So I’m having a hard time with some people thinking this will be like when the app store or play store was released.
My main concern is that it will be similar to the plugins experience, where almost no one made money there. I think the hardest part is I don’t see a paid feature or a simplification of the paid process. I get it, it’s good to leave most of the work for the developer to do what they wish for.
I think it would be cool to have a free version and a paid version, with all the payment system already in place. Would be nice to have hosting options + domain + payment system offered for the builders. Otherwise, I suspect most gpt will be really low quality, builders will lose traction since there is no money to be made and no one will spend money since no assistant is worth more then 0.99 cents
But, I could be wrong. I do wonder, what are your guys expectations? Why do you think GPTs will be different then the Plugins? I think there should also be a gpt review of some sorts as publishing an app to any mobile store takes a few days to go public!
Also, there is nothing in place to see if its a duplicate gpt! I guess my concern is that 99% of the gpts will be simple gpts, 1% will be useful but no one will spend time to test it as 99.9% of gpt store visitors are gpt builders themselves looking to make money.
There is a specific issue about the topic of getting paid for LLM apps, especially if it’s a “small” consumer app. And they all point towards the creation of a store.
Many offerings currently use a “Bring your own Key” approach which is problematic because now we have end-users with developer like status and end-users don’t always know about the dangers of putting one’s key into a random input field on the web.
OpenAI is aware but also acknowledges that they have not, yet, offered a robust solution to this problem. The Plug-in store was a first attempt. The GPT store is now the second attempt, hopefully better than the first.
Additionally, both approaches are compatible, in the sense that previous developments can be re-deployed.
You can also read from the announcements that making a store work is something they really want. Probably because there is extra money to be made for them, too.
So, yes I am sure there will be a store. But will it work this time? Time will tell.
To be fair, I don’t think the plugins failed either. It was really cool to test out some of the plugins, but in terms of the plugin store, I feel like all of them became obsolete with the new models or were a really hard sell for a very niche market. They must have statistics on usage of plugins in comparison to the total user base and how much each plugin profited. By all means, look at the most popular ones:
In 1 year we went from oh we can’t handle pdfs to we have in real time conversations with a text model about a pdf I sent code interpreter. So where will the base models of chatgpt be in 1 year after its release?
So what I’m saying is, yes, the plugins were a really cool addition and let’s us run our own code in chatgpt, and gpts are cool because you can set configurations by text, but in terms of a store - I’m not sure they are there yet! I hope I am wrong about this as I’ve created my own gpts too!
Fair and I agree with everything you said. I think the biggest issue I have with it is the store part. I feel like that for a proper release of a store they need a couple more features for their GPT builders, like a unified payment system.
Any platform where you can make money off you normally have a system for those that meet the criteria to make money, instead of relying on them creating their own payment system. The more the builder can focus on building the better.
Also, chatgpt doesn’t really have a social media aspect to it, like a feed where you can have potential gpt shoppers strolling by… not sure if they even want that to be honest
another thing is the plugins&gpt builders are most likely stuck with not being able to manipulate components, so the window of creativity for a plugin is always limited by that. so you can’t release deliver a native vr experience inside gpt for example, you’d have to drag the user to your website by outputting a link to it. Same thing for having a simple game running on gpt. You’d have to download a gpt clone and mod it accordingly to be able to do that, but chatgpt is limited to that. You can’t even display a video using plugins. You can only link the user to it.
Also, I hope they give more space as the plugin store has very little space for each plugin. I hope they succeed, I know they care and I do care too, so that is why I’m making this post. I have a gut feeling that if they just release it as it was that it wont get the results they want, maybe that is why it was delayed the first time. Hopefully.
It addresses common feature request: “how can I save and select my favorite custom instructions?”
Why successful for OpenAI?
It puts sharing of customization prompt and continuing upon that behind a paywall after the disabling of “continue conversation” on shared links.
I suspect that’s going to be the 99%, as the development does not itself target exclusively those that can extend the agent capabilities to the fullest.
