20:39are written by other engineers, is that
20:41a problem? If no one actually knows
20:43what's going on under the hood, is that
20:45bad? because for malicious reasons or
20:48just generally understanding what's
20:50happening. There are a lot of great
20:51questions you asking moving forward.
20:54Okay, a year later iOS and Android
20:56update their operating systems. Now you
20:57need to upgrade them. Dunkable deals
20:59with all of that under the hood. But now
21:01it's even more different than you
21:02realized before. There are a lot of
21:04really good questions and I've thought a
21:06lot about this and spoken a bunch about
21:08this. I think we are we are way past
21:10time. So I think it's a dialogue worth
21:12having. Um for me I think the answer is
21:15you need to be able to see the code
21:16knows what's going on and actually um
21:19there are AI debuggers, there are
21:21security checkers, there are all these
21:22things in this infrastructure that will
21:24be built around these prompt engineering
21:27platforms. But to be able to v code your
21:29way to something functional in minutes
21:31is super exciting. And so I think we
21:33have to both be really excited about
21:34what's happening as well as ask these
21:36hard questions and actually come up with
21:37answers to these questions. And it's
21:39something I know that we take very
21:40seriously. I thought, well, we work very
21:41hard on the security side, on the
21:43compliance side, etc. And I know a
21:45number of other companies take this
21:46seriously as well. Um, but it's a great
21:48question and the real answer is we're
21:50offering it out in real time. Let's
21:52build this plane while we fly it. Any
21:54other questions, comments, concerns?
21:57>> If not, thank you guys so much for being
21:59here today. It's real treat. I'm glad
22:01you actually got to get hands- on with
22:02this. And now every one of you can say
22:04you have prompt engineered at least an
22:06app or two. If you have any further
22:08questions about Funkable specifically,
22:10you can email us directly. Just hello at
22:12Funkable We'll go to me and Paul
22:13actually. So, please feel free to email
22:15us directly. If you have any further
22:17questions, let me know. Otherwise, thank
22:20you so much for being here. We have one
22:21more session with the CEO Grammarly in
22:23the main hall and then a little coffee
22:24break. I'll be here to hang out. Thanks
22:26so much. Take care.