I upgraded my Claude account to Max x 20 to take advantage of Fable being available under the subscription model until June 22nd before it’s removed to pay as you go (meaning I won’t be using it again until it comes back into subscription). It’s really absolutely fascinating to see it in action and it is already making me sad to lose it on the 22nd although they have said they will bring it back into subscription options as soon as they can. (I calculated I’ve spent well over $500 CAD on AI in the past month or so under the guise of experimentation but of course that isn’t sustainable, until it starts making me some money). I’ve only been doing fun / creative stuff so far..as Claude is already so good with documents so for run of the mill stuff I’m not sure you would be able to tell the difference. But for my creative / gaming / fun let’s see what it can do brain it’s actually remarkable. As I have a ton of credits to get through (although at this pace I won’t make it through to next Thursday when it resets with any left) it’s allowed me to experiment in a major way. And the beauty of Claude code is if you have the availability of credits you can be coding as many apps as you want simultaneously. So I used the following prompt on 6 games that I’ve previously tried to clone as a starting point to then twist them into something different for my Anomaly Arcade project. In all previous cases the results were never very usable without a ton of tweaking / and normally they are so far off the mark it wouldn’t be worth it. When you would end up with something playable like with simple games like Galaga you would get a working shoot em up that might belong from that era but it certainly wouldn’t feel at all like the original in the slightest, sharing just a tiny piece of DNA. So the prompt is at the bottom of this message. Very simple. And I used it in 6 games just changing the name of the game for each prompt. And I have run them in Opus 4.8 and also Fable 5 on High settings (which is the middle option). Opus would generally create something playable, you might recognize it as a distant cousin to what we were looking for, sometimes it would be playable but be nothing like the original in any way. But very hit or miss. If you got very lucky it may give you a starting point. More complex games like Outrun have no chance at all of getting anything playable on first go let alone similar to the original. Fable on the other hand scores 6 for 6. It obviously did its research very carefully in the case of outrun it found a repo where someone reverse engineered the code. However , as its guardrails won’t allow it to use that it just took all the key values if needed to recreate the overall feel and structure but every sound, graphic, level layout etc is original of its own creation inspired by the original. Or as it calls it “a tribute”. 😁

Nathan Campbell asked in another post how to best get started with game development, especially when using AI tools. That question stuck with me, because I think a lot of beginners run into the same trap: They start polishing graphics, assets, menus, sounds, effects, and tiny details way too early. So I made this small workflow infographic as a reminder. My main takeaway: Don’t try to make the game beautiful first.Make it clear.Make it playable.Then make it beautiful. Curious how you all approach this:At what point do you usually start caring about final assets and polish?

Hey, I see their ads quite offen, did anyone try it out? Is it worth to take a closer Look? https://www.seeles.ai/ Looks a bit scammy to me with their AI generated images of games you will not be able to just prompt it.

I jumped in right away, this single feature is actually something I was building for myself. It really feels like Anthropic has their dev's reading my mind now. This is I think the 2nd feature recently released that literally is a solution to something I was in the middle of building out on my own.

xAi partnerships with Anthropic is new. Anthropic gets more compute power and they doubled quotas for Pro/Max tier subscriptions. xAI recently partnered with Cursor too, this looks good for Grok to get all that data on coding from Cursor