Hacker News Reader

Comment by erellsworth | Hacker News Reader

Don't worry, they will just send this message to their "prompt engineer" who will spend weeks tweaking prompts to get something that's "good enough." In the end it will cost just as much as if they paid a human developer but with the added benefit of being shittier.

View on HN

Maybe. I see a lot of whistling past the graveyard here in tech world. Design is a critical and nuanced world, but if an LLM can get you 80% or even 50% there it's a time (and possibly resource) saver.

Don't existing templates get 90% of people 90% of the way there?

Yes - and that is basically what tools like this are doing. Automating the work of matching a template to a request, plugging in the relevant compatible components (storefront, blog, whatever) and then just customizing the the theme and filling in specific content. They aren't writing raw html and css most the time, but rather orchestrating an existing design.

Almost - This tool helps select a template, adding business applications and creating default content. But then, you have the Wix drag & drop builder to continue customizing the website, adding animations, changing content, adding or removing pages / sections, changing design, etc.

The strength of this offering is that it is seamless with the experience Wix is based on.

I've seen the monstrosities non-professional web designers can create. They ain't pretty but often serve their purpose well enough. As soon as you need a professional looking and functioning site designers and developers are needed regardless.

I agree. Web Designers and Developers will always be needed for the more unique look and feel, more unique animations and specific functionality.

I think the role of AI generated site is to help getting a better result fast, both for non professionals, but also for professionals as a starting point.

Sure, but they're far more rigid and don't take a ton of input beyond specific SCSS-type variables.

I am somewhat curious about this though. It seems like this kind of workflow would be more flexible than a dumb template, but in practice I am not so sure these days. The back and forth of it gets slightly tiring and you start to miss the simpler, unambiguous tools. Perhaps its just something about the apparently extreme complexity of prompt engineering, but the feeling of freedom/flexibility in the prompt based workflow at the very least feels comparable to a good template based one.

This is not to wave away the other things it can do like generating copy.

That’s a plus when you need someone to support your site later, not a minus.

This reminds me of the #1 reason everybody hates WordPress....their first experience with it was having to fix some cheap contractor's mistakes.

Most human-developed websites are pretty shitty. It's what people are used to, so if they can get it faster and without the excuses that mediocre web shops have for why stuff doesn't work very well, they will go for it.

People will pay some money just to differentiate from shittier to… maybe just shit.

except it scales to 10s of thousands of sites.

what value are you providing to this conversation other than unwarranted pessimism?

> except it scales to 10s of thousands of sites.

Does it though? In what sense, other than content, would a site created by AI scale any better than one created by a human? Would each of these 10s of thousands of sites be unique? Or would they be slight variations on a theme? Because just about any human dev could write a script in a few hours to spin up WordPress sites with slight variations on a single theme, limited only by how much they were willing to spend on hosting.

What value are you providing to this conversation other than unwarranted optimism?

> Because just about any human dev could write a script in a few hours

what are you even talking about? because I'm talking about AI scaling for the Wix website builder. where people who don't even know how to write copy for their website, let alone code for, use to build a web presence.

.. which will all look and feel exactly the same and the only thing differentiating you from the competition is doing non-LLM and probably very human design.

I don't actually work in the design part of the company, just occasionally implementing UX designs, but I've frequently been in the client meetings, and the point you make here cannot be overstated. This is the prime motivator of most clients when presented with a "templated" website - they want theirs to look different enough not to be recognized as the template. They want this so much they will push for sub-optimal UX, just for the sake of differentiation. I honestly don't blame them, as our templates, just as most templates, produce recognizable results - pretty much same site, different colors and logo.

All that being said, for more than 3/4 of them, they change their mind pretty quick when they find out the price difference between going with the templates or having custom UX developed. And this is just with the farmed out UX people who are cheap - not our good designers who the big customers use. Templates are good 'cause they let the little guy at least have the form and functionality of the big guys as long as they can tolerate they're essentially a row house.

I don't disagree with you, but there are a lot of companies (mine included) that don't care if they stand out. I need a website so people believe I'm a real company; as a place to convey basic information about our product and team; and as a place to provide information on how to get in touch. It's a chore to make it and I'm happy to just phone it in.

yeah, I'm sure people that have no idea how to build a website are really good at making their website stand out, right?

I work on a competitor website builder and I can tell you that this is definitely not the case. they'll leave in the default helper text we provide half the time.

Reply on news.ycombinator.com