We build to augment, never to replace

This is the first thing to know about us, because it decides what we work on and what we turn down.

A system we build should leave the person using it more capable than it found them. Not faster at producing an output they no longer understand, and not dependent on something they cannot check.

That is a design requirement, not a preference. It determines what we count as a working system: if a person could not do the thing better after using it, we have not built what we set out to build, however good the output looks.

The industry's default question is what a system can do instead of a person. It is a narrower problem than it appears, and a less interesting one. The harder question, and the one our research is actually about, is what makes somebody more capable and whether that can be measured rather than asserted.

We think the second question is where the durable work is. It is also the only one an institution can adopt without quietly hollowing out the people it employs.

• Systems whose value is that a person is no longer needed.
• Products that produce a result the user cannot inspect, question, or reproduce.
• Automating a judgement while leaving the person nominally responsible for it.