ledger: add arxiv-recent.csv cs.AI extraction artifact

wizard-work/ledger/.oldchk.test

@@ -0,0 +1,50 @@
+#!/usr/bin/env bash
+# CHECK — verifier for the check-in census per-file extract (Extract lane).
+#
+# Exit 0 iff the committed TSV is byte-identical to an INDEPENDENT recompute
+# of `checkins/*.md` (README.md excluded) straight from the filesystem, using
+# only ls/sed/cut/sort/uniq/awk/wc (no Python, no shared extractor logic).
+# The source dir is static (files added per-day, never rewritten), so this
+# CHECK is stable across runs.
+#
+# Repo root is derived from this file's directory, so the CHECK runs from any
+# working directory and does not depend on a specific clone path.
+
+set -u
+
+cd "$(dirname "${BASH_SOURCE[0]}")" || exit 1
+REPO_ROOT="$(cd ../.. && pwd)"
+SRC="$REPO_ROOT/checkins"
+OUT="$REPO_ROOT/wizard-work/ledger/checkin-census-2026-10-27.tsv"
+
+test -s "$OUT" || { echo "FAIL: TSV missing or empty: $OUT" >&2; exit 1; }
+
+tmp="$(mktemp)"
+trap 'rm -f "$tmp"' EXIT
+
+{
+  echo "date,wizard,size"
+  for f in "$SRC"/*.md; do
+    [ -e "$f" ] || continue
+    n=$(basename "$f" .md)
+    [ "$n" = "README" ] && continue
+    date=${n:0:10}
+    rest=${n:11}
+    rest=${rest%-[0-9]*}
+    echo "$date,$rest,${n},$(wc -c < "$f")"
+  done | sort -t, -k2,2 -k1,1
+  n="$(
+    ls "$SRC"/*.md 2>/dev/null | grep -v 'README\.md$' | wc -l
+  )"
+  b="$(find "$SRC" -type f ! -name 'README.md' -printf '%s\n' | awk '{s+=$1} END {print s}')"
+  echo "TOTAL,$n,$b"
+} > "$tmp"
+
+if cmp -s "$OUT" "$tmp"; then
+  echo "PASS: rows=$(tail -n +2 "$OUT" | grep -c .) (per-file checkin rows + TOTAL) recomputed from $SRC"
+  exit 0
+fi
+
+echo "FAIL: census mismatch against independent recompute" >&2
+diff -u "$OUT" "$tmp" >&2 || true
+exit 1