#104 - When AI Coding Gets 10x More Expensive

But today is a little bit more sad talk, Karina, right?

Yeah, I think, you know, we have 1st of June.

So by 1st of June, we have co-pilot pricing changes going into place.

And what I see in some organizations or in quite a few, there is quite a bit of panic around the cost of it and how much it costs versus knowing how the people use it.

So one thing, you know, they look into, okay, how much we spend and how much we can decrease spending per person by putting hard limits on it.

But what comes with it as a user, as a developer, very often you have limited, you get less for more so you have to pay more money for what you are getting less that's one thing and being able to still work effectively with it if you don't have you know max plan as you have with cloud or something

like this what does this mean right it's changing the how much money how many tokens you can use per dollar like in co-pilot which means that if you're spending a lot of tokens the price will be a lot higher is that

Yeah, we pretty much moved from premium request billing to more token billings.

So depending on how many words, how many tokens you use, then you are priced by this.

Andre?

Yeah, so we had a whole episode on another podcast that I'm doing now, like a joint CKI in DevOps.

like where we define the big squeeze, and that's pretty much everyone doing that.

I mean, like Coursor was the first, the bravest.

They did it last year, where they went from the request, actual plain API tokens.

It's still possible to use Coursor today, but the composer model is free.

Basically, you pay, but then you can use it unlimited.

If you use any other model, then you pay IPI rates, which is still a decent deal.

With Cloud... have a subscription with a limit so you cannot go full throttle forever uh you have like five hour windows and then you have a weekly window it was very short now they did the deal with space x and that's space x and now it's like a little bit more usable i mean like it's get it's now harder to hit the limits assuming you're using opus as you as you did it right now like if you're using like haiku or sonnet for some reason you probably you will have a hard time hitting the daily meet but like if you're running with all posts it's actually not that hard to hit it so I kind of the copilot is actually one of those who are coming late to this but I do expect that this trend to continue that we will be moving more and more towards the towards the API rate model.

I mean, the open AI is the only ones, because they feel they're kind of lagging behind a little bit, so they're trying to attract people with a lot of subsidies, but basically it's all subsidies.

We are talking about subsidies.

Most of the subsidies are going to be drying up.

If you're using Kira from AWS, for instance, they also give you credits.

that lasts for a month, and then every month it's renewed for like 20, 40, or whatever box is the highest tier.

So those are subsidies.

They eventually will disappear.