You've surely heard of the GPT-5 launch in August 2025, but did you catch the part of the release where OpenAI recommended a frontend tech stack for use with ChatGPT? Here's the scoop.

I've always been a realist that tech decisions are overwhelmingly made by "what the guy or gal who used to work here was most comfortable with."

But I've been screwed over by type errors caused by untyped languages like JavaScript in production apps so many times I've lost count.

So I definitely have strong opinions as to what is the best frontend stack for building web apps, and apparently OpenAI does too.

It's no problem at all if you missed OpenAI's "announcement" about who won the frontend stack wars because I did the boring work of reading their entire GPT-5 prompting guide. Here's what I found.

I Guess OpenAI Has Tribal Favorites for Stacks Too

As I was saying before, I think it's entirely fine and preferable, in fact, to have strong opinions about your choice of technologies.

In fact, that's how I define a "career programmer": a specialist programmer with strong opinions about which technologies are best, based on reality. If you're curious, I've written about that before:

That "based on reality" phrase may be a little strong because there are different use cases and types of software engineering.

Of course, different stacks are more appropriate for different purposes, but the question I always ask is "what's the best tool for the job?"

Frankly, though, nobody likes being forced to go with a stack they don't know, which is why I so strongly advocate finding a tech stack that you personally believe in and learning how to debate its merits with others.

That, in fact, is what I mean by becoming a specialist programmer: narrowing your career's focus for the next three to five years by becoming a world-class expert in a small number of tools, those that you like best.

I don't care if you prefer Svelte or vanilla JavaScript or even vibe-coding a prototype and hiring people to clean up after you without even thinking about what programming language to use. You do you and I'm fine with it.

But what I'm not fine with is the "tribalism" around the "stack wars" — the blind allegiance without seriously considering alternatives.

Look, I don't use Go or Rust on a regular basis, but I still think hard enough about it to have an opinion that I publish on my blog.

And unfortunately, I think OpenAI's recommendation is a little bit more based on tribalism than advocating for the merits of the software, frameworks, and programming languages in a reasoned and measured way. So let's get to those recommendations.

Exactly What OpenAI Suggests for Frontend Dev

By far the best place to get LLM tips (other than my publication, According to Context) is by reading comments on Hacker News.

And one of the things that I see recurrently in comments from SWEs who seem pretty happy using LLMs is that they seem to work better on typed languages.

That's why I'm not surprised at all to see TypeScript being not just my preferred development language for frontend development, but OpenAI's. Here's the exact quote.

• Frameworks: Next.js (TypeScript), React, HTML
• Styling/UI: Tailwind CSS, Shadcn/UI, Radix Themes
• Icons: Material Symbols, Heroicons, Lucide
• Animation: Motion [now separate from Framer Motion]
• Fonts: Sans Serif, Inter, Geist, Mono Sans, IBM Plex Sans, Manrope

I think performance is critical to a world-class UX, so I'm not surprised at all to see Next.js being the recommendation.

After all, incremental static generation, server-side rendering, and automatic image optimization, as well as having API routes, are all the reasons I specialized my own frontend career in working in Next.js.

Similarly, I find that I develop at least twice as fast and I maintain existing code bases somewhere between three to five times faster with Tailwind CSS. And it's usually better for performance because it can be compressed more efficiently than traditional CSS.

Shadcn/UI is the current UI package darling of the world, according to the internet. But please let me know in the responses if I should start using that instead of using Headless UI/Tailwind UI.

Similarly, I haven't used Radix Themes or Material Symbols or Lucide. So please, if you agree or disagree with those, let me know in the responses at the end of the article because I am most interested in your opinion, dear reader, not OpenAI's.

I'm just as obsessed with motion design and animation as I am with performance, developer experience, and user experience. So I'm very happy to see Motion (previously known as Framer Motion) in the list.

If you haven't used it, you should because a little bit of animation goes a really long way — and yes, I also use CSS animations (through Tailwind CSS), depending on the use case.

(Unfortunately, you can't use Motion or CSS animations in React Native. So if that's your stack, then I'd suggest checking out Moti because I enjoy using Moti a lot more than writing React Native Reanimated animations by hand.)

Last but not least, OpenAI literally told us to use Sans Serif fonts. So I think the font section could have been left off this list and is kind of hilarious, but at least Comic Sans isn't on there, right?

Why I Think OpenAI's Recommendations Are Tribal

Despite the GPT-5 prompting guide and the OpenAI cookbook being incredibly long, Sammy Boy and team failed to justify any of their recommendations.

Doesn't that strike you as weird?

I would never tell you to use a tool because it's popular. I would only ever recommend a tool to you because my understanding is it's the best tool for the job, usually based on my extensive personal experience as an SWE for more than 20 years now.

In fact, the OpenAI team put some weird stuff in this frontend app development section of their prompting guide. Let me quote it:

"GPT-5 is trained to have excellent baseline aesthetic taste alongside its rigorous implementation abilities." — Source

Yeah, this is exactly the type of baseless assumption that makes me never want to read any AI hype. And this is, again, just one of those statements that I would love to have justified with some examples, data, links, and research.

They don't stop there, though. They go on:

"We're confident in its ability to use all types of web development frameworks and packages; however, for new apps, we recommend using the following frameworks and packages to get the most out of the model's frontend capabilities:" — Ibid (same source as before)

And that's a paragraph immediately preceding the list of frontend tools I quoted earlier.

So where's the justification? Is this like a training data issue thing? Are these just popular tools? Are these the best tools? Really, like, I genuinely want to know.

And I think it's bizarre that the company that makes the most downloaded app in history wouldn't bother to treat its SWE users with more respect than that.

OpenAI Named the Frontend Winner. Do You Agree?

I think it would be really easy to sit here and feel vindicated since I have been using exactly that tech stack that OpenAI recommends for frontend development for the past six years.

But I don't feel that way at all.

Instead, I feel insulted by whoever put this OpenAI GPT-5 prompt guide together because of the complete lack of justification. It makes me think that there is almost no justification at all; they're just trying to hop on a bandwagon.

But hopping on a bandwagon wouldn't be anything like OpenAI and Sam Altman's marketing model, would it? 😂😂

You know what? You let me know in the comments whether we should take OpenAI's unjustified recommendations for frontend development seriously.

Do you agree with my arguments for using Next.js, Tailwind, and Motion?

Or do you go in the exact opposite direction: because OpenAI named "the winner of the frontend stack wars," you are pointedly going to use something else instead?

Or if you're a GPT-5 user, have you tried those tools in this LLM model? Do they work well? Do they work better than in other tools, like Claude, Cursor, or Gemini 2.5 Pro?

Please, let me know. And I can't wait to read what you have to say.

Happy coding!