Begin with what is worth studying
Automated research often begins by generating candidates, but the value of the research question cannot be skipped. Before committing to experiments, we need to clarify the evidence behind the question, its relationship to prior work, and the observations that could change our judgment.
Carry failures into the next round
An exploration's value extends beyond the candidate that survives. Refuted branches, the conditions under which attempts failed, and comparisons under the same budget also determine whether the next round makes real progress. In Research Hub and my research discovery tools, I am working to preserve these records as research assets that future work can build on.
Keep the process understandable
A complete engineering loop does not, by itself, establish the validity of a scientific discovery. I want literature, search, refutation, and experiments with frozen protocols to form an evidence-based process. I also want people using the tools to understand why each experiment is being done. Connecting research with teaching remains a central direction as I develop this workbench.