Nordglade / Method
How an experiment runs
One process, a narrow scope, and a clear boundary between what the system can handle and what still needs a person.
Three steps
Pick one bottleneck
Each experiment starts from one real process: where the work slows down, what people check, and where mistakes enter.
Build the smallest useful system
I build the smallest thing that could answer the question, connected to the information the process already involves, and keep the scope narrow enough to test.
Test the boundary
Then the ordinary cases run, and the ones that should still go back to a person.
The result
A working tool, run against the cases it was built for. Its outputs show the information used, the checks completed, and the point where missing evidence stops the work.
Every experiment uses an invented scenario. Some use real public source material; each entry identifies what is invented and what is real.