I designed the wedding ring for my wife, so I had a specific ring in mind to see how well it would do. I used the prompt:
A titanium ring with wavy rails on the sides, incorporating flowers in the middle of the band, where the wavy rails move farthest away from each other. The flowers should be a lily, sunflower, and then a spot for a gem, followed by a lily, sunflower, and then a spot for a gem and then a sunflower. The flowers should be evenly spaced around the whole band.
And it got the basics right, looked good, but it put the gem settings between the flowers, where the wavy rails were closest together. I tried several attempts at putting prompts in the "How should we modify this?" box such as "No, the gem settings should be where the flowers are, but every other flower" and "Don't put the gem settings at the points where the wavy edges are closest together" and it just failed hilariously. No matter how I phrased it, it kept putting the gem settings where the rails were closest together. After three (or it might have been four) attempts I gave up.
I could have explained the entire ring to a jeweler in a short paragraph, and any issues with it could be fixed in a sentence or two. I'm not sure why I need a frustrating middleman. And don't feel bad, this applies to nearly everything I use built on LLMs.
It's a cute toy, but boy are LLMs not remotely ready for production anything. I didn't even bother to give it the other aspects of the ring because if it can't get one of the most basic things in four tries of exceedingly clear language, then why would I keep going?
Put the login gate after you generate a ring, not before. I bounced immediately because it's not worth creating an account without really knowing whether I would be interested. Let potential users invest a little time and start dreaming, so you can take advantage of loss aversion to get signups. Probably also worth adding an option more lowkey than creating an account (which I assume includes setting up a password etc.), like a newsletter that aggregates the best designs every week.
Also, it wasn't obvious to me that the examples in the "Need inspiration?" section were generated, or that I could tweak those designs.
Ha I only added this because my wife doesn't have gmail. I think this is a problem I don't particularly care about for now - I think the parent commenter is correct in saying that the account wall is in the wrong place, so if I move the wall to where the value is (speak to a jeweler) I can defer email verification for a while.
I asked it to "add a hamburger to the design" of a 'need inspiration' example and it served an image of a ring with a life like mini hamburger next to the jewel.
I designed the wedding ring for my wife, so I had a specific ring in mind to see how well it would do. I used the prompt:
A titanium ring with wavy rails on the sides, incorporating flowers in the middle of the band, where the wavy rails move farthest away from each other. The flowers should be a lily, sunflower, and then a spot for a gem, followed by a lily, sunflower, and then a spot for a gem and then a sunflower. The flowers should be evenly spaced around the whole band.
And it got the basics right, looked good, but it put the gem settings between the flowers, where the wavy rails were closest together. I tried several attempts at putting prompts in the "How should we modify this?" box such as "No, the gem settings should be where the flowers are, but every other flower" and "Don't put the gem settings at the points where the wavy edges are closest together" and it just failed hilariously. No matter how I phrased it, it kept putting the gem settings where the rails were closest together. After three (or it might have been four) attempts I gave up.
I could have explained the entire ring to a jeweler in a short paragraph, and any issues with it could be fixed in a sentence or two. I'm not sure why I need a frustrating middleman. And don't feel bad, this applies to nearly everything I use built on LLMs.
It's a cute toy, but boy are LLMs not remotely ready for production anything. I didn't even bother to give it the other aspects of the ring because if it can't get one of the most basic things in four tries of exceedingly clear language, then why would I keep going?
Put the login gate after you generate a ring, not before. I bounced immediately because it's not worth creating an account without really knowing whether I would be interested. Let potential users invest a little time and start dreaming, so you can take advantage of loss aversion to get signups. Probably also worth adding an option more lowkey than creating an account (which I assume includes setting up a password etc.), like a newsletter that aggregates the best designs every week.
Also, it wasn't obvious to me that the examples in the "Need inspiration?" section were generated, or that I could tweak those designs.
Thank you, this feedback is so so useful to me. Login gate when you ask to speak to a jeweller maybe?
> Probably also worth adding [..] a newsletter that aggregates the best designs every week.
Yep this is a great idea, thank you.
> Also, it wasn't obvious to me that the examples in the "Need inspiration?" section were generated, or that I could tweak those designs.
No, its not. I will work on this too, thank you.
Agree. FWIw you can quickly put any email/pw to make an account and it lets you through (i.e. no email verification)
Ha I only added this because my wife doesn't have gmail. I think this is a problem I don't particularly care about for now - I think the parent commenter is correct in saying that the account wall is in the wrong place, so if I move the wall to where the value is (speak to a jeweler) I can defer email verification for a while.
I asked it to "add a hamburger to the design" of a 'need inspiration' example and it served an image of a ring with a life like mini hamburger next to the jewel.
Lucky you, I asked it to design a ring like someone gutted a pig and wrapped the intestines around a finger and all I got was a lousy login prompt.
[dead]
Takes too long to load
It does! I had this running on a potato and I didn't realise people were actually going to see this! Very exciting!
EDIT: can you be a bit more specific about which part takes too long to load?
This Serverless Function has timed out.
Your connection is working correctly.
Vercel is working correctly.
504: GATEWAY_TIMEOUT Code: FUNCTION_INVOCATION_TIMEOUT ID: iad1::b6d7d-1751062063140-14fa915b731f
Thank you this is extremely helpful and now fixed
The website