Charge on the input tokens, then you will naturally optimise for fewer output tokens.

Theoretically.

In reality, one sessions output tokens become the next sessions input tokens (at least if you continue the topic) so, its not as aligned as all that.

But the parent is right, when incentives are not aligned, friction will happen. Its inevitable.

▲

demibabs 4 hours ago | parent | next [-]

“Claude, spend the next 10 hours trying to solve the Reimann Hypothesis”.

I agree that incentives are misaligned but there’s several competing model providers. If one gets funny with their costs people will jump ship, especially if the gap between the top 2 labs and everyone else keeps shrinking.

▲

dijit 3 hours ago | parent [-]

“I can take current sources and tell you how solved this is, but I am not willing to work to a timeframe or to solve things that aren’t yet solved by mathematicians or science”

These safeguards already exist when they get a whiff that you might be using Claude to fix security issues. Doesn’t seem farfetched given the incentives I outlined that they would apply to this kind of abuse.

How loose those controls are becomes a market force.

▲

dullcrisp 3 hours ago | parent [-]

Perfect, a coding agent that refuses to do things that haven’t already been done before.

	▲	dijit 3 hours ago \| parent [-]
		Hahahaha, I think you’ve misunderstood what LLMs are.

▲

cthaeh 4 hours ago | parent | prev [-]

[dead]