The main issue is there isn't enough time to implement such a solution. You have to develop this new system of handling API keys, communicate this change to customers, develop a process to migrate users, and figure out what to do with people who paid for an entire year of access 3 months ago. All of that is a herculean task for a small dev team to accomplish in 30 days.

That's not even dealing with the fact that this process would be difficult for users to actually use, and may run afoul of Apple's app store rules.

While that solution may be appealing to tech-savy end users, it's completely untenable for a popular app, especially given the tight time window required.

> All of that is a herculean task for a small dev team to accomplish in 30 days.

Why 30 days? I mean, I know that Reddit announced that that's when they'll kill the API... but that's a completely arbitrary deadline.

Here’s an anecdote. I was subscribed to ChatGPT Plus for a while, to get access to GPT4.

I stopped subscribing after after I got GPT4 API access because I developed a little personal app which used the OpenAI API to just read and write directly from plain text files and that suits my workflow better than the ChatGPT website.

But it sucks because I’m constantly thinking about how much I’m using, and how many tokens I’m putting into my query, because each API call costs me money. It was way nicer just paying a flat fee and using it “as much as I want”, even though this actually costs me way less because I use don’t use $20USD worth of API calls in a month, even with GPT4.

It would be a nightmare to use Reddit if it cost money to scroll down or post a comment. On the other hand, that might actually be a good disincentive to help me spend less time on it.

I completely take your point, and would _love_ some kind of negative externality to keep from scrolling to much. I'd see that as a positive.

If I was designing such an app I would prominently display your ongoing monthly costs and add a budget/limit feature.

My problem isn’t that I’m worried I’m spending too much, I know for a fact that I’m spending less than I did on ChatGPT Plus, I can check my usage regularly, and I do.

My problem is that before I send each query, I think to myself “is this worth a fraction of a penny?” and, when I have a bunch of follow up queries, I think “the previous query/response history is getting pretty big, should I trim some of it before I send my next query?” or “Should I skimp on this one and just use GPT3.5?”

This is the same reason I can't ever see microtransactions for news sites gaining traction. While I can't afford to subscribe to every news site I would like to, I also don't want to have to decide whether every news article I might read is going to be worth my penny before I read it. So I stick with the status quo of subscribing to a couple of key sites and skipping out when I hit a paywall. It's suboptimal for both sides but I can't see pay per article working any better realistically.

I had the same reaction when Kagi search changed their pricing to a fixed number 1,000 searches per month for $10. I couldn’t imagine trying to use search thinking “is this one worth searching for?”

… I now pay $25 for their unlimited option even though I probably use less than 1k searches per month anyways

Good news on that- they’re experimenting with lower cost unlimited right now; people who had an annual membership before the switch are back to unlimited searches for now.

Which is why when designing pricing "per use" pricing models, prefer something like $0.0001 per search, or $15/month, whichever is less in a month

This is an interesting option. I'd assume the added friction for an average user will drop the Apollo user base drastically, but it would still allow it to stay alive.

Don't forget that they're getting their tokens from an OAuth endpoint. In otherwords it's already tied to a specific user. Reddit could have simply said "third party api support is only for reddit gold users".

> I have no idea if they prorate charges if you use less than 1000 calls so lets assume they don't, so the minimum daily cost for a user is $0.24.

The way these things work nearly everywhere is that $0.24 for 1000 API calls means your cost in a given billing period for N API calls during a billing period is 0.24 ⌈N/1000⌉ or 0.24 N/1000. The first is if they do not prorate, the second is if they do.

If it takes on average 345 requests per day per user, that would be 10 500 per month per user, which would be $2.64 per month per user if they do not prorate and $2.52 per month per user if they do prorate.

A Twitter client I used (Spring) had that, and it worked for months after Twitter killed off Tweetbot and all other clients.

However, it naturally stopped working when Twitter basically killed the free tier of their API.

It would likely still work on the "Basic" API tier, but I'm not paying $100 a month to use a Twitter app.

I remember Falcon Pro doing this 10 years ago when it hit Twitter's 100,000 user limit on its token. You could press 4 secret buttons in the corners of the login screen and it would take you to a secret page where you could put your own tokens.