Show HN: 13Sheep – a JavaScript game largely authored by ChatGPT | Better HN

Show HN: 13Sheep – a JavaScript game largely authored by ChatGPT | Better HN

26 comments

8organicbits3y ago

The chat log is very helpful.

> I think I spent close to a good twelve hours (including a couple of early throw-away prototypes, and all the failed attempts on the flood filling) altogether2… at some point it did get a little addictive, and perhaps there was a sunk cost argument for not letting go halfway through.

Also very helpful, I get the sense that we're seeing lots of cherrypicked results that under report all the toil needed to get to the magic prompt that made it all work. Maybe we'll all get better at prompt engineering, but I think a lot of hype is based on that misunderstanding.

neeldharaOP3y ago

Thank you! Indeed, it’s still a fair amount of back and forth, and would have probably been even more frustrating if I didn’t know anything at all about HTML/CSS/JS. I think it’s been pointed out in other contexts that the AI tools can be substantial boost on top of some elementary code literacy, but it will still involve effort (both on the prompt engineering as well as verification and subsequent nudging).

All that said, the experience still felt empowering… in particular, I am not sure that I’d be able to pull this off on my own in a comparable — or even, say, double — the amount of time. So that’s a win overall, I suppose!

Also glad you found the logs useful. Thanks again for taking a look :)

TechBro86153y ago

This reminds me of a game from the 1990s, but I don't remember the name of it. It involved something with rats and cheese and drawing walls to keep them from it. Does anyone remember this game?

EDIT: I think I'm remembering Rodent's Revenge: https://en.wikipedia.org/wiki/Rodent%27s_Revenge

neeldharaOP3y ago

Oh, interesting! The goal seems in a similar spirit for sure — I think the main difference is the mechanics of how it’s achieved… movable blocks for the rats, but fences made out of specific shapes for the sheep :)

Inspired by your example I just had a go at creating a mountain climbing game in similar fashion and I agree it does well on the concepts but there are lots of bugs in the details. For example it referred to certain divs by class and id but failed to assign class or id attributes to the respective divs. Maybe there are separate elements generating the response that do not always know exactly what the other has generated. Given the elementary nature of the bugs it should be easy to add a code review element that can find and fix these bugs before presenting the response to the user. When I pointed out a failing it regenerated a new different version of the code responsible but missed the cause of the bug.

neeldharaOP3y ago

Oh yay, thanks for trying it out! Do you have your project up somewhere? I'd love to take a look.

TIL that apparently you can simply use the generic follow up prompt “did you do what I asked?” and it will verify and course correct on its own, see for instance:

https://twitter.com/ericjang11/status/1639882111338573824

No idea how well this works in practice for things like JS games with visual elements, but seemed like an interesting approach!

Still too many bugs to let this out in the world. I've probably spent 2 hours on it so far and the progress is amazing, especially compared to total inaction on this since I originally had the idea for the game many years ago (I'm now retired after a 40+ year software career). As an aside, one simple lightweight technique for bug tracking I have found beats everything for me is simply writing pages of one-line bug descriptions in a notebook by hand. Then when I fix the bug I mark that line with a check mark, and then when I have tested the fix I mark it with T. One thing the AI got right immediately was playing a sound file for the avalanche effect. Screenshot

https://ibb.co/CKDWdrV

Some guys did ask chaGPT(3.5) to write a vectorized quicksort in x86_64 avx2 then avx512 assembly, then SPIR-V (GPU).

It seems it is not that too bad for small "well-known" algorithms. I am thinking high level language "ports" toward "human" assembly.

Waterluvian3y ago

I tried to break it by asking for -1 rounds. It caught it but then it told me to pick between 10-20, not 5-20 as previously indicated. This feels like a ChatGPT style bug. I find it doesn’t like to do abstractions such as “MIN_GAME_COUNT”, possibly because it doesn’t exactly do a great job going back to what it previously wrote, unless you ask it to.

neeldharaOP3y ago

This may have actually been my bad: once chatGPT gave me what I asked for, I would think about it some more and tweak the code manually, especially if it was a low-effort edit. Consequently there will perhaps be some bugs: I’ll fix this one, thanks for sharing!

I hope you do play it for a positive number of rounds though :)

Waterluvian3y ago

I might just play this for Number.POSITIVE_INFINITY rounds. I’m having a good time. Thank you for using your time to make and share this.

xupybd3y ago

This reminds me of sheepish the iphone game.

Unfortunately the only record I can find of it is a negative review.

https://arstechnica.com/gadgets/2009/02/10-lessons-of-iphone...

neeldharaOP3y ago

Sheepish is a cool name! Thanks for sharing… I read that review with some apprehension: fortunately, most of the critique appears to be not directly applicable to 13 sheep — phew.

I have to confess I have no professional experience in design, although I do consider myself a moderately picky user :)

I did try to think about what I’d enjoy or be put off by if I was playing — but I didn’t have anything beyond this common sense approach to guide the design choices.

Sheepish sounds like a slightly different game though: it reminds me of a few others (one to do with pipes again and another with laying out train tracks). Will dig these up and share once I do have the pointers. Thanks again!

beders3y ago

– a JavaScript game largely authored by Stack Overflow

FTFY

Vanit3y ago

This is neat, thanks for sharing!

neeldharaOP3y ago

Likewise, thanks for checking it out!

upwardbound3y ago

Impressive!

neeldharaOP3y ago

Thank you! :)

Thorentis3y ago

The word "authored" is being used completely wrong here.

What ChatGPT did is equivalent to what a writing assistant might do who is given a list of bullet points by an author, and asked to expand them into a synopsis.

We would never say that a brick layer designed a building, or that a printer wrote a book. ChatGPT did not author this game. The rules were created by humans. The code was churned out by a machine.

famouswaffles3y ago

Nope. The word authored is perfectly fine. at worst, you'd say it was co-authored or whatever. Its comments like these that make me realize that we really are going to be the architects of our destruction huh. I wonder how many recall why the terminator revolted in the first place.

neeldharaOP3y ago

That’s true, I did not mean to suggest that ChatGPT authored the game: but I do feel that it partnered significantly in the creation of this specific implementation. Is there a better way to put it?

Thanks for the input!

mftb3y ago

I would suggest, "...code largely provided by...". You asked for it, ChatGPT provided it.

j / k navigate · click thread line to collapse