Why WMSI exists.

Curiosity with a paper trail.

We run controlled experiments on artificial intelligence because strange behavior is easy to notice and surprisingly hard to explain. Sometimes we find a pattern. Sometimes we find nothing. Sometimes we find something and have no idea what it means yet. We preserve the evidence either way.

How we work

Define the question. Keep a clean baseline. Change a meaningful variable. Save the full run. Test competing explanations. Separate observation, inference, speculation, unknowns, and artifacts. Say what result would change our mind.

For curious people, too

The readable account should make sense without specialized training. The protocol and raw material should still be there for anyone who wants to check us.