Pareto efficiency

This is what we're trying to achieve here.

A situation is Pareto efficient when all possible Pareto improvements have already been made.

That means there are no changes you can possibly make that would make one person better off without hurting someone else.

It means you've hit the frontier, the maximum.

Everyone can perform, everyone can belong, everyone can reach their potential.

What we're really saying is we are pursuing Pareto efficiency.

We are looking for Pareto improvements available to us right now.

But we're not playing the zero sum game.

From the research world.

A paper out of the University of Washington and Hong Kong University of Science and Technology introduced APO, Approximate Pareto Optimality.

a framework for personalizing large language models using just 20 samples per user.

The approach clusters users by shared preference patterns, meta-trains cluster-specific initializations, and then adapts for new users from a tiny sample.

Capabilities Research Expands the Safety Usefulness Pareto Frontier 2 by Alex Malen Published on October 2, 2026 Subtitle A model of when and why research improves or hurts safety It's tempting to define safety research as research that enables developers to deploy an AI system more safely without making the deployment much more expensive or much less useful.

You can visualize this definition of safety research as pushing out the safety usefulness Pareto frontier.

At any given level of usefulness, there's greater safety available.

There's an image here.

Awkwardly, this definition counts basically all capabilities research as safety research.

For example, consider performance optimization for inference.

By making inference more efficient you can use weaker, safer models more extensively than you would otherwise be able to, pushing out the Pareto frontier.

Likewise, any successful research whatsoever pushes out this Pareto frontier because research can only ever create more options.