What Was Sploder?

If you went to sploder.com at any time between May 2024 and April 2026, you would've been met with an error page. This was the final death blow of a platform that had existed in some form since 2004, and which was publicly available since 2007.

Sploder was a site where you could make and publish small games online using various Adobe Flash based editors. In its heyday, it was nice. Making and sharing games was simple, easy, and fun. And there was no shortage of games to play — when Sploder died, there were 4.44 million games [1]. Sure, most of them were the basic "Little Timmy's first game" sort of deal, but there were plenty of gems in there.

Of course, by the time the death of Adobe Flash came around at the end of 2020, the site's creator (who I will refer to using the username he uses today), neurofuzzy, had not updated the editors to HTML5. With the writing on the wall, he disabled new game uploads, and that was that for Sploder. Over the next few years, the site lingered as a frozen time capsule, now dysfunctional, until it finally fully weathered away with the full takedown in 2024.

Sploder was gone.

...

Until it wasn't.

Zombification!!

On June 15, 2026, I was randomly searching around stuff related to Sploder. I forgot why exactly, but I think it was because I wanted to see if there was anything new about running Sploder games on Linux easily (because right now, if I wanna play my old Sploder games, I gotta boot up a Windows VM and play them there).

So I end up on the Sploder subreddit, because at least I can use an alternative frontend for Reddit (and not Discord, which is seemingly the only other place where concentrated Sploder discussion is ocurring in 2026). And that's when I see this post [2]:

Geoff released new sploder

https://beta.sploder.com It's not the original creators but it's definitely something...

Ok... interesting... but it's not official, right? I mean that's what the post says—

...but I will just send this https://bsky.app/profile/shadross.bsky.social/post/3mkzyt5jfl224

Huh. So it is official. This is an official revival. Cool! Let's see what it has to offer—

New Sploder haha.. It's all AI so the UI isn't the best, though..

Oh no.

---

I cautiously open up the site... [3]

At this point, I'm thinking ok. This kinda looks AI generated but I can maybe give it the benefit of the doubt, because if it was generated then it isn't obvious at this first part of the page. And then I scroll down and immediately I'm met with stuff that looks like this:

And so, I am faced with the horror. The self-advertising tone. The em dashes used for embellishing points. The text that becomes meaningless when it is looked at closely. The random widget on the right that looks straight out of every other LLM generated site fitting in here as well as chocolate fits with spaghetti.

The only hint of anything human made is the logo, which admittedly looks alright and fits the vibe of "Sploder but modern." Other than that, it's all AI.

Sploder has been brought back as an LLM-ridden zombie.

But What Did It Look Like Before?

I have to imagine that if you're here you probably haven't seen the original Sploder. So, to provide context, this section will show what Sploder looked like back then.

Before I start, I should clarify one thing. Yes, I know that New Sploder only got its start about a month before writing this, so it will be less feature-rich if nothing old came back. However, this isn't my main concern. I could accept a smaller scope for a revival if it was crafted by hand, and not just entirely made with LLMs. Because of this, I won't be pointing out where the new site lacks features compared to the old — rather, I will point out where details in the original site were lost from the usage of AI. (Though I will still point out some new features or lack thereof which I find strange)

Anyways, here's the homepage:

The difference between the old homepage and the new homepage is night and day. Rather than just having two visible buttons and a bunch of text that doesn't convey what the games actually look like, this homepage invites the user to check out what has been made and to be inspired to make something themself. It shows not only what games will actually end up looking like, but also highlights the users that make them.

Of course, the site also has its own style. Rather than looking like literally every other site, elements have been made to look unique. It is reasonably legible, and although it doesn't work on mobile, neither does the new site. Sure, this is a result of not being fundamentally changed since 2007 (which is another reason why it doesn't work on mobile), but it's not like something with a unique style can't be made in 2026 (just look at my concept for a possible social media site).

But for what it's worth, just look at the two sites. Which one is more inviting? The one that shows off human made creations, or the one that tries to convince you to click on one of the two buttons at the top using LLM-speak?

Looking Deeper at the New Site

But that was just the homepage. What do the other pages on the site look like?

Let's start with the catalog page, which you get to after clicking that little "Explore the Catalog" link right at the top of the homepage:

In my opinion this is the saddest page on the site. It's literally all just LLM text, complete with the tone that makes me wonder if neurofuzzy even looked closely at this page at any point. Not much more to say here.

Moving past that, let's see what the editor actually looks like. I got to this page by clicking "Work Offline":

We'll get more into those sample games later. For now, let's just focus on the UI. Unlike the old site, where the editors were Flash apps inside the main HTML window, here the entire site is the editor. And by the entire site, I mean the entire site — there's seemingly no way to browse games others have uploaded (at least, without an account, but I don't want to make an account...).

It isn't as clear here as it is with the main pages that LLMs were used here. There are a few possible signs: The generic looking UI seems out of place for Sploder and the little number tag next to "Sample Games" especially has the vibe. However, after clicking that little help button on the top right, we can enter an "Animated Overview" of the editor, which looks like so:

This is more clearly LLM generated: the random glowing dot on the top left, the grid background behind the left side of the content, and the LLM language all show signs of it. If I had a thought that maybe the editor had at least a bit of human intervention, it would've been crushed by this fucked up little thing.

Unfortunately, even in the editor, it doesn't fare much better once we go to another tab:

This just isn't good when it is looked at more closely. Like what's up with that bottom bar? It just duplicates the name of the tab again as well as the entity count (with yet another random dot). The whole row of info in the bottom right is a bit strange, and I'm unsure how helpful it is to see that info on the main project page.

By now, you probably get the point. It's all LLM generated code, and as a result the website is generic and uninviting. But there's one aspect of it that baffles me. Remember those sample games? Here's one of them being played:

Without any sort of AI integration into the editor, it's hard to tell if the sample games were made by AI or not. But they're functional enough that I think that these were made by neurofuzzy himself.

One thing to note that isn't in the screenshot, however, is the lack of music. None of the sample games have it, which is weird. I'll talk more about this later.

With this, I received a curveball. Neurofuzzy must've cared at least somewhat about this New Sploder to make some games with it by hand. He must've cared at least a bit to bring back Sploder at all. But to do it with LLMs is another thing, one that begs the question...

How Much Does He Care?

From what I gather online, I have come to two conclusions.

My first conclusion is that neurofuzzy still cares a lot about Sploder. Based on past Bluesky posts [4] it is clear he still thinks about it quite a bit. Sploder was his creation, the site that he was the head of for 13 years, and whose forums he was in for a bit longer than that.

My second conclusion is that he doesn't have significant qualms with LLM usage in general. Once again, going off his Bluesky posts, over time while it seems he has grown to recognize that there are issues with LLMs, he doesn't see the fundamental issues of using them [5].

From these, I have to assume that he cares about Sploder, but he isn't too aware of just how bad LLM usage is.

But of course, he has other interests too, and these seem to be more important to him now. And I know as well as he does that sometimes other interests get priority over software creation. But taking a shortcut with LLMs is not the way to go about a revival if this is something that he cherishes. Sure, he may not be fully aware of the fundamental concerns of simply using generative AI, but bringing Sploder back by using it all over does lead to some bad consequences that should be considered.

The Impact

Aside from the issues with using LLMs, I can't help but think about what this could mean to the kind of internet user who would visit this kind of site.

I first discovered Sploder in 2013, back when I was very little. My games on Sploder were some of my first ever games, and I cherish them greatly. What if someone who is the same age I was back then found this New Sploder today? They wouldn't be met by an inviting website full of fun games, they would be met by a registration wall, an overly complex editor, and a bunch of AI generated text. It would be uninteresting to them, they'd go away, and a potential formative experience would've been lost.

It kinda points to a bigger trend of everything new on the internet always having the risk of being tainted by generative AI usage. How is one supposed to know if their tool is worse or less ethically sourced without the experience of being able to tell if it was made with it?

And because it was made with LLMs, it will be designed as the LLM thinks of it. As a result, this editor ended up more as a tool than a fun experience. All the personality and ease of use the original editors had is lost here.

Overall, I am left disappointed. This is no way to treat something that so many loved. It would've been better to have had Sploder stay dead.

Some Final Notes

• My argument here was based on some best educated guesses based on prior knowledge and a small bit of research. While I am somewhat confident I am accurate in my assessment of how much LLMs were used in the making of the New Sploder site, it could be the case that they weren't really used. If this is the case, I will update this article.
  • One interesting thing I found that could help in making this prediction more accurate is looking at this repo for a SFX generator tool neurofuzzy made, as well as the corresponding demo. The demo site seems to have very similar if not identical capabilities as the instrument creation tool in the New Sploder editor, so I have to guess that some code was reused there, in which case an analysis on LLM usage in this tool would probably have some indication for the New Sploder editor.
• I left my qualms about the lack of music in the sample games until here, and that's because this concern is less related to the fact that New Sploder uses LLMs and more the design of the editor. That is to say, why is there a music editor at all? Sure, having a music editor is nice for a game creation tool, but in the original Sploder, the editors were focused on building games and building games only. You could make a game without making any assets, and that's what made it so easy. Sure, the Physics Puzzle Editor had the Sprite Editor, but that was separate — you could make do with the various assets given to you in that editor itself. But my main concern is the fact that you have to make music yourself, and that's because that's hard! Even for me today, making music is hard for me; imagine how hard it would've been for 2013 me...
  • And putting on my nostalgia glasses, damn it I really adored all the Amiga music that you could select from in the original Sploder. It really helped to give Sploder games their own vibe, even between editors. It might've been limiting, but sometimes limits are good for creativity.
• One big caveat with this article was that I didn't actually log in. And yeah, that is a big concern, but I just didn't want to bother. I barely wanted to touch the site anyways, and the fact that you can log into New Sploder with Discord or Google (as well as email) gave me a few concerns about registration. Perhaps if I take a deeper dive I might try it, but for now I'll gladly leave this Sploder behind once more.
• Ending things off with some fun: If you happen to have been a Sploderian back in the day, here's a phrase that can sum up my opinions on the New Sploder: splode this.

---

Footnotes

1

I got this number from https://web.archive.org/web/20210117115452/https://www.sploder.com/games/4441140/. This is a snapshot of Sploder in 2021, just after game uploading stopped. The last number in that URL is an approximate count of all the games in Sploder: it counts up from the newest games to the oldest games.

2

This is the original link. I recommend you view it with an alternative frontend like Redlib.

3

I opened up beta.sploder.com first, and later I discovered the changes were live on the root domain sploder.com. The sites seem identical, apart from the small "beta" icon on the top of the beta page. This article only goes through the main sploder.com.

4

These all can be found on neurofuzzy's Bluesky page, which I found with the aforementioned Reddit post. I'd like to highlight specifically a string of posts from February 23 of last year, which includes Showcase 1, Showcase 2, and Showcase 3, as well as this thread containing a bunch of footage of an unreleased Classic Sploder editor.

5

If you're here, you probably know this already, but to those who don't know, here's a link to a bunch of fancy academic articles with a nice bullet point summary in between the article links,

But here's a common-sense TL;DR: AI (or more specifically, the generative AI that's been pushed by big tech companies since 2023) generally fucks over communities and the environment where data centers are being built, and these companies want to build a fuck ton of them everywhere. It fucks with people because having access to a "yes-man" constantly can affirm dangerous beliefs, and it fucks over creative people by making plagiarism trivial.

For programming specifically, using AI is like putting lead or asbestos in there. Sure, it can work wonderfully, but there are many issues that will only surface later, not being immediately apparent, which then blow up in your face.