docs(benchmarks): parity-row note matches the committed snapshot

The committed data predates the per-linter parity split, so it
carries a single `mdsmith-parity` row — but the generated results
note described per-linter `mdsmith-<linter>-parity` rows that
aren't in the table, which read as a contradiction.

Make gen_fragments.py emit the note that matches the rows the
data actually holds: the single pre-split column now, and the
per-peer wording automatically once a refresh adds the four
columns. Align the "Like-for-like" parenthetical to the same
framing (single column reads closest to mado-parity; the next
refresh replaces it with one column per peer).

Regenerated; results table byte-identical; `mdsmith check .`
passes.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01PThu9DcNDDSAKaw1upanY4

Author: claude

docs/research/benchmarks/README.md

@@ -109,10 +109,10 @@ file: results.fragment.md
 docs/research/benchmarks/data/*.json — do not edit by hand. Re-run
 the harness (run.sh) and `mdsmith fix` to refresh. -->
 
-`mdsmith` is the default rule set. Each `mdsmith-<linter>-parity`
-row runs the rule set that peer enables by default, for a
-like-for-like comparison against that peer (the
-`bench-<linter>-parity.mdsmith.yml` profiles).
+`mdsmith` is the default rule set. The `mdsmith-parity`
+row is the single pre-split parity column; the per-linter
+`mdsmith-<linter>-parity` columns, one per peer, land when
+the benchmark is next re-run.
 
 **Repo corpus — 766 Markdown files** (median wall time, lower is
 better; `vs mado` is the ratio to mado's median):
@@ -185,9 +185,9 @@ The `mdsmith` → `mdsmith-parity` delta is the measured cost of
 the cross-file and generated-content layer, work users opt into.
 The residual gap to mado on long prose is engine headroom: the
 number to drive down, and the profiler loop below is how. (These
-figures predate the per-linter split; the `mdsmith-parity`
-column now selects `mado-parity`, whose 27-rule set is close to
-the rule class measured here.)
+figures predate the per-linter split: the single `mdsmith-parity`
+column shown here reads closest to the `mado-parity` 27-rule set,
+and the next refresh replaces it with one column per peer.)
 
 ### Why parity trails gomarklint
 

docs/research/benchmarks/gen_fragments.py

@@ -80,11 +80,24 @@ def main() -> None:
     repo_tbl, repo = rows(json_dir, "corpus_repo")
     neut_tbl, _ = rows(json_dir, "corpus_neutral")
 
-    note = ("`mdsmith` is the default rule set. Each "
-            "`mdsmith-<linter>-parity`\nrow runs the rule set that "
-            "peer enables by default, for a\nlike-for-like comparison "
-            "against that peer (the\n`bench-<linter>-parity.mdsmith.yml` "
-            "profiles).\n\n")
+    # The parity columns differ across the per-linter split: the pre-split
+    # snapshot carries a single `mdsmith-parity` row, while the current
+    # harness emits one `mdsmith-<peer>-parity` column per peer. Describe
+    # whichever the data actually holds so the note never contradicts the
+    # table below it.
+    per_linter_parity = [c for c in repo if c.startswith("mdsmith-")
+                         and c.endswith("-parity") and c != "mdsmith-parity"]
+    if per_linter_parity:
+        note = ("`mdsmith` is the default rule set. Each "
+                "`mdsmith-<linter>-parity`\nrow runs the rule set that "
+                "peer enables by default, for a\nlike-for-like comparison "
+                "against that peer (the\n`bench-<linter>-parity.mdsmith.yml` "
+                "profiles).\n\n")
+    else:
+        note = ("`mdsmith` is the default rule set. The `mdsmith-parity`\n"
+                "row is the single pre-split parity column; the per-linter\n"
+                "`mdsmith-<linter>-parity` columns, one per peer, land when\n"
+                "the benchmark is next re-run.\n\n")
     results = (GEN + "\n" + note +
                f"**Repo corpus — {repo_n} Markdown files** (median "
                "wall time, lower is\nbetter; `vs mado` is the ratio "

docs/research/benchmarks/results.fragment.md

@@ -2,10 +2,10 @@
 docs/research/benchmarks/data/*.json — do not edit by hand. Re-run
 the harness (run.sh) and `mdsmith fix` to refresh. -->
 
-`mdsmith` is the default rule set. Each `mdsmith-<linter>-parity`
-row runs the rule set that peer enables by default, for a
-like-for-like comparison against that peer (the
-`bench-<linter>-parity.mdsmith.yml` profiles).
+`mdsmith` is the default rule set. The `mdsmith-parity`
+row is the single pre-split parity column; the per-linter
+`mdsmith-<linter>-parity` columns, one per peer, land when
+the benchmark is next re-run.
 
 **Repo corpus — 766 Markdown files** (median wall time, lower is
 better; `vs mado` is the ratio to mado's median):