# HOW MANY README LINES CARRY A NUMBER THAT NOTHING ACCOUNTS FOR.
#
# Generated by `python ci.py claims-baseline`. Enforced by
# `test_the_unaccounted_claim_count_is_exactly_the_committed_baseline`, which is in the DEFAULT
# gate -- unlike the report, because this number is green the day it arrives and the report is not.
#
# THE RULE DIGEST IS THE POINT OF THIS FILE. A ratchet whose denominator moves silently is a lie
# with a number on it: a sibling project's first baseline fell 696 -> 335 and only about half was
# work, the rest was the literal-detection rule tightening. So the count is stored WITH a digest of
# the rule that produced it, and the test refuses to compare across a change rather than reporting
# a fall that nobody earned. Move the rule and the test tells you to re-measure; it does not quietly
# accept the new number.
#
# TWO DIGESTS, BECAUSE ONE COULD NOT ATTRIBUTE. `detection` covers what counts as a claim --
# `NUMBER`, the `_KINDS` name/pattern pairs, `_SUPERSEDED_DECLARES` -- and a change to it moves the
# count with NO WORK DONE, so the gate refuses to compare. `bindings` covers `_BOUND` and `ci.py`'s
# `_SCALE_FIGURES`, and a change there lowers the count because somebody bound a claim, which IS
# the work. One digest fired "not comparable" on both, so honest work tripped the alarm meant for a
# fall nobody earned -- and an alarm that fires on good news is one people learn to silence.
# Neither covers the kinds' REASONS: documentation for a reader, changing no verdict.
#
# AN EQUALITY, NOT A CEILING. `<=` would let the file rot upward while real work went unrecorded,
# and a ratchet that never tightens is not a ratchet. Going DOWN is as red as going up, and the
# failure message says which -- down means regenerate this file, up means a claim arrived that
# nothing accounts for.
detection 56cf5a2ef4679db6
bindings 1d6bb490ea3f7932
population 407
unaccounted 55
historical 68
