tl;dr
- I wanted to test building a small website comparing WordPress to a custom AI generated site to support people with a medical issue.
- Bottom line is the generative site is easily competitive with WordPress, and very flexible.
- Final choice is to to go with the reliable WordPress option though; so I don’t have to worry about ongoing issues, but the generative site capabilities were somewhere between kind of cool and astounding.
- The end results include the ACL support site generated with a Gen AI tool, (which is not content complete at this point), and the go to market site, aclsupport.com, which is managed hosting WordPress. (No attempt to make them match visually or with content; just get the basics up.)
That’s it! if you want the full story, here you go…
What’s the Goal?
Build a simple, small information site to support a user/patient medical community, and also try out an AI site generator, (we’ll use Lovable.dev), then compare it to longtime content management system WordPress. Both sites to be Mobile responsive, but no native apps.
My goal is to not burn more than a day or two effort for the initial build. For the record: I do not believe this is a smart way to build production sites that would have any important functionality. However, this fast attack approach is fine for hobbyist sites and prototypes. When done, I’ll decide which one makes sense to keep long term.
Why Do This?
- It’s something I want to create because a) I think there’s a need, and b) as a thank you to others who helped me through a painful journey.
- It’s a personal experiment. I’m a product manager, not a developer. While often working at a strategic or team management level, I also like to test and understand the tools of our craft. And I wanted to do a couple of fast builds with a real project to get a more solid feel for the generative builder capabilities. I’ve seen some articles that compare AI to WordPress, but they’re theoretical, speculating on pros/cons based on general capabilities. I’m going to actually build with both.
- We’ve heard buzz that we don’t need programmers now because AI empowers entrepreneurs and product folks to just build tech. It’s just so easy. But we don’t need product managers because coders or CEOs can use AI for all product, marketing and all strategic tasks. I’m not sure whom we’re left with at that point. These ideas are great clickbait, but shortsighted and wrong for all manner of reasons. Still, used properly, these tools can help with team efficiency, speed, testing, and more. My goal is to see how easily a “somewhat technical” product person, (though with some degree of long atrophied dev experience), can build alone.
More About the Product and the Why
This information resource site is for those who have suffered from a leg injury involving Anterior Cruciate Ligament (ACL) damage, and associated injuries. It will include content, references, and product / service suggestions. The goal is to satisfy constantly recurring questions in one place. Having been injured somewhat badly before myself, and as a long time community EMS volunteer I’ve seen all manner of injuries. But my own most recent sporting event injury was a bad one and I wanted to put something out there for others that can maybe help get them past the kind of information seeking challenges I faced post-injury, post-surgery, and during recovery rehab.
The Build
My budget is $200 for both sites, including hosting for a year. (Edit: I was off a bit. It cost $12 for the first year, though ok, I already had hosting for one of the sites with an existing account.) All in, this took about 8 – 10 hours over the course of a week, plus another few hours for content writing / touch-ups. Most of the effort was on the WordPress site.
WordPress Effort
For the WordPress ACLSupport.com site, I’m using a Managed WordPress plan where I won’t even have to do basic updates. It’s cheap the first year, and not bad beyond. I can pretty much set it and forget it except updating content. (I’m still messing with the theme/look/feel a bit. It’s “ok” for now, but not all that great. So any graphics here might not be representative for long.)
AI Generated Effort
For the AI generated ACLSupport site, I used Lovable.dev to build, and hosted the files on my already in use server. Lovable can sync with GitHub which can then automatically deploy the files for me. I’m just dumping everything into a folder and giving it a sub-domain.
Product Requirements
There should be a Product Requirements Document (PRD) or at least feature stories. Since we’re just messing around here, this will be simple. We only need some prompt(s) for the AI generator. (Changes with additional prompts could be new user stories.)
Here’s a Google Doc with my initial prompts for the website generator, which can be considered as a weak PRD. You’ve heard of “The One Minute Manager?” Well, this is the 3 minute PRD.
Results Assessment
The Winner is WordPress. But just barely. My concern here is having a known stable platform for what is a solo minimal time hobby effort with intermittent updates. If making a site with custom functionality, it could easily be an AI builder win. Though I’m not talking about “vibe coding.” I’m talking about using GenAI as good starting places. They’re not yet trustworthy for serious production without professional review and management. (No matter how many YouTubers are telling us how they’re making $$$ per week off their Vibeware.)
General Impression of using an AI Builder
Pros
- Useful for prototypes or jumpstarting something real.
- Speed to build is astonishing. Initial site generation was about 10 minutes. Then an hour or so for adjustment prompts. Creating the initial admin section section took a handful of prompts and 20 – 30 minutes. Typically, this might be dozens of user stories, (likely more), days of design work, (at least), and likely a couple of sprints involving at least two developers. In other words, 20 – 30 minutes, vs. 4 – 5 person months of effort. That is simply astonishing.
- Easy to iterate. (Without having a meeting to discuss, waiting for design to get assets to development, etc. Just tell the machine what to do, grab a cup of tea, and that’s it.)
Cons:
- Can take awhile to undo undesirable elements.
- Can be slower. Like any product, over time, it’s faster as you learn tricks.
Mixed
- I told the machine what to do and it did it. No meetings. No discussion. I didn’t have to wait for anyone. Many say this is the benefit. But if you believe, (as I do), that talented teams do better together – for the most part – than even solid product managers or CEOs, then this isn’t necessarily all good. In fact, for many it may change “speed to market” into “speed to pure crap.” And this may be “pure crap at scale.” Yes, we’ll see some successful cases. And yes, sometimes the more you throw at the wall, the more sticks. At the same time, it’s not always the case that doing more wrong things fast naturally leads to ah-hah momenta. It’s possible, (likely?), in the wrong hands these tools will just make motion feel like progress when some may just be moving in wrong directions. Though I suppose that’s no different than any other craft. Better tools sometimes level up everyone, but talented craftspeople will still shine more brightly.
WordPress Specific Results
- Build
- Getting the basic site up took about an hour.
- Adjusting themes and getting plugins working took several frustrating hours. Lots of flexibility / capability, but still constrained and sometimes challenging to find the right switch. Even then, it’s useful to use GenAI to help with custom code to inject in some places. (This also isn’t the prettiest site in the world. Some more theme tweaking is likely in order.)
- Bottom Line: If you have content together and aren’t looking for a lot of customization to start, it’s maybe half a day effort to have something passable up and running.
- Pros
- Trivially easy to initiate and set up, in spite of hassles tweaking plugins and themes.
- Decades of industry support services and tools. Themes, Plugins, help sites, contractors, etc.
- Cons
- WordPress and associated tools have gotten somewhat bloated. Getting themes right can be challenging. Lots of dependencies, some of which are complicated to sort out.
Lovable.Dev Specific Results
- Build
- Started at: 3:00 PM
- First version: 3:10 PM
- That’s right. Version 1.0. Ten Minutes. OK, I had the prompt written. So if you like, add in another few minutes.
- It took a couple dozen prompts to adjust. And another dozen to set up a content admin tool.
- Bottom Line: Same as WordPress, it’s maybe half a day effort to have something passable up and running. But probably a whole day or somewhat longer to adjust various bits and pieces.
- Pros
- Amazing capabilities. Shocking actually.
- Overall design is decent. Nothing thrilling, but could adjust further.
- Can quickly modify a prompt for fast changes.
- Cons
- Can be challenging to get changes exactly as desired. This may be true when working with people as well. Still, it may take several iterations for simple changes. The tool can repeatedly make mistakes. If using a paid account with credits per iteration, you can burn through credits in spite of mistakes.
- Without a content admin tool, generating pages and adding content is a hassle. More robust workflow tools would be high on the product backlog.
- Making changes can be done directly manipulating code. But besides needing to know what you’re doing, it’s going to take time to understand how. And then if you adjust something, some or most of what you’ve learned may change on another prompt.
- Quick safety note: If you use generative tools, take care about things like API keys or similar. If you port directly to GitHub, (even a private repo), you may risk exposing private keys. So you’ll may want to handle such things carefully.
Next Steps
These were super quick stand ups with minimal thought or planning. So next steps would include adding analytics, some basic plugins for SEO, maybe advertising to see if there’s enough traffic at some point to cover the hosting. As well, I’ll probably toss in a GPT Chat plugin because, well, I mean… it’s 2025. Without AI I might as well just grab an “Under Construction” .gif from the 1990s. Or maybe add some glassy shiny beveled buttons.
Anyway, Minimal Viable Product (MVP) 1.0 is done. I guess I did ok from an MVP perspective because I’m not at all proud of it. Needs work. But it was a fun experiment.
Final Thoughts…
Fast Gen AI site builders… vibe coding… super fast deployment…
- Using generative site building is an arguably good idea, at least for fast prototypes. I can easily see myself, teammates, many others just spinning things up on occasion to get a better feel for an idea than otherwise would have been possible. It’s trivially easy to do. (Challenging to hone it just right, but easy to jumpstart.)
- Using generative site building is also an arguably bad idea. As in using them for MVPs and launching to real customers by those who just don’t care at all about consequences. There’s plenty of folks running for the next shiny thing, maybe legitimately scared they’re going to miss their target market if they don’t launch something by lunchtime. And maybe they’re right. However, I’d argue that if your timeframe is that tight, chances are you don’t have much marketplace protectable in the first place. Not in terms of strategy, intellectual property, probably not in terms of network effects, or anything else original. But, there’s also those that argue that doing whatever you have to for a first go is ok.
Bottom Line? These are just tools like any others. Powerful? Absolutely. They’ll be useful and highly valuable for those that wield them well, and dangerous as hell for others, (from producers to customers), when used with reckless abandon. I’d say, “Let the games begin,” but clearly, that starting gun went off maybe a year or so ago as the most modern Gen AI builders started to gain real traction. Let’s enjoy what’s next.