[drunken_thor]

AI services are only minorly incentivized to reduce token usage. They want high token usage, it makes you pay more. They are going to continually test where the limit is, what is the max token usage before you get angry. All AI companies will continue to trade places for token use and cost as cost increases. We are in tepid water pretending it is a bath pretending we aren’t about to be boiled frogs.

[jedberg]

People said this about AWS too. "Why would they save you money??". It turns out that every time they reduce prices, they make more money, because more people use their services.

AI companies have the same incentive. Make it cheaper and people will use it more, making you more money (assuming your price is still above cost). And of course they have every reason to reduce their on costs.

[esperent]

AWS is notorious for being extremely expensive, so it's not like they became cheap. They just reduce prices from extremely expensive to slightly less extremely expensive, and that makes more people decide to start using it.

Since the price they are charging is still way, way above their operating costs there's no surprise really that they end up making more from small price reductions.

If competition drove them to reduce costs to the point where their operating costs started to be a large factor, the paradox would disappear.

[zormino]

jevons paradox

[minimaxir]

To an extent. That economic incentive stops making sense when a) capacity is an actual constraint and b) Anthropic is not a monopoly and is subject to pressure from competitors who are more user-friendly.

[estimator7292]

I severely doubt it. Token spend translates to real cost for the provider. Each token involves real and expensive compute. They aren't free monopoly money you get billed arbitrarily for. You're paying for electricity and infrastructure involved in generating each token.

Less spend means less real cost to the provider while your flat monthly subscription stays the same price. As well, reducing token use per customer means you can over-subscribe even harder, allowing for more flat monthly subscriptions.

Less tokens = more free capacity = more subscription income.

[GodelNumbering]

I am betting on the fact that people will get increasingly frustrated at closed agent lock-ins. I built (cline fork) and open-sourced https://github.com/dirac-run/dirac with the sole focus on token efficiency expecting that the closed-lock-in vendors will do enough to frustrate their users over time. Looking for contributors

[y42]

That's what I am thinking, too. It sound's like a conspiracy theory, but at the end Anthropic et al benefits from models that don't finish their jobs. I recently read about this "over editing phenomenon". The machine is never done. It doesn't want to.

It's like dating apps. They don't want you to find a good match, because then you cancel the subscription.

[biglyburrito]

Which works fine, right up until China releases a new DeepSeek model that's 85% as capable as an Anthropic or OpenAI premium model but costs a fraction of what either of those US companies are charging.

Speaking of which:

https://www.cnbc.com/2026/04/24/deepseek-v4-llm-preview-open...

[nananana9]

Up to a point. There is incentive when they get to the point where they literally can't serve their userbase and customers start leaving.

[zzzeek]

Well that's why threads like this are important to upvote. On hacker news , they're angry !