Figure 15.5: The trace-to-dataset flywheel, the loop that turns firefighting into improvement
Figure 15.5 The trace-to-dataset flywheel, the loop that turns firefighting into
improvement. Production runs are recorded as traces; the failures in them are sorted and
become new eval cases (in accent, the handoff to the next chapter); those cases join the
regression suite; the suite makes every later change safer; and safer changes go back to
production to be traced again. Turn the loop once and a resolved failure becomes a test
that guards every future release.
