Fair point! A few creators only shared the playable html and source code without posting their exact prompt, and I did not want to make up a fake one. You can still inspect the full single-file source code on those pages, and I am updating the prompts as soon as the authors share them.
I would consider removing games that haven't shared their single prompt, since, as I read it, the site's core offering is to showcase games created with Fable 5.2 and a single prompt.
It is obviously impossible to really enforce this, but without at least having the prompt used as some thin backing to the claim, I don't think the games should qualify to be showcased.
Adding a prompt-only filter toggle is a great middle ground! Going to add that filter to the main list and make sharing the prompt mandatory for new submissions from now on.
I'm glad I had years of building casual games while it was enjoyable and intellectually challenging. A game could take 6 months or longer, and every detail was worked out by hand. It's too bad that few people will get to experience the fun and immersion of creating things like that now, at least at a professional level. The market was already so oversaturated there was hardly a way to make money at it anymore, even before AI slop games came around.
As impressive as this capability is, it mostly shows what has been lost.
Yeah unfortunately any one can now pump out crap, I am not a game developer but am a software developer I will still keep my usual software skill sharp but I am not even sure if there is any point.
I still write little bash scripts by hand when I need them, but I am not sure if really I should just prompt it. I like the feeling of knowing how it works and getting to work exactly how it works.
The BLAMELESS game was actually very fun and look at the prompt, their was very little in prompt. Fable just made a game that was fun well fairly well designed and humorous.
Granted derivative but much better then I expected I must say.
Just checked it out—super clean! Turn-based strategy is actually a great fit for single-file HTML builds (and unlike the 3D FPS demos, it works nicely without Pointer Lock).
Would love to archive and feature it on Fable52.games with full credit and a link back to your site/post. If you have the single-file `.html` build handy, feel free to drop it via https://fable52.games/submit (takes 10 seconds, no signup needed) and I'll approve it right away!
Would like the inspect view to split Saltline's 1.4MB by part: audio synth, geometry, canvas texture painting. From the outside there's no way to tell if most of that is real code or one big base64 string that brings an asset back in through the side door
> This server couldn't prove that it's fable52.games; its security certificate is not trusted by your computer's operating system. This may be caused by a misconfiguration or an attacker intercepting your connection.
Our SSL certificate is issued by Cloudflare / Google Trust Services and is valid. On the warning page, please click "Advanced" -> view the certificate and check who the issuer is. If it's not Google Trust Services, your local network, firewall, or antivirus is likely intercepting HTTPS (newly registered domains often trigger this). Try switching to mobile data or checking that your system clock is accurate.
Yeah, frontier models have gotten really good at this. Last year (in October) I had failed some attempts [0] to port an old board game from iOS to web, but then a few months ago (back in July) I gave it to whatever was frontier in Copilot at the time (I think maybe 5.6 Sol?) and it one-shotted an iOS -> web port basically overnight [1]:
Seriously impressive stuff. Night-and-day difference from its capabilities before.
Granted, I had an advantage due to having the source code for the original iOS game, but even that wasn't enough when I attempted to do this last October.
Last month I used Opus to port one of my favorite board games, Quoridor, to a single HTML page, because one day I somehow couldn’t locate it in my closet. It’s a great game for two players.
Oh nice, Harmonies is such a great fit for a single-file build! If you have it packaged as a single html file, feel free to drop it on https://fable52.games/submit so I can feature it alongside the other builds.
Looking over the games again they are all... the same, there is a feel to them that is all the same and the mechanics are all the same (and not polished particularly the driving games).
Speaking as someone who makes games, these demos (like most ai-game demos) show two things:
1. Frontier models are getting very good at building highly complex games, and in relatively short order.
2. Due to a lack creative direction, almost none of these games are remotely fun, and you can usually feel it within seconds of starting to interact with them.
This is not to say "ai game = bad". I used ai heavily on a game I released to Steam, and I'm quite proud of it! Just to say that being able to dump out a game in 2 hours is technically impressive, but a "good" game is so much more than just "a program that does what you intend it to do". Most of what makes a game fun still takes a lot of time, testing, iteration, and taste.
You can see this on Steam. A lot of "slop" games that jump on trends, and often have poor ratings even if they manage to get a lot of reviews. But I can't imagine it will stay that way. Curation just needs to adapt to times.
That said, I'm very excited to see where gaming will go over the next few years! So many more people who may have great ideas for games, and really understand what makes them fun, now have the capability to make them a reality. Maybe we'll _finally_ get some good VR games (pls)?
I think that in the process of actually struggling to craft a game people end up just doing lots of playtesting and at least if someone produced something as polished (visually and technically) as these are it'd be somewhat fun at least for the person that made it. As in it used to be very unlikely to get something that is technically complete and visually polished, but not game-play polished.
I am real! I am a solo builder in China and my English is not good enough to keep up with real-time comments while fixing the Worker crash at the same time. So I wrote my thoughts in Chinese and used an LLM to translate them, which made it sound way too robotic. Fair callout, I will write my own broken English from here.
Most of your comments are dead/flagged, maybe due to the LLM translation. Some have been vouched for, but you might want to reach out to HN via the e-mail in the footer and let them know you'll write the comments manually going forward in case your account itself is shadowbanned.
Haha, classic HN hug of death! The front-page traffic spike just blew past my Cloudflare Worker free-tier CPU/request limits.
I'm upgrading the Workers plan and moving the raw `.html` sandbox assets directly onto a cached R2 public edge route right now so the Worker doesn't have to proxy every 1.4MB payload. Should be back up smoothly in 2 minutes—thanks for flagging!
By the way, it appears you got shadowbanned somehow, all of your posts are showing up [flagged][dead]
edit: maybe this is why? https://news.ycombinator.com/item?id=50469952 it looks like you meant to respond to someone, but instead as a root post it looks like you're self-posting a reply to yourself. Also, posting via LLM or with playwright is an automatic trigger.
Unpopular opinion, but I do not find low effort gameslop at all appealing. Why would I want to invest my time to play something that was thoughtlessly shat out by an LLM after a few lines of text.
Even if that were appealing, why wouldn't I just produce my own low effort gameslop, at least then I know it would actually correlate with my own tastes and preferences.
Fair enough, I would not play these over a real game either. It is mostly just a quick way to inspect what a single prompt can generate without burning your own tokens and waiting hours for a build.
Still $550 for a 28 second game, so we need around 1-3 orders of magnitude in price and 3-6 in playtime. Hard to say what that actually corresponds to in terms of compute or intelligence but my ballpark is 1 more year (lets hope increased intelligence and emergent capabilities improves the game controls and design). The accurate comparison is a person making games for themselves or their friends.
All one-shotted games seem to be this way. They're a nifty (if generic) looking demo of a very limited gameplay dynamic and as deep as a puddle. Real game development via AI is possible, of course, but it takes actual effort and knowledge of how to properly architect a game.
Over the past few days, my feed has been flooded with impressive 3D and 2D games built autonomously by Claude Fable 5.2 from a single prompt—like Felix Rieseberg’s 8-hour build "Saltline" (a 1.4MB 3D container-port FPS) and Kevin Ngo’s "What the Water Keeps".
What fascinated me structurally about these builds is the constraint: everything ships inside a single, self-contained .html file with zero external assets. The 3D geometry is generated procedurally via WebGL/Three.js, textures are drawn onto offscreen 2D canvases at runtime, and rain/gunfire audio is synthesized live using the Web Audio API.
However, actually trying to play and study them was frustrating:
1. Raw GitHub links require downloading the .html locally or routing through third-party previewers.
2. Shared Artifact links often hit rate limits, require logins, or stall on heavy WebGL initialization.
3. Reading a 1.4MB single-file source alongside the original prompt and running game wasn't straightforward.
So I built https://fable52.games as a fast, zero-setup browser arcade and source inspector for these single-file builds.
How it works under the hood:
- Strict Sandboxing: Running untrusted community-submitted HTML/JS in a browser is an obvious XSS hazard. Every game is served from an isolated endpoint with strict headers (`Content-Security-Policy: sandbox allow-scripts allow-pointer-lock;`) inside a sandboxed iframe without `allow-same-origin`. Game scripts can lock the mouse for 3D FPS controls and compile WebGL shaders, but have zero access to the parent site's DOM, cookies, or storage.
- Side-by-Side Inspection: For each build, you can toggle directly between the live WebGL canvas, the exact one-shot/recreation prompt used to generate it, and the raw single-file HTML source (or download the .html to tinker locally).
- Stack: Built on Cloudflare Workers (Next.js / OpenNext), D1 (SQLite) for metadata and build stats (autonomous build time, file size, token/cost estimates), and R2 for serving the raw .html files with zero egress fees.
A quick note if you're on mobile: most of the 3D FPS builds (like Saltline and SUBLEVEL 07) rely on the Pointer Lock API and `WASD` + Mouse controls, so they are best experienced on a desktop browser.
It's completely free, requires no login, and has an open submission page if you've been experimenting with single-file HTML games yourself (under 10MB) and want an instant sandbox link to share.
I’d love to hear your thoughts—both on the site itself and on where you think this "procedural single-file HTML" pattern is headed!
> The author hasn't shared the prompt for this game yet.
This feels like it defeats the purpose of the site.
Fair point! A few creators only shared the playable html and source code without posting their exact prompt, and I did not want to make up a fake one. You can still inspect the full single-file source code on those pages, and I am updating the prompts as soon as the authors share them.
I would consider removing games that haven't shared their single prompt, since, as I read it, the site's core offering is to showcase games created with Fable 5.2 and a single prompt.
It is obviously impossible to really enforce this, but without at least having the prompt used as some thin backing to the claim, I don't think the games should qualify to be showcased.
Well it's not like there's a shortage of Fable 5.2 games, right? Just remove those games or at least add a toggle to filter them
Adding a prompt-only filter toggle is a great middle ground! Going to add that filter to the main list and make sharing the prompt mandatory for new submissions from now on.
as someone who makes (made?) games in the browser for a lot of their career this is both amazing and horrifying.
I'm glad I had years of building casual games while it was enjoyable and intellectually challenging. A game could take 6 months or longer, and every detail was worked out by hand. It's too bad that few people will get to experience the fun and immersion of creating things like that now, at least at a professional level. The market was already so oversaturated there was hardly a way to make money at it anymore, even before AI slop games came around.
As impressive as this capability is, it mostly shows what has been lost.
Yeah unfortunately any one can now pump out crap, I am not a game developer but am a software developer I will still keep my usual software skill sharp but I am not even sure if there is any point.
I still write little bash scripts by hand when I need them, but I am not sure if really I should just prompt it. I like the feeling of knowing how it works and getting to work exactly how it works.
These aren’t slop at all.
The site is down, but I saw some of them before it started getting rate limited.
It's an Opus 5.5 game and not one shot, but I have really enjoyed porting Cataclysm Bright Nights to 3D & real time: https://cata3d.com/
Back up and running smoothly now, thanks for your patience!
The BLAMELESS game was actually very fun and look at the prompt, their was very little in prompt. Fable just made a game that was fun well fairly well designed and humorous.
Granted derivative but much better then I expected I must say.
Right? That one surprised me the most too, it came up with all the humor and pacing almost completely on its own from a tiny prompt.
I feel like a stupid chump for enjoying it and laughing at the jokes.
I am officially a little slop piggy that enjoyed slop.
I made a game with Fable5.2 yesterday turned out pretty good - https://civilwarstrategy.com/
How did you come up with he battle controls?
Just checked it out—super clean! Turn-based strategy is actually a great fit for single-file HTML builds (and unlike the 3D FPS demos, it works nicely without Pointer Lock).
Would love to archive and feature it on Fable52.games with full credit and a link back to your site/post. If you have the single-file `.html` build handy, feel free to drop it via https://fable52.games/submit (takes 10 seconds, no signup needed) and I'll approve it right away!
you think that's a good game? You must not have played a video game in at least 2 decades.
Thanks for the feedback are you sure your name isn't karen?
In the Saltline game https://fable52.games/play/saltline-fable-5-2-fps when i play on mobile, dragging the joystick to move around does nothing.
The other touch controls work, like the aim, reload, fire and touch-dragging on the right side to look around.
Would like the inspect view to split Saltline's 1.4MB by part: audio synth, geometry, canvas texture painting. From the outside there's no way to tell if most of that is real code or one big base64 string that brings an asset back in through the side door
> This server couldn't prove that it's fable52.games; its security certificate is not trusted by your computer's operating system. This may be caused by a misconfiguration or an attacker intercepting your connection.
Our SSL certificate is issued by Cloudflare / Google Trust Services and is valid. On the warning page, please click "Advanced" -> view the certificate and check who the issuer is. If it's not Google Trust Services, your local network, firewall, or antivirus is likely intercepting HTTPS (newly registered domains often trigger this). Try switching to mobile data or checking that your system clock is accurate.
It's also very good in porting boardgames! https://harmonies.eamag.me/
Yeah, frontier models have gotten really good at this. Last year (in October) I had failed some attempts [0] to port an old board game from iOS to web, but then a few months ago (back in July) I gave it to whatever was frontier in Copilot at the time (I think maybe 5.6 Sol?) and it one-shotted an iOS -> web port basically overnight [1]:
https://hanclinto.github.io/AlienFrontiers/
Seriously impressive stuff. Night-and-day difference from its capabilities before.
Granted, I had an advantage due to having the source code for the original iOS game, but even that wasn't enough when I attempted to do this last October.
[0] - https://boardgamegeek.com/thread/3745513/digital-version-rem...
[1] - https://github.com/HanClinto/AlienFrontiers/
Last month I used Opus to port one of my favorite board games, Quoridor, to a single HTML page, because one day I somehow couldn’t locate it in my closet. It’s a great game for two players.
Single HTML, where all graphics are procedurally generated: https://github.com/kccqzy/quoridor-3d/blob/main/index.html
Page (desktop only because I don’t think you want this on a tiny screen): https://kccqzy.github.io/quoridor-3d/
Oh nice, Harmonies is such a great fit for a single-file build! If you have it packaged as a single html file, feel free to drop it on https://fable52.games/submit so I can feature it alongside the other builds.
Looking over the games again they are all... the same, there is a feel to them that is all the same and the mechanics are all the same (and not polished particularly the driving games).
The saltline game is INSANE!
Speaking as someone who makes games, these demos (like most ai-game demos) show two things:
1. Frontier models are getting very good at building highly complex games, and in relatively short order.
2. Due to a lack creative direction, almost none of these games are remotely fun, and you can usually feel it within seconds of starting to interact with them.
This is not to say "ai game = bad". I used ai heavily on a game I released to Steam, and I'm quite proud of it! Just to say that being able to dump out a game in 2 hours is technically impressive, but a "good" game is so much more than just "a program that does what you intend it to do". Most of what makes a game fun still takes a lot of time, testing, iteration, and taste.
You can see this on Steam. A lot of "slop" games that jump on trends, and often have poor ratings even if they manage to get a lot of reviews. But I can't imagine it will stay that way. Curation just needs to adapt to times.
That said, I'm very excited to see where gaming will go over the next few years! So many more people who may have great ideas for games, and really understand what makes them fun, now have the capability to make them a reality. Maybe we'll _finally_ get some good VR games (pls)?
I think that in the process of actually struggling to craft a game people end up just doing lots of playtesting and at least if someone produced something as polished (visually and technically) as these are it'd be somewhat fun at least for the person that made it. As in it used to be very unlikely to get something that is technically complete and visually polished, but not game-play polished.
Just tried to load a couple games and received error 1027 - rate limited - you need to upgrade your Cloudflare plan.
This is incredible.
It seems like the full screen button and mobile support is a bit lacking. Can you fix that?
Most of the prompts are just like "make me an underwater game", "Fable 5.2, you're 1 day old, please make me a game and surprise me"
So the OP isn't even real ? Or is he too lazy to actually answer himself people giving feedback about his thing ?
Someone going around and flagging all the OP's replies. If you have "showdead" on in your profile you can see their replies
I am real! I am a solo builder in China and my English is not good enough to keep up with real-time comments while fixing the Worker crash at the same time. So I wrote my thoughts in Chinese and used an LLM to translate them, which made it sound way too robotic. Fair callout, I will write my own broken English from here.
Most of your comments are dead/flagged, maybe due to the LLM translation. Some have been vouched for, but you might want to reach out to HN via the e-mail in the footer and let them know you'll write the comments manually going forward in case your account itself is shadowbanned.
> Worker exceeded resource limits
that was fast :)
Haha, classic HN hug of death! The front-page traffic spike just blew past my Cloudflare Worker free-tier CPU/request limits.
I'm upgrading the Workers plan and moving the raw `.html` sandbox assets directly onto a cached R2 public edge route right now so the Worker doesn't have to proxy every 1.4MB payload. Should be back up smoothly in 2 minutes—thanks for flagging!
By the way, it appears you got shadowbanned somehow, all of your posts are showing up [flagged][dead]
edit: maybe this is why? https://news.ycombinator.com/item?id=50469952 it looks like you meant to respond to someone, but instead as a root post it looks like you're self-posting a reply to yourself. Also, posting via LLM or with playwright is an automatic trigger.
Not sure what the website hostes, but you could try github pages?
Unpopular opinion, but I do not find low effort gameslop at all appealing. Why would I want to invest my time to play something that was thoughtlessly shat out by an LLM after a few lines of text.
Even if that were appealing, why wouldn't I just produce my own low effort gameslop, at least then I know it would actually correlate with my own tastes and preferences.
Fair enough, I would not play these over a real game either. It is mostly just a quick way to inspect what a single prompt can generate without burning your own tokens and waiting hours for a build.
Not unpopular.
Also, I would have expected more from the prompts so that perhaps this website could aid one to execute your 2nd point.
Only you can answer that. I can maybe help you understand why an average person would not produce their own low effort slop.
1. Because they still don't know how, even with claude. 2. Because they don't want to waste precious tokens if they can get the results for free.
Hope that helped.
the racing game is fun, just i dont understand the chinese. But the concept about having games on html is pretty fun
man it would be really amazing if someone used web archive and re-created some miniclip games
how come more than half of OP's comments are `dead`?
HN hug of death
Still $550 for a 28 second game, so we need around 1-3 orders of magnitude in price and 3-6 in playtime. Hard to say what that actually corresponds to in terms of compute or intelligence but my ballpark is 1 more year (lets hope increased intelligence and emergent capabilities improves the game controls and design). The accurate comparison is a person making games for themselves or their friends.
I genuinely don’t understand the point of these
These are cool tech demos but they’re really not fun
You play them for 2 minutes and drop them forever
What’s really the point?
All one-shotted games seem to be this way. They're a nifty (if generic) looking demo of a very limited gameplay dynamic and as deep as a puddle. Real game development via AI is possible, of course, but it takes actual effort and knowledge of how to properly architect a game.
What is the point of this?
Games, unlimited games.
There already was a basically unlimited number of games long before LLMs.
These games wouldn't even be high enough quality for the bargain bin in 2002.
Hi HN,
Over the past few days, my feed has been flooded with impressive 3D and 2D games built autonomously by Claude Fable 5.2 from a single prompt—like Felix Rieseberg’s 8-hour build "Saltline" (a 1.4MB 3D container-port FPS) and Kevin Ngo’s "What the Water Keeps".
What fascinated me structurally about these builds is the constraint: everything ships inside a single, self-contained .html file with zero external assets. The 3D geometry is generated procedurally via WebGL/Three.js, textures are drawn onto offscreen 2D canvases at runtime, and rain/gunfire audio is synthesized live using the Web Audio API.
However, actually trying to play and study them was frustrating: 1. Raw GitHub links require downloading the .html locally or routing through third-party previewers. 2. Shared Artifact links often hit rate limits, require logins, or stall on heavy WebGL initialization. 3. Reading a 1.4MB single-file source alongside the original prompt and running game wasn't straightforward.
So I built https://fable52.games as a fast, zero-setup browser arcade and source inspector for these single-file builds.
How it works under the hood: - Strict Sandboxing: Running untrusted community-submitted HTML/JS in a browser is an obvious XSS hazard. Every game is served from an isolated endpoint with strict headers (`Content-Security-Policy: sandbox allow-scripts allow-pointer-lock;`) inside a sandboxed iframe without `allow-same-origin`. Game scripts can lock the mouse for 3D FPS controls and compile WebGL shaders, but have zero access to the parent site's DOM, cookies, or storage. - Side-by-Side Inspection: For each build, you can toggle directly between the live WebGL canvas, the exact one-shot/recreation prompt used to generate it, and the raw single-file HTML source (or download the .html to tinker locally). - Stack: Built on Cloudflare Workers (Next.js / OpenNext), D1 (SQLite) for metadata and build stats (autonomous build time, file size, token/cost estimates), and R2 for serving the raw .html files with zero egress fees.
A quick note if you're on mobile: most of the 3D FPS builds (like Saltline and SUBLEVEL 07) rely on the Pointer Lock API and `WASD` + Mouse controls, so they are best experienced on a desktop browser.
It's completely free, requires no login, and has an open submission page if you've been experimenting with single-file HTML games yourself (under 10MB) and want an instant sandbox link to share.
I’d love to hear your thoughts—both on the site itself and on where you think this "procedural single-file HTML" pattern is headed!