AI truth and verification
Testing sycophancy, confident errors, softened facts and the difference between an answer that sounds right and one that can be checked.
Field notes
What I test, not what I promise.
This is the work in plain language. No service menu and no claim that every experiment becomes a finished system.
The current testing ground
Four areas. One proof rule.
If an idea cannot survive a real check, it does not earn a success story.
Operator systems
Building practical workflows for research, writing, records and follow-through, with visible proof paths and clear human gates.
Practical AI for regular people
Finding out whether a tool works for someone who speaks naturally, learns by doing and does not want to become a software engineer.
Writing and publication
Using AI as part of a serious writing process without letting fluent language hide weak thinking, stale facts or unfinished work.
Public proof paths
Follow the work at the source.
These are current public profiles. Their posts and publication records are stronger evidence than promotional copy written for this page.
One public proof object
The field work became a book.
The Invisible Code examines what happens when AI rewards confidence and agreement instead of truth. The paperback is live.