Superadditive

EXPERIMENTS / AN OPEN LAB

Experiments in making knowledge functional.

There's good research on how organizations work, and most of it never reaches the people who could use it. Each experiment below takes something we know, from research and from our own client work, and turns it into a tool you can put to work today.

01

Start with who you are

A senior leader responsible for results

Someone leading change from the middle

A team that keeps hitting the same wall

02

Or browse all seven experiments

Each one works on its own
03

This week in The Feed

IN REVIEW · UPDATED SEP 27, 2026

Alongside the tools, we scan new research and practice on AI at work from a vetted list of sources and people. Here's what's made the cut.

What the latest reading suggests

  1. Two working papers found open-weight and post-trained LLMs discriminate by gender, race and age in hiring simulations, with exclusion rates rising from 5.6% to 17.3% after post-training.
  2. An end-to-end audit of a real public employment agency's AI hiring system found age, salary and gender disparities that aggregate parity metrics had masked.
  3. In a simulated multicenter trial, junior clinicians caught only 15.8% of AI hallucinations; one real neuroradiology department using structured change management raised active review of AI outputs from 14% to 61%.
Explore the full Feed