I tried to build UI page in all LLMs - ChatGPT, V0, Claude, Cursor, Gemini, Replit. The only one which came out 100% right with no fix is Bolt. WoW 🤩 don’t know how I came across video. Big Thank You, you got a new subscriber!!!
@@briandesign but what happens with the backend code to make the app work and be hosted online? how is the best way to complement this prototype with a backend solution to allow users interactions and real time activity on the app and host them online? by now it's only frontend design
Been playing with this all day and it's amazing, definitely replaces v0 for me, but as a QA engineer Cursor is still in the toolbox until this thing gets the ability to define something like a .cursorrules file, the ability to scrape the web for most recent documentation, the ability to download a Playwright browser or do anything with docker-related. And it just handles existing (messy) codebases better
@@bollvigblack Search for Cursor AI tutorial for beginners (you want the 40 minute one from Greg Isenberg). Will show you everything I just mentioned and more.
I tried it yesterday and honestly it’s pretty good. I’m a designer so I use it so I don’t make my designs from scratch especially when I don’t have access to competitor products for reference
@@micoberss5579 , I am just a hobbyist programmer, and despite seeing a slightly better code with Claude, in my experience loses the context quicker after some iterations. So ChatGPT, still my option. 10 times better? Definitely not.
Sonnet 3.5 was top notch for coding and even better than chatgpt up until a month ago. Then it became retarded and gpt 4o blows it out of the water easy.
Okay, here's the down side. In order to use this service for any commercial purposes (i.e., you want to make money with a web app that you generate with it), you need to purchase AND MAINTAIN the commercial paid subscription, which they can change the terms of at any time. This license to use the code created under this commercial subscription EXPIRES when you stop paying them monthly. So, basically, it's useless unless you decide you love this platform so much that you'll always want to pay them a monthly fee to maintain that right to use what it produces. Now, what happens if they shut down and go out of business? Now you can't pay them, and you can't use the code either. I will never feel good about that. So, for me, for now... it's just a toy. A tech demo. I'm done.
Haha if the code is generated by AI, there's no way I'm going to care about whatever copyright they tried to claim on the generated code. I doubt that that would even hold up in court
Been playing with it today, and I've already hit the daily limit twice (two accounts)! It is amazing, and I have just installed the local version (which is painfully slow)!! I don't know if it will replace cursor, I think you will find a few other new products will come from this, which can only benefit us all. As for now, I am going to upgrade to the pro version as it has effectively build me two basic but well designed SaaS platforms in two hours, for nothing! Which is a game changer. This is going to allow me to work faster, and build anything, undercut all my competitors and still make a very decent living, while only working half the time each week. What is there not to like??
That's the way of things now isn't it? I still hear the drumbeat of awesome new LLM, stable diffusion, etc. models but I don't bother to chase them down, because always boils down to requiring and managing a cluster and a bunch of time consuming installing and tweaking to get something anyone but a mad tweaker would enjoy using and is not painfully slow. Having said that, Google has announced local AI services in an upcoming version of Chrome. Maybe they've created a bunch of distilled models that still do plenty of useful things and can be run successfully on your local PC.
but what happens with the backend code to make the app work and be hosted online? how is the best way to complement this prototype with a backend solution to allow users interactions and real time activity on the app and host them online? by now it's only frontend design
Agree this could be a great app. And thanks for the video. But you use up the free credits in just a few prompts. Then it's $9/month, which gives you a few more fixed credits, but only enough for maybe another half-day / day use (because text-to-code is very verbose and conversational). I ended up using so many credits just trying to get it to fix its own code! Which is crazy. There's no option to buy more credits, you just have to wait til the next month,. So it's not actually useable for real projects. I also noticed sometimes it would break parts of the design when I try to add new page sections. So yeah - it's okay as part of AI hype, but right now, not usable for anything useful.
@@portalteam5832really tho, this is what I keep wondering why people aren’t doing… Is there any reason why this wouldn’t work? I’m currently learning to do this with “Coding With Antonio” & his Canva-clone, copying his source code and creating a custom .env file. Once I’m familiar with it this was my next goal, and Bolt being both OSS and also apparently good output, it seems like the logical choice for me to learn to fork (or clone, not sure the proper term in this case)
Thanks for this. As someone who has been using v0 for a few days over a month now. This did the one thing I requested in their discord, to display a filestructure system for the code preview. So I don't have to locally host and test the code that way. Thanks for this.
but what happens with the backend code to make the app work and be hosted online? how is the best way to complement this prototype with a backend solution to allow users interactions and real time activity on the app and host them online? by now it's only frontend design
I have ran the same prompt and got a totally different result :) Having an Hero page etc... ;) And there is quota (very limited) before being ask to pay
Is it open source? Why would anyone want to use an online service and give them access to your ideas/IP? Can you even take your source from it after you are done?
Sometimes it won’t run the npm run dev You need to open the code section go to terminal and type: npm run dev click enter. Give it few seconds and it should load. Otherwise just tell the AI I have a blank page. Hope that works for you
is your wifi slow? sometimes it would be really slow to show the preview or maybe it's an issue on their end if too many users are trying to use it and causing it to lag. That was a big issue with chatgpt in the early days
I don't know if it makes sense to give $20. Yes, it can work for simple projects. They said they increased the number of tokens. Free tokens are not enough to solve this problem. After the tokens ran out, I felt like I was scammed :) There's something strange.
you'll prob end up learning code by default when using the tools, but def nice to know some code so you can write better prompts/save time debugging issues when the AI doesn't work
This tool is driving me insane. It produces good work, then if you ask it to add a feature, it will break what it has made already. Its one step forward, 2 steps back. Spent far too long on a simple screen, could of built it myself quicker. When the AI is more intelligent, this will be great.
my buddy was just telling me this same issue with Cursor… I think it wasn’t until he tried it outside of cursor with o1 mini, it solved at least some of his issues.
@@sixtyfivewatts65 it's not a problem. you need to build checks and logic in. people expect a very specific thing, and that is the problem with people.
Looks great, have yet to see an answer to how the backend might be set up and solved, or DB management. At least though this gets you a huge head start on the UI.
Code, test, build, run on tje fly is the next big thing. I dont wanna code anymore. As a dev I wanna just set the rules and let the AI sit on my server waiting and when a user comes the AI present the UI all on the fly and saves the data. If a computer can generate the UI and run it at any time, why bother pre building a UI?
yeah I understand. Once AI gets to that point where it can do most of the work then there's no point in coding from scratch. Still would need to know code and how to prompt/edit mistakes, but manually typing every letter doesn't make sense anymore when AI can just auto fill it instantly
Could you give me the steps to access the platform? As you said, I'm having a lot of difficulty using it, how did you do it? I couldn't, and I really wanted to use the platform, I'm full of ideas to put into action
it was a little difficult because you had to go to the bolt new site and then create an account, but they didn't give an email confirmation pop up, so I manually had to check my email and then log in through the link. It took a few tries though so just keep trying to log in with the email and password
The Ai hype is overwhelming the internet news and that is too much for a step-ahead in the IT technology. I have tested most the AI tools in the App Gen and non of them prove to be more than a TOOL or at most a Template +. Let us take what you have done in the video. It can be done by WP with simple theme with much better maintenance and operation for non-coder. I fully agree that this is very helpful for the economic but it might explode as bubble.
This is great.. but this is not for the everyday consumer. Normal folks like me don't know how to use github. Etc. Is there any tutorials on where to start. ?
The irony about these new technologies is they are causing more people to learn code even though there is a whole lot of spam on the net discouraging the learning of code and or no need to learn code jargon.😆
No, it only deals with JS. I asked it some basic PHP stuff and it failed at some things. When I tried to 'fix' it, it decided to rewrite the whole thing in JS and got rid of the PHP. Not good.
i used it 15 mins ( i only know basic html and css) and made a full stack web game and deployed it into vercel and it was amazing. by the way bolt uses claude but trained it towards programming and enhanced it
@@briandesign was just a program but if you start to use you will see what kind of issues you will get. Just click a fix button or write a fix issue taking tousands of tokens, and you will get always envoriement errors or like that errors from server and programside of backen not my faults or usage i couldnt create 2 functional page with this. Cursor is better ever 🫡
It doesnt matter its eating up tooo much its need to be like cursor ai message based or must use a prompt cache and fixing debugging could not be our problem because wetr getting stackblitz envoriement and server issues
I've tried all these tools and they are far behind what we expect from them. They do basic coding that too with a lot of errors. If you try to make modifications, the code doesn't work. All of this AI hype is fake. AI is in development phase and it's goings to take another decade to reach it's true potential.
I would say it's both amazing and kind of nonsense. You just created a "website" you can't use and it doesn't take hours for a human to code that. I would do that in 20 minutes. However as I said, it is also amazing. As it is now it opens up for people to create cheap really boring websites. But how will the AI-apps work in 10 years, 20 years? How many years will it take for AI-apps to create more amazing websites/apps in 10 minutes compared with 10 humans working on it for 6 month? 5 years, 20 years?
yeah you could do it in 20 mins since you know code, but for someone who never watched a single tutorial or just started, it would take them a few days if not weeks to learn all the basics of html, css then JS, react etc. just to be able to make something similar by themselves. Also, I just did a basic test, but I'm sure there are others using it to build more complex apps and things that would take much longer. regarding how many years it would take ai to make X apps, well people are already doing it now and they don't know code. I've already seen plenty of apps/companies that have made/raised a lot of money and they use AI softwares to help them build it
Best resource to learn code: v2.scrimba.com/?via=briandesign
I tried to build UI page in all LLMs - ChatGPT, V0, Claude, Cursor, Gemini, Replit. The only one which came out 100% right with no fix is Bolt. WoW 🤩 don’t know how I came across video. Big Thank You, you got a new subscriber!!!
that's epic! thanks for the sub
@@briandesign but what happens with the backend code to make the app work and be hosted online? how is the best way to complement this prototype with a backend solution to allow users interactions and real time activity on the app and host them online? by now it's only frontend design
Pack up your bags Devs, we will be leaving soon
You came across the video because you talk about it
@@holo406 it also does backend code you can run two projects at the same time.
Been playing with this all day and it's amazing, definitely replaces v0 for me, but as a QA engineer Cursor is still in the toolbox until this thing gets the ability to define something like a .cursorrules file, the ability to scrape the web for most recent documentation, the ability to download a Playwright browser or do anything with docker-related. And it just handles existing (messy) codebases better
yeah agree. Adding instructions is a key thing i like from Cursor.
can you please elaborate more what you mean exactly? i just switched from vscode i don't know what do you mean
yeah this is still in the early stages so can't imagine how it will be once it gets new features
@@bollvigblack Search for Cursor AI tutorial for beginners (you want the 40 minute one from Greg Isenberg). Will show you everything I just mentioned and more.
@@DJ_Locks thanks i just saw it yesterday. btw did you see new E2B Fragments another New Ai tool xD just discovered its worth checking
I tried it yesterday and honestly it’s pretty good. I’m a designer so I use it so I don’t make my designs from scratch especially when I don’t have access to competitor products for reference
I have been using Claude recently. I am impressed hiw good it is . Ten times better than chatgpt to write code
which model did you use for chatgpt?
@@briandesignSonnet and opus I’ve found to be much better at writing code compared to ChatGPT
@@micoberss5579 , I am just a hobbyist programmer, and despite seeing a slightly better code with Claude, in my experience loses the context quicker after some iterations. So ChatGPT, still my option.
10 times better? Definitely not.
Sonnet 3.5 was top notch for coding and even better than chatgpt up until a month ago. Then it became retarded and gpt 4o blows it out of the water easy.
@@woahluca8037nah o1 baby
Okay, here's the down side. In order to use this service for any commercial purposes (i.e., you want to make money with a web app that you generate with it), you need to purchase AND MAINTAIN the commercial paid subscription, which they can change the terms of at any time. This license to use the code created under this commercial subscription EXPIRES when you stop paying them monthly. So, basically, it's useless unless you decide you love this platform so much that you'll always want to pay them a monthly fee to maintain that right to use what it produces. Now, what happens if they shut down and go out of business? Now you can't pay them, and you can't use the code either.
I will never feel good about that. So, for me, for now... it's just a toy. A tech demo. I'm done.
thats true with any tool
Mate the most important thing is the code duh
Haha if the code is generated by AI, there's no way I'm going to care about whatever copyright they tried to claim on the generated code. I doubt that that would even hold up in court
@@kasforai Not my router.
Do they force their own APIs in the returned code?
Been playing with it today, and I've already hit the daily limit twice (two accounts)! It is amazing, and I have just installed the local version (which is painfully slow)!!
I don't know if it will replace cursor, I think you will find a few other new products will come from this, which can only benefit us all. As for now, I am going to upgrade to the pro version as it has effectively build me two basic but well designed SaaS platforms in two hours, for nothing! Which is a game changer. This is going to allow me to work faster, and build anything, undercut all my competitors and still make a very decent living, while only working half the time each week. What is there not to like??
yeah seems like it's in the very early stages, so once they figure out the issues and add new updates then it could be a game changer
That's the way of things now isn't it? I still hear the drumbeat of awesome new LLM, stable diffusion, etc. models but I don't bother to chase them down, because always boils down to requiring and managing a cluster and a bunch of time consuming installing and tweaking to get something anyone but a mad tweaker would enjoy using and is not painfully slow.
Having said that, Google has announced local AI services in an upcoming version of Chrome. Maybe they've created a bunch of distilled models that still do plenty of useful things and can be run successfully on your local PC.
but what happens with the backend code to make the app work and be hosted online? how is the best way to complement this prototype with a backend solution to allow users interactions and real time activity on the app and host them online? by now it's only frontend design
@@holo406you would have to
Generate that as well. This example is using statistic code but I’m sure in the ide you can begin with backend
i am greatful to be at your video. it's really game changer to see how it works. thank you.
thanks for watching
Agree this could be a great app. And thanks for the video. But you use up the free credits in just a few prompts. Then it's $9/month, which gives you a few more fixed credits, but only enough for maybe another half-day / day use (because text-to-code is very verbose and conversational). I ended up using so many credits just trying to get it to fix its own code! Which is crazy. There's no option to buy more credits, you just have to wait til the next month,. So it's not actually useable for real projects. I also noticed sometimes it would break parts of the design when I try to add new page sections. So yeah - it's okay as part of AI hype, but right now, not usable for anything useful.
yeah still early stages, I'm sure in the future they can get cost down and increase the limits
It's open source - take the repo and spin up your own local version then just plug in your own API keys
@@portalteam5832really tho, this is what I keep wondering why people aren’t doing… Is there any reason why this wouldn’t work? I’m currently learning to do this with “Coding With Antonio” & his Canva-clone, copying his source code and creating a custom .env file.
Once I’m familiar with it this was my next goal, and Bolt being both OSS and also apparently good output, it seems like the logical choice for me to learn to fork (or clone, not sure the proper term in this case)
Thanks for this. As someone who has been using v0 for a few days over a month now. This did the one thing I requested in their discord, to display a filestructure system for the code preview. So I don't have to locally host and test the code that way.
Thanks for this.
but what happens with the backend code to make the app work and be hosted online? how is the best way to complement this prototype with a backend solution to allow users interactions and real time activity on the app and host them online? by now it's only frontend design
bolt was smart enough to break down the code into components. I just switched to pro
This is legit. Great find!
Yes , I tested yesterday , it is the greatest of all without a doubt
I have ran the same prompt and got a totally different result :) Having an Hero page etc... ;) And there is quota (very limited) before being ask to pay
Is it open source? Why would anyone want to use an online service and give them access to your ideas/IP? Can you even take your source from it after you are done?
@ 1:37. So how the do you sign in then. You find a problem but don't show/tell the solution ???.
My preview keeps showing me a blank white screen for every app I created. Did you have to do anything prior
Just copy and paste what u said to it. It will fix it 😂 I tried and now no more white page . Make it understand
Sometimes it won’t run the npm run dev
You need to open the code section go to terminal and type: npm run dev click enter.
Give it few seconds and it should load.
Otherwise just tell the AI I have a blank page.
Hope that works for you
is your wifi slow? sometimes it would be really slow to show the preview or maybe it's an issue on their end if too many users are trying to use it and causing it to lag. That was a big issue with chatgpt in the early days
@@Gtencio will try it thanks
@@briandesign will wait it out
I don't know if it makes sense to give $20. Yes, it can work for simple projects. They said they increased the number of tokens. Free tokens are not enough to solve this problem. After the tokens ran out, I felt like I was scammed :) There's something strange.
Definitely not killing Cursor or the others if all bolt.new handles are web apps.
is it for free and unlimited or does it run out and you have to buy it?
bro this is sooooo good
yeah I'm surprised no other major company has added these features in their app already
is it wise to learn coding now ? or jst using those tools without any prior knowledge?
you'll prob end up learning code by default when using the tools, but def nice to know some code so you can write better prompts/save time debugging issues when the AI doesn't work
Learn the basic like data type, structure, if else, looping, try catch, configuration server, environment, etc.
Does this work with Flutter?????
You made static page, every other Ai agent can do it, the problem is, when you start to add more features to it.
"Say goodbye"? maybe to you...
but they can't allow you to edit code directly while showing all the files + deploy all in the browser.
Also when you're a developer yo
Just for Macs? I will wait for the Windows version, meanwhile ChatGPT (with Canvas now) or Claude, will be the options
oh I didn't know it didn't work on windows
I'm on Windows and it works for me using Chrome.
Does it commit to your own Github, or theirs?
it's on their netlify hosting acc temporarily unless you attach yours, but it would auto delete according to them if you don't sync your acc.
es una locura como genera y como entiende el contexto completo de un proyecto, no solo es generar componentes o ventanas como otros asistentes
This is a game changer. They will get funded 100%
yeah crazy it's already like this without funding
Can you ask bolt to generate code based on design image or figma?
This tool is driving me insane. It produces good work, then if you ask it to add a feature, it will break what it has made already. Its one step forward, 2 steps back. Spent far too long on a simple screen, could of built it myself quicker. When the AI is more intelligent, this will be great.
yeah maybe if they get funding or fix the bugs then it could be really good
thats the problem with ai
my buddy was just telling me this same issue with Cursor… I think it wasn’t until he tried it outside of cursor with o1 mini, it solved at least some of his issues.
@@sixtyfivewatts65 it's not a problem. you need to build checks and logic in. people expect a very specific thing, and that is the problem with people.
woahh.. this is too cool
yeah it's pretty dope
bro that was good bro i am using
Have you heard of Replit agent?
yes
Replit is limited to only python
Looks like GPT engineer which is pretty awesome except this one looks like it can actually build it using the stake I want. Will check it out.
nice
I tried it it’s weak, still needs to improve
yeah it's still new so not surprised, needs a few more updates
+1
your pagination update worked properly before you deployed.
yeah I noticed that too, but not sure why it broke once I deployed it
Not creating an account to try Bolt.
Looks great, have yet to see an answer to how the backend might be set up and solved, or DB management. At least though this gets you a huge head start on the UI.
Code, test, build, run on tje fly is the next big thing. I dont wanna code anymore. As a dev I wanna just set the rules and let the AI sit on my server waiting and when a user comes the AI present the UI all on the fly and saves the data.
If a computer can generate the UI and run it at any time, why bother pre building a UI?
yeah I understand. Once AI gets to that point where it can do most of the work then there's no point in coding from scratch. Still would need to know code and how to prompt/edit mistakes, but manually typing every letter doesn't make sense anymore when AI can just auto fill it instantly
Does it support React Native
In free version it is on Sonnet 3.5 . Only.
hmm
Could you give me the steps to access the platform? As you said, I'm having a lot of difficulty using it, how did you do it? I couldn't, and I really wanted to use the platform, I'm full of ideas to put into action
it was a little difficult because you had to go to the bolt new site and then create an account, but they didn't give an email confirmation pop up, so I manually had to check my email and then log in through the link. It took a few tries though so just keep trying to log in with the email and password
The Ai hype is overwhelming the internet news and that is too much for a step-ahead in the IT technology. I have tested most the AI tools in the App Gen and non of them prove to be more than a TOOL or at most a Template +.
Let us take what you have done in the video. It can be done by WP with simple theme with much better maintenance and operation for non-coder.
I fully agree that this is very helpful for the economic but it might explode as bubble.
i dont think clicking on view etc will works because need functions in the backend
English, please.
Is it paid product? Do they have a free tier to try it?
its free for personal use, but paid for commercial projects
It DOES NOT do it all! It is javascript only. You can't build anything with Python or PHP like you can with ChatGPT and Claude.
Sad. This means it doesn't work for Flutter
goodybye chatgpt, then goodbye bolt :)
lol just gotta wait for the others to update and add similar features
When I tried to use it. It has a limit in tokens. Even though it said 40k tokens it exhausted after 3 messages
Do they have funding?
not sure, not that I'm aware of
It's created by Stackblitz, they make that very obvious, not sure why youtube creators keep getting confused about that.
This is great.. but this is not for the everyday consumer. Normal folks like me don't know how to use github. Etc. Is there any tutorials on where to start. ?
How to connect database with this
I can`t believed it.
me either
If I was a kid thinking about a programming career, I'd re-think it.. In 3-5 years devs could be obsolete.. In 10y definitely obsolete.
The irony about these new technologies is they are causing more people to learn code even though there is a whole lot of spam on the net discouraging the learning of code and or no need to learn code jargon.😆
Why I can’t register
No, it only deals with JS. I asked it some basic PHP stuff and it failed at some things. When I tried to 'fix' it, it decided to rewrite the whole thing in JS and got rid of the PHP. Not good.
yeah seems like this focuses on react mainly
@@briandesignso much for "DOES IT ALL"
i used it 15 mins ( i only know basic html and css) and made a full stack web game and deployed it into vercel and it was amazing.
by the way bolt uses claude but trained it towards programming and enhanced it
10m token finished first day
dang what you building
@@briandesign was just a program but if you start to use you will see what kind of issues you will get. Just click a fix button or write a fix issue taking tousands of tokens, and you will get always envoriement errors or like that errors from server and programside of backen not my faults or usage i couldnt create 2 functional page with this. Cursor is better ever 🫡
Mine was finishing without 12 hours(6am to 6pm)
It doesnt matter its eating up tooo much its need to be like cursor ai message based or must use a prompt cache and fixing debugging could not be our problem because wetr getting stackblitz envoriement and server issues
I've tried all these tools and they are far behind what we expect from them. They do basic coding that too with a lot of errors. If you try to make modifications, the code doesn't work. All of this AI hype is fake. AI is in development phase and it's goings to take another decade to reach it's true potential.
For tutorials sake you should probably get it to work
technically the pagination was working during the preview when I was editing the vid, but once it deployed it stopped
I would say it's both amazing and kind of nonsense. You just created a "website" you can't use and it doesn't take hours for a human to code that. I would do that in 20 minutes. However as I said, it is also amazing. As it is now it opens up for people to create cheap really boring websites. But how will the AI-apps work in 10 years, 20 years? How many years will it take for AI-apps to create more amazing websites/apps in 10 minutes compared with 10 humans working on it for 6 month? 5 years, 20 years?
yeah you could do it in 20 mins since you know code, but for someone who never watched a single tutorial or just started, it would take them a few days if not weeks to learn all the basics of html, css then JS, react etc. just to be able to make something similar by themselves. Also, I just did a basic test, but I'm sure there are others using it to build more complex apps and things that would take much longer. regarding how many years it would take ai to make X apps, well people are already doing it now and they don't know code. I've already seen plenty of apps/companies that have made/raised a lot of money and they use AI softwares to help them build it
It does not. Bolt's own website is not responsive and looks absolutely horrible on mobile devices.
The plural of software is software, not softwares.
Will fsd lose their jobs because of these AI??
Lol, killed?
lol
There are tools that are way better than this. If I used this it would be a backwards step.
which ones?
Not Replit. That's awful
Yeah, which ones?
How does the user insert the data like house prices ?
I told it to use made up data for the example, but if you used an API you'd have to update the price value to be shown dynamically which is easy to do
but v0 better
NOT
Another exaggerated video