Quality information, but really needs work on editing. 1) no need for bg music. We want to hear your voice explaining things, not elevator music. 2) careful with the constant zooming (as others have noted). 3) the fill videos with random people are a serious turn off. It has nothing to do with coding. If you are talking while not necessarily doing anything on scree, that's ok. Just keep the camera/screen on the project while you talk about it. No need to all the extra filler. These are instructional videos, not marketing campaigns. But please, do keep up the great work and explanations. Content is great.
Luke from Cursor here, thanks for the video man! I think you'll really like Composer (our newest feature, just released), which integrates a bunch of the stuff you were doing with Claude directly into Cursor (and more). Let me know if you try it out and what you think :)
Thanks Luke, will check it out when I get the chance! I'm using Claude for the image-to-code part of my workflow because the Claude UI makes it easier to interact with images in an iterative fashion. Cursor as an IDE focuses on the codebase as the main window contents.
@@ai-in-practice I've not tested this myself, but I believe there are preview extensions in VSCode/Cursor that can be used similarly to Claude's preview to close that loop in your workflow.
He's right @lukemelas, the image to code in Cursor is lacking a little bit. Or at least I could not tweak my prompts well enough to get the code to match the image. My project uses react-native, so it was for a mobile design
interesting topic but also the music makes it really hard to follow your voice. Please avoid as well additional atmospheric videos, and the above mentioned in out zooming.
Something to note about cursor: the chat will give you suggestions, where as the composer will actually enact changes to the files you give it access to.
Using AI tools like Galileo, Claude, and Cursor to build an app can result in impressive outcomes. These platforms offer advanced capabilities in design, code generation, and user experience enhancements. The final product can be highly innovative and efficient, showcasing the power of AI in app development.
I found claude better than gpt but I think in upcoming months this space is gonna be wildin on next level. Get ready for next gen developers powered by AI.
I actually really like the way your edit looks. Keeping the audience aware of mouse is key. What app did you use to do this? It looks like it’s something you are controlling during the screencast. Is that true- or are you adding all this motion in editing tool(premiere or after effects)? Thanks!
I have! The v0 of this app was made by first generating images in mid journey to get the right tone and style, before generating the app itself with the workflow in the video www.scratch-app.net/
Nice work man! Really cool seeing new tools like Galileo speeding up the development process. One thing Im struggling with myself is when the project grows in Cursor, the AI can hallucinate a bit and create copies of already existing files I use the "with codebase" command to much. Have you experienced that?
Yeah you need to manage context a little better when a project grows. Also because you quickly run out of fast LLM requests when you chat with codebase for every question. I usually either work on a file-by-file basis to make sure the LLM has the right context, or ask very specific questions like 'fix this bug' or 'change this one thing'.
This is pretty cool! It's a nice way to see the practical parts of using these services. Would this also work in adding backend services? (seeing as this was primarily focused on frontend)
Backend is still more manual work in my experience, also because it's harder to write out the requirements for the LLM. If you have an image, the LLM has enough information to generate the code from. For backend, I guess you could try writing an API definition but in practice most backend frameworks have too many caveats to make generating code for them straightforward. I just finished integrating payload CMS into my app, which still took quite a bit of manual work. There's also the question of framework version, which is where a lot of the errors come from. Even for example Nextjs or React, Claude 3.5 sonnet will sometimes generate code that works for an earlier version and you need to fix it manually.
More expectation from you.... also while making video please also consider Nontechnical people like me.... it will make easy for me to build things from these tools... thanks.. liked & subscribed...
Good overall but that was hard to follow especially for an inexperienced software dev like me. Also, I'm not sure what those people dancing were. Was that a screensaver? Will sub with anticipation that your vids will be cleaner in future. If you can make vids like NetworkChuck, you will be a winner. Keep it up. Love the topic.
Thanks tyrone! I thought the dancer looked a lot like Claude, and I needed a visual to show the AI was generating :) Probably not the best idea in the history of cinematography, but hey. Will check out networkchuck, thanks for the suggestion!
I put everything together in Davinci Resolve. The screen recording is done with Screen Studio, which zooms into wherever your cursor is pointing. The AI video editing tools I've seen aren't quite there yet.
Most software development roles will probably look more like AI management in the future - the same thing that happened in digital marketing with the rise of programmatic advertising. Not sure that's a bad thing tbh :) And you still need to understand the basics of cs to develop a proper backend, make sure everything works well together, and there are no performance issues.
I did the exact same thing in Flutter and it took 2 hours. Then I did in React(not Next.js) in 30 minutes or less.( With Firebase) Mobile development is harder than Web development
@@ai-in-practice most of the AI tools seems tailored for Web Development than Mobile development. They'll quickly give convert their outputs to HTML/CSS/Javascript code, but other options are limited.
Thanks! I've found the 'chat with code base' feature to do a good job identifying relevant files, but I've mostly used cursor on smaller projects (max 1000 files). No idea how it'd perform on a giant monorepo 😆
Is there an advantage to using image to code, instead of just grabbing the html from Galileo and have cursor/claude convert the html to react/tailwind? I feel like the end result would be more similar, no?
I found it easier to inspect the initial result in the Claude Anthropic app when you use artifacts, which allows for faster iteration over the design. That's also one of the main reasons I'm using the Anthropic app (instead of just generating the React code with Claude 3.5 sonnet from Cursor).
if you need to rely on AI to create as simple page as this, its so over for you , what I would rather do it that you at least solve bugs on your own and learn the basic structure of your project so that you can easily scale up and modify the code on your own, I wonder how will development look like after like 5 years or so
Horrible video editing bro, 😂 Content is good though. You should have to work on video script too. You work on making an app using AI, then why not use AI to generate script for better presentation. All the best 😁
Quality information, but really needs work on editing. 1) no need for bg music. We want to hear your voice explaining things, not elevator music. 2) careful with the constant zooming (as others have noted). 3) the fill videos with random people are a serious turn off. It has nothing to do with coding. If you are talking while not necessarily doing anything on scree, that's ok. Just keep the camera/screen on the project while you talk about it. No need to all the extra filler. These are instructional videos, not marketing campaigns. But please, do keep up the great work and explanations. Content is great.
Luke from Cursor here, thanks for the video man! I think you'll really like Composer (our newest feature, just released), which integrates a bunch of the stuff you were doing with Claude directly into Cursor (and more). Let me know if you try it out and what you think :)
Will the cursor composer do the work of Galileo as well and what other things can you describe
Thanks Luke, will check it out when I get the chance! I'm using Claude for the image-to-code part of my workflow because the Claude UI makes it easier to interact with images in an iterative fashion. Cursor as an IDE focuses on the codebase as the main window contents.
@@ai-in-practice I've not tested this myself, but I believe there are preview extensions in VSCode/Cursor that can be used similarly to Claude's preview to close that loop in your workflow.
@@Technology_Forum not yet just the coding part
He's right @lukemelas, the image to code in Cursor is lacking a little bit. Or at least I could not tweak my prompts well enough to get the code to match the image. My project uses react-native, so it was for a mobile design
Agreed, the quality output from Claude is pretty wild.
i'm getting dizzy with all the zoom in/out screen moving
Noted! will make sure there is less movement in the next one.
interesting topic but also the music makes it really hard to follow your voice. Please avoid as well additional atmospheric videos, and the above mentioned in out zooming.
@@ai-in-practiceThe zooming is fine, I think it’s the Motion Blur that’s making the dizzy effect. Overall really nice video!
@@ai-in-practicemore explanations in words, less movement
@@ai-in-practicedon’t need to apologize. You are giving free advice and he’s complaining about zooming and zooming out, give me a break.
Thank you for the shoutout! Really appreciate it!
Of course! Love your stuff, it's really helped us out a lot!
Something to note about cursor: the chat will give you suggestions, where as the composer will actually enact changes to the files you give it access to.
Great video without the random people added and background music
Please do more videos like this with backend
A lot of congrats! Got a subscription from Brazil ! 👏👏👏👏
Awesome! Obrigado :)
Using AI tools like Galileo, Claude, and Cursor to build an app can result in impressive outcomes. These platforms offer advanced capabilities in design, code generation, and user experience enhancements. The final product can be highly innovative and efficient, showcasing the power of AI in app development.
Thank you very much about "Galileo" i can now have UI designs which can be developed into an app.
I found claude better than gpt but I think in upcoming months this space is gonna be wildin on next level. Get ready for next gen developers powered by AI.
I actually really like the way your edit looks. Keeping the audience aware of mouse is key. What app did you use to do this? It looks like it’s something you are controlling during the screencast. Is that true- or are you adding all this motion in editing tool(premiere or after effects)?
Thanks!
Thank you!
Thanks for covering Galileo! Never heard of it. Have you tried uploading photos to influence the design? In the video it seems you skipped this
I have! The v0 of this app was made by first generating images in mid journey to get the right tone and style, before generating the app itself with the workflow in the video www.scratch-app.net/
Good job, it makes me eye-opening! Would you like to finish the backend functions?
I probably won't, but will do a video on how to handle data soon!
Whoa....had no clue this was possible. Thanks.
Nice work man! Really cool seeing new tools like Galileo speeding up the development process. One thing Im struggling with myself is when the project grows in Cursor, the AI can hallucinate a bit and create copies of already existing files I use the "with codebase" command to much. Have you experienced that?
Yeah you need to manage context a little better when a project grows. Also because you quickly run out of fast LLM requests when you chat with codebase for every question. I usually either work on a file-by-file basis to make sure the LLM has the right context, or ask very specific questions like 'fix this bug' or 'change this one thing'.
@@ai-in-practice started doing that. Works wayyy better. Thanks!
This is pretty cool! It's a nice way to see the practical parts of using these services. Would this also work in adding backend services? (seeing as this was primarily focused on frontend)
Backend is still more manual work in my experience, also because it's harder to write out the requirements for the LLM. If you have an image, the LLM has enough information to generate the code from. For backend, I guess you could try writing an API definition but in practice most backend frameworks have too many caveats to make generating code for them straightforward. I just finished integrating payload CMS into my app, which still took quite a bit of manual work. There's also the question of framework version, which is where a lot of the errors come from. Even for example Nextjs or React, Claude 3.5 sonnet will sometimes generate code that works for an earlier version and you need to fix it manually.
More expectation from you.... also while making video please also consider Nontechnical people like me.... it will make easy for me to build things from these tools... thanks.. liked & subscribed...
Good overall but that was hard to follow especially for an inexperienced software dev like me. Also, I'm not sure what those people dancing were. Was that a screensaver?
Will sub with anticipation that your vids will be cleaner in future. If you can make vids like NetworkChuck, you will be a winner.
Keep it up. Love the topic.
Thanks tyrone! I thought the dancer looked a lot like Claude, and I needed a visual to show the AI was generating :) Probably not the best idea in the history of cinematography, but hey.
Will check out networkchuck, thanks for the suggestion!
This is amazing!
It's mind blowing!
This is pretty neat!
Good tutorial. I don't get it why there are some dancers and girls in the video.
Video is edited by AI 😢
Did you use an AI tool for the editing, It actually has a very professional looks that I like. Curious to know
I put everything together in Davinci Resolve. The screen recording is done with Screen Studio, which zooms into wherever your cursor is pointing. The AI video editing tools I've seen aren't quite there yet.
a big ol' BRUHHH moment lol loved this
Thanks man!
As cs first year scares me
But good video and what thing i have to do for a job ready in the future
Most software development roles will probably look more like AI management in the future - the same thing that happened in digital marketing with the rise of programmatic advertising. Not sure that's a bad thing tbh :) And you still need to understand the basics of cs to develop a proper backend, make sure everything works well together, and there are no performance issues.
@@ai-in-practice thanks
I did the exact same thing in Flutter and it took 2 hours. Then I did in React(not Next.js) in 30 minutes or less.( With Firebase) Mobile development is harder than Web development
Yeah this workflow definitely works better for some use cases than for others.
@@ai-in-practice most of the AI tools seems tailored for Web Development than Mobile development. They'll quickly give convert their outputs to HTML/CSS/Javascript code, but other options are limited.
Love the content, what do you use to record your videos?
Thanks! I'm using Screen Studio: www.screen.studio/
Very helpful
Ik ben ook bezig met cursor en Claude dev werkt goed :)
Which Boilerplate are you using? Thanks man for the content.
Thanks Dominik! This one: nextjs.org/docs/pages/api-reference/cli/create-next-app
New subscription..AI tools are getting amazing. Thanks for sharing.
thanks you , suscribed and liked
I like your experimentation with video format in general, but maybe too much zooming. I was often confused. Thanks for the interesting presentation.
Will fix that for the next video, thanks!
Got subscribe. Thanks
thanks
Great video! How would you approach adding a feature on an existing legacy codebase?
Thanks! I've found the 'chat with code base' feature to do a good job identifying relevant files, but I've mostly used cursor on smaller projects (max 1000 files).
No idea how it'd perform on a giant monorepo 😆
Thx
Adding the people randomly to video makes no sense and makes the video look like shit and all the zooming in is not needed.
Thanks for the feedback! There's definitely room for improvement 😆
i use v0 dev to generate components , it uses shadcn ui so its gr8
Nice! yeah I need to make some time to check out v0.
Is there an advantage to using image to code, instead of just grabbing the html from Galileo and have cursor/claude convert the html to react/tailwind? I feel like the end result would be more similar, no?
I found it easier to inspect the initial result in the Claude Anthropic app when you use artifacts, which allows for faster iteration over the design. That's also one of the main reasons I'm using the Anthropic app (instead of just generating the React code with Claude 3.5 sonnet from Cursor).
@@ai-in-practice Yes, but why not just paste in Galileo's HTML and have it convert the exact structure and pure css into tailwind/react?
Which app do you use for the zoom in and out?
Screen studio
Which app you are using for recording screen
Screen studio
nice:)
all good, but zoom on mouse move is really annoying.
👍
What’s with the random dance scenes? 🙄 dude keep it professional!!
Nice video, but what's up with the dancing girls?
You don't like interpretative dance?
Isnt v0 better than this galileo? Makes me think it was sponsored
Will check it out, it does the same thing as Galileo AI? I wish this video was sponsored or had affiliate links, but it doesn't :)
3:04 which movie sean.
Stock footage :)
How r u zooming in and out ?
Screen studio does this automatically
♥️
Great content but you should remove the pictures (of women, men) next times
Noted
if you need to rely on AI to create as simple page as this, its so over for you , what I would rather do it that you at least solve bugs on your own and learn the basic structure of your project so that you can easily scale up and modify the code on your own,
I wonder how will development look like after like 5 years or so
Yeah we'll definitely forget some of the basic skills without practice, same as people have gotten worse at doing maths in their head
wtf those images that appear out of nothing and out of context??
Interpretative dance?
Wayyy too much Screen Studio
😂
And how about backend
Backend is a lot harder in my experience because you have to give a lot more explicit instructions.
@@ai-in-practice So it’s not complete? I don’t know about this AI Stuff currently.
And approaches like „no need for 10 Years Software development“
Are just wrong and hype phrases bc SWEs are doing more than just a frontend ui..
Galileo site is down?
Looks to be up rn?
@@ai-in-practice it was weirdly blocked on chrome… I was able to use another browser.
Horrible video editing bro, 😂
Content is good though. You should have to work on video script too.
You work on making an app using AI, then why not use AI to generate script for better presentation. All the best 😁
😂 thanks! i guess
Sorry, but I fell asleep after 5 mins
dont put that dancing couple, its annoying
:)
I am wondering what will happen to web developers and designers
They will need to learn how to adopt AI tools to remain competitive.