Vierla
Navigation

Our approach

How we study structure.

Learning creates patterns. Design shapes the conditions around it. Our research asks how each can inform the other.

  1. 01

    Start with the question

    An architecture is a hypothesis about how a system should learn and behave. We start with the problem, state the assumptions and ask what evidence would challenge them.

  2. 02

    Study the relationships

    A capable part can behave differently when connected to another. We examine what a connection enables, what it restricts and what happens when we change it.

  3. 03

    Keep alternatives testable

    A useful comparison isolates a choice. Change a representation, a memory or an interface, then measure what it contributes and where it fails.

  4. 04

    Design with a reason

    An objective or a constraint can make some patterns easier to learn and others harder to express. We want to understand those effects, including the possibilities a design excludes.

  5. 05

    Make the evidence legible

    An illustration explains an idea; an experiment tests it. A result needs its conditions, comparisons and limits alongside it. The diagrams here introduce research questions rather than report findings.

  6. 06

    Let learning inform design

    Unexpected learned behaviour can expose an assumption in the design. A change in design can make different behaviour possible. We study that exchange across representation, memory, prediction and coordination.

    Our methods and architectural choices remain open to revision.

An open question

What would you like AI to solve?

Tell us what you are trying to do, what gets in the way and what a useful outcome would look like.