This week, ChatGPT 5.0 was released! And while I could sit here and regurgitate what other dev influencers are saying about it (Which would make me like any other newsletter, which, BOO to that. But if you'd like to know the general news on it, here's the FireShip video on the newest model), I'd rather talk about the Frontend side of AI.
So about a month ago, I created a post in my Daily.Dev squad with the simple question
How are you using AI in Frontend Development?
And we got some pretty interesting responses! Others... made me worry. So let's take the time to review some exceptional responses, shall we?
As with most "what do you use AI for?" type answers, I saw a general pattern that repeated in a few responses. Some of these weren't surprising. These included:
1) Using GitHub Pilot or similar tools for code suggestions or completion.
2) For creating the "starting point" of a project, then continuing the rest yourself.
3) Some people mentioned general bugfixing, doing CSS (think working with external libraries or animations), or creating basic layouts of containers.
To me, these are good use cases. Whenever there is grunt work that you know AI can do well, fast, and is easy to verify yourself, these are quick wins that speed up development. But, well, most devs who've used AI probably know this. So, no point in rehashing it.
What I'm interested in here is the surprising or controversial use cases that might push some boundaries and create interesting discussion around AI use.
For example, a response from one reader read:
I use it as my assistant; I discuss with it and ask what would be the best option (e.g., in the app router, what would be the best option to implement the language switcher), then use Shadcn for building the component in the header and then gradually follow the steps to complete the implementation. Also, to enhance the UI and heavy tailwind, I rely mostly on AI.
This answer stood out to me here because, in this case, AI is actively helping make architecture and implementation decisions. Unlike earlier answers, where AI helped with bugfixing or suggestions, it sounds like this dev is working 1:1 with the AI, asking for the best options and getting help with using libraries they might not be familiar with.
This, to me, is probably the most ground-breaking aspect AI has had with Frontend development in the last few years. Mostly because:
The hardest part about Frontend development for many devs has always been the infinite number of new libraries and frameworks ready and recommended to use.
Things always felt like they changed quickly in our space. New, "better" and more trendy ways of working always pop up. Thus leaving devs with the sense that they were always "behind" in their craft. This has always been a huge cause for "imposter syndrome" in our field, and both I and others have felt this way.
Except now, you might not necessarily need to spend so much time learning libraries. You could, at a high level, understand what needs to be done or what the library has to do, and have the AI help you connect it to your project.
Syntax and "connecting with this calendar library" type issues seem less relevant with an AI assistant.
All you have to work on then is understanding what "needs to be done" or what's important that AI might leave out (like accessibility or security). Which, to be fair, is probably the hard part of being a dev/engineer anyway.
Okay, new response! This one from André Rodrigues stood out to me:
I just started experimenting with generating dot graphs for a project. e.g., a graph containing the project’s components and their relationships.
I just ask for dot graphs of the components and their relationships, and get a description I can paste in graphviz or alternative software. I have yet to try more models and more precise prompts.
Overall I think it’s a fun experiment.
Now, this DOES sound like a fun experiment. So I decided to try it myself using the newest ChatGPT model 😎
(For those that don't know, because I sure as heck didn't before trying this, a dot file is "just a Graphviz script — it describes the nodes, edges, and styling of your graph. To see it as an actual diagram, you need to render it." Neato!)
Feeling a little cocky, I spun up a new Chat window and spat out in a single prompt:
Hey, can I give you a GitHub repo, and you generate a dot graph, for let's say this repo?
https://github.com/facebook/react
It said yes, and after waiting 12 seconds, it gave me a .dot file which it described as a "high-level package/relationship DOT graph for the React monorepo". I copied and pasted the .dot file into a Graphviz visual editor, and saw this:
Cool.
Is it correct?
No idea 😂
I've never even opened the actual React repo, so I have no way of knowing
So it's a bad test. Plus, if I wanted a full graph, it involved having to clone the repo, download an external library, set up the project (which, by the way, immediately gave me issues when I tried doing an npm install), and run various CLI commands.
So instead, we went with a simpler test that most devs could understand:
1) Use this simple React landing page project as the base.
2) Download the project as a .zip file
3) Paste it into the ChatGPT window and have it generate the .dot file based on the project
4) View the .dot file in the graph visualizer.
This was the result:
This is cool! I like this experiment, and it gives you a good view of which components are coupled together and where each one is used.
I'm sure other "view your components as a tree" tools exist, but it's interesting how AI can create one for you as well.
Alright, this is becoming a long email, last response we'll go over!
I think it is time to evolve. Using for code is now normal usage and we need use ai that lives our website in interesting way that pointed to users.
This... This SPEAKS to me.
Yes
YES
WE DO "NEED TO EVOLVE"
WE HAVEN'T MADE OUR JOBS "EASIER" TO DO ENOUGH
GUYS, LET'S JUST ADD MORE FRAMEWORKS IN THE FRONTEND
I PROPOSE A NEW PARADIGM SHIFT WHERE WE ADD ANOTHER SERVER TO THE FRONTEND
WE'LL CALL IT... SSSR
SERVER SERVER SIDE RENDERING
WE'LL ALL BECOME DOUBLE-FULL-STACK DEVELOPERS
WITH GPT 5, WE CAN'T STOP NOW.
WE, AS EXCEPTIONAL FRONTEND DEVELOPERS, "NEED USE AI THAT LIVES OUR WEBSITE IN INTERESTING WAY THAT POINTED TO USERS"
Truly words to live by 😭
Okay, okay, in all seriousness (and to give the poor guy some slack), AI has changed things. It's one hell of a tool. Companies are hungry to apply it to everything and throw AI at the wall to see if it generates any more new money. A lot of times, to the detriment of actual productivity and project quality.
I just get worried when somebody says, "We need to evolve".
It's the same as somebody coming to you and saying, "We need to innovate!!"
No bud. You have it wrong.
That's not how you evolve or innovate.
Evolving or innovation is like happiness. The more you chase it, the less of it you have.
Evolving and innovation come when you find ways to solve problems in a cheaper, faster, or higher-quality way. Many times in a context that nobody has thought of before.
Yes, AI is cheaper and faster, but you can't just ignore how much it leaves to be desired when it comes to real quality.
The more corners you cut, the more trust erodes whenever something goes wrong. Whether that's a client or user who finds a new bug, or there's a security leak, or the website unexpectedly breaks, and the AI can't fix it immediately.
Though I believe most devs understand this already. In most areas where we've found tasks that were repetitive, annoying, or automated, we've already used AI to whisk them away or minimize them.
Obviously, because we don't want to do it 😄
But how much more can we "evolve" if we just keep trying to cram AI in places that might not need it?
If we use this hammer for everything, doesn't everything just look like a nail?
If we really want to innovate or evolve with AI, I wouldn't keep concentrating only on the tech space or on websites. Why not use this amazing "hammer" we have and see what problems in other contexts we can solve?
How are farmers using AI? Are they using AI?
What about coal miners? Submarines? Satellites? Doctors? The government? The stock market?
See your curiosity growing? Wondering?
That's how innovation is created. By changing contexts and seeing what problems you can solve in a new way.
P.S.
Hope you have a wonderful weekend! Reply with a "hey" if you actually finished reading the email. I want to know if it's worth writing emails this long every once in a while 😂