Recorded by commits 9279596 through f14e1d2
Experiment Note
An independent split-form Cahn-Hilliard suite was added. The phase field was prioritized, Pareto guards were added, and acceptance was then relaxed safely. The presence of smoke, residual, reproducibility, isolation, u-first-policy, and Pareto-policy tests shows that this was treated as a serious controller stress test rather than a configuration rename.
This is the last date represented in the current Git history. Later work exists as uncommitted paper, report, result, and submission artifacts.
Technical Essay
Cahn-Hilliard is a difficult high-order phase-field problem. A split form introduces a chemical-potential field so fourth-order dynamics can be represented through coupled lower-order residuals. That creates a new multi-objective issue: improving the auxiliary chemical potential is not necessarily more important than reconstructing the primary phase field. The controller was therefore made "u first," with support-only or capped influence from the auxiliary field.
The Pareto acceptance threshold initially appears to have been too strict for the noisy coupled problem. The final commit relaxed it safely rather than removing the guard. This is a recurring design pattern: preserve the invariant, adjust the tolerance based on noise and scale, and retain explicit tests.
Reading Note
Phase-field literature emphasizes sharp diffuse interfaces, stiffness, mass behavior, and the importance of resolving the primary order parameter. The lesson for a general controller was that variable awareness must also encode semantic priority: not every output field is equally important merely because it appears in the network.
Progress Update
By the end of tracked code development, the repository contained steady and transient flows, manufactured scalar PDEs, and a coupled phase-field system. The scientific focus then shifted from adding mechanisms to consolidating evidence and writing the paper.