This page holds longer-form writing where ideas have been worked through carefully.
These pieces reflect slower thinking, synthesis, and revision, aimed at clarity rather than immediacy.
Compared to notes, essays here are fewer, more deliberate, and updated only when the thinking has settled.
Current Essay
Stop Trusting AI Explanations
Coming soon
You get an explanation that seems solid enough to approve.
The review passes. The decision moves on.
Three weeks later, something quietly fails, and no one can trace why.
I kept seeing this pattern: confident outputs built on weak footing, passing review because the reasoning sounded coherent. Standard explainability did not help. In practice, it often made the problem worse.
This essay walks through what I built instead: a protocol that makes the shaping influences visible, so the ones that do not belong can be removed or constrained before the next decision. The point is not explanation. It is intervention.
I am not saying bias disappears. I am saying bias becomes selectable and reducible.
Not theory.
What I use when decisions have stakes.