Skip to content
Human Choice

Research + tools for human flourishing

Make room for a better human life.

We build and study AI systems that help people accomplish meaningful work, gain time, and retain control over their choices.

Our research asks how the benefits of automation reach the people whose lives it changes.

Explore the work Read our principles

What guides the work

Progress people can feel in their lives.

More time, stronger choices, and greater security count as progress—even when output stays the same.

Capability

More of what matters.

Useful tools should expand what people can accomplish, learn, and create.

Agency

A meaningful say.

People should be able to understand, direct, change, and refuse how automation acts for them.

Participation

A share in the possibilities.

People affected by automation should have a voice in its terms and how its benefits are shared.

A concrete place to start

Delegate the task.
Keep a say in the terms.

Wallet Kernel explores one part of human agency: letting an AI agent buy a resource within limits its operator understands and sets.

Wallet Kernel · offline prototype

Try the spending-control demo

The idea in four steps

  1. 01
    You set the terms.

    Choose permitted sellers, budgets, and approval rules.

  2. 02
    The agent requests a resource.

    It asks within the authority you have given it.

  3. 03
    The kernel checks the request.

    Allow, decline, or ask for an exact approval.

  4. 04
    The outcome stays inspectable.

    Keep authorization, payment, and useful results distinct.

Explore a browser simulation and inspect the separate engineering evidence.

The research agenda

Questions worth staying with.

Our first focus: people whose expertise and daily work are being reshaped by AI.

01

When AI saves time, who gets that time back?

Follow the distance between a faster task and a better day.

Proposed
02

Can people delegate without giving up control?

Study useful autonomy, understandable limits, and the burden of supervision.

Proposed
03

Does AI assistance help us become more capable?

Look beyond the immediate answer to learning, judgment, and lasting skill.

Proposed
Explore all research questions

A purpose we can hold ourselves to

AI should expand the lives people can choose.

We will judge our work by what changes for people, publish what we learn, and change direction when the evidence calls for it.

Meet the project