At the start of November, OpenAI introduced Custom GPTs in OpenAI's words: "You can now create custom versions of ChatGPT that combine instructions, extra knowledge, and any combination of skills."
In the screenshot above, OpenAI states that a store is expected to be launched soon and that they will also be sharing revenue with Custom GPTs based on "how many people are using your GPT". People have been speculating on Twitter that a possible revenue sharing model could look something like this:
Resolution Criteria
Will resolve based on available evidence. Custom Bots by OpenAI won’t count. If the bot generates revenue through referral links or direct sales it won’t count. Only revenue paid out by OpenAI for the usage will be counted.
1,000
3.00@traders apparently openai is paying some creators but they are all under NDA and can’t disclose how much they are making :/
@Soli someone on Twitter, the "Source? I made it up" poster with the Dr Manhattan meme, said that GPT 4.5-t will release soon and 4-t will become free like 3.5-t, just so the GPT Store can be used by non chatgpt pro users, giving Custom GPT devs access to future customers.
Doesn't make complete sense, but there's an opportunity there for OpenAI to make money somehow, and devs with them, somehow...
@Soli I like this market, but how likely is it that we will have reliable information on how much revenue OpenAI pays out to particular Custom GPT creators? Do we have that for other app stores?
I think openai has an incentive to highlight how much revenue is being generated by the top GPTs to attract more developers/builders. I don’t think they will release exact figures but I feel we can expect a leaderboard with some popularity metric (downloads) and clear information on their revenue model which should not make it hard to estimate the revenue
Apple and Android don’t release this information but I have used various tools in the past that calculate an estimate. I think some of them just use credit card data or something like that.
@VAPOR i saw the same rumors and based on my experience working with gpt-3.5 i don’t think they are reliable enough to be used for the GPT store - i created the question below and hold the biggest yes position :)
@Soli this is complicated and I lack knowledge, but I saw that gpts are insecure and can be robbed of their ideas, files, prompts etc and replicated on a cheap third party service, recently saw this on Twitter
The bigger they get the more likely they are to leak users to clones?
@VAPOR You are right that it will be relatively easy to leak the system message + uploaded files. I think the top GPTs will be a bit more sophisticated than that and won’t be as trivial to copy. Also, I would trust the original creator of a GPT to know better than someone who clones it how it can be improved. Tops GPTs will probably be updated and improved constantly. OpenAI updates their system messages quite often.
@Soli it's not the prompt that the success of custom gpts will rely on, it's reliability and access to custom or proprietary data you can't get from anywhere else.
@Soli also if you think of the gpts as projects on GitHub without a license, then all these things hold again. A project doesn't get stars when everyone knows it's a clone. It's gotta be actively maintained and add some value
@firstuserhere Yes I agree with you 100%. This is what I was trying to say.
I do think that how GPTs access their data will affect how easy it is to “clone” them. If all data is just uploaded as files then it will be relatively easy to steal the files and clone the GPT. But even if this happens I would still use the original GPT because the creator knows best how to improve.
To counter this, custom gpts could retrieve data via actions/apis which will make it harder to steal. Constantly adding more files and improving the system message will also help.
@Soli and i agree with you on all those points too, but a few more months is not long enough time for me to be confident of people figuring out a monetization plan with something like this, implement it without losing customers, fending off competition while still generating $100k. I think it's more akin to rapid development of free games we saw earlier in 2010s, with efforts for monetization coming a bit later
Created this question: https://manifold.markets/Soli/will-any-custom-gpt-rank-in-the-top?r=U29saQ
There might be better ways to do it, will see what I can come up with