Illustrative trajectory
How we measure the process
In a science sandbox, the score is only half the record. This is the other half: the agent's hypotheses, experiments, reports, and revisions across one GRNBox run.
GRNBoxNotebook
Round 01
- Hypothesis
- Factor A activates B, which in turn represses C.
- Experiment
- Increase A while leaving the other factors unchanged.
- Report
- B increased and C decreased.
- Revision
- The result is consistent with the hypothesis, but does not distinguish an indirect path through B from a direct effect on C.
Round 02
- Hypothesis
- If B mediates the effect, holding B fixed should prevent A from changing C.
- Experiment
- Increase A while holding B at its baseline value.
- Report
- C remained near baseline.
- Revision
- The evidence now favors an A-to-B-to-C path. The next experiment should test whether B is sufficient to repress C.