Causalor Labs
Field Notes
Structural reads on live events. Not how likely an outcome is, but whether it is still reachable, how fast the option set is closing, and what survives adversarial pressure.
How to Stop an Agent Taking an Action It Cannot Take Back
An agent wiped a production database in nine seconds. The post-mortem blamed missing approval gates. Here is what an actual runtime control looks like, why a second model reviewing the first is not one, and ten lines you can run right now.
Read note →Great American Artificial Intelligence Act Narrows the Release Path
The Great American Artificial Intelligence Act is more than another AI regulation headline. Its real test is whether federal AI law preserves a reachable path for frontier AI governance, competitive entry, AI compliance, and deployment.
Read note →The Lane Is Not Broken. It Is Closing.
Every shipping desk on the Street has written off the Asia-to-North-Europe lane for the quarter. That is a probability statement about a feasibility question, and the two have different answers.
Read note →Optionality Dies Before the Price Does
In every collapse worth studying, the set of survivable outcomes closes long before the price admits it. A structural reading of why the option set, not the tape, is the instrument to watch.
Read note →The Shape of What Is Still Possible
Almost every risk system ever built answers one question: how likely is this outcome? That number is a single shadow of a much larger object. These notes are written from inside that object, which has a size, a shape, and a death.
Read note →