From 337e0cedaa7f6b19187794a33f90865d158e076d Mon Sep 17 00:00:00 2001
From: rNLKJA <62606765+rNLKJA@users.noreply.github.com>
Date: Fri, 9 Oct 2026 20:11:14 +1030
Subject: [PATCH] fix(content): label the A* admissibility claim as corrected
 in 2026

The Part A report argued that both Manhattan and Euclidean heuristics were
admissible. The paired study shows neither is on the hex grid, and the site
said so without noting that the original report claimed the opposite. The
landing page, the A* Lab findings and the README now say what the report
claimed and that it was corrected in 2026.

Content audit: every other landing, methods, agent-card, decision-record and
tour figure was traced to the original code, notebooks, report or the
reference-study data and matched. The original benchmark (144/160) was re-run
with uv and reproduced byte for byte.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
---
 README.md                  | 3 ++-
 web/src/app/astar/page.tsx | 9 +++++----
 web/src/app/page.tsx       | 5 +++++
 3 files changed, 12 insertions(+), 5 deletions(-)

diff --git a/README.md b/README.md
index 375d060..255f91e 100644
--- a/README.md
+++ b/README.md
@@ -174,7 +174,8 @@ From the 2026 evaluation (details, seeds and downloadable data on the site):
   but returns a shortest path on 74.3% of solvable boards (71.4% to 77.0%) against 85.6% (83.2% to
   87.7%) for Euclidean. On the same 931 boards that is 11.3 points fewer (paired bootstrap 95% CI
   9.1 to 13.3); where only one heuristic was optimal it was Manhattan on 1 board and Euclidean on
-  106 (exact McNemar p < 0.001). Neither heuristic is admissible on this hex grid.
+  106 (exact McNemar p < 0.001). Neither heuristic is admissible on this hex grid, although
+  our 2022 report argued that both were.
 - **No detectable disagreement with the original Python** on the pairings feasible in Python
   (480 games; every Newcombe 95% interval for the six win-rate differences contains 0). With 80
   games per pairing this check can only detect differences larger than about ±15 points.
diff --git a/web/src/app/astar/page.tsx b/web/src/app/astar/page.tsx
index 36b5ac1..38510b9 100644
--- a/web/src/app/astar/page.tsx
+++ b/web/src/app/astar/page.tsx
@@ -68,10 +68,11 @@ export default function AstarPage() {
               {formatPStatement(o.mcnemar.pValue)}).
             </Finding>
             <Finding>
-              <strong>Why: neither heuristic is admissible here.</strong> On this hex grid a step
-              like (1, −1) costs 1, but Manhattan scores it 2 and Euclidean √2. Overestimating makes
-              the search greedier: fewer expansions, longer paths. The two sample inputs happen to
-              be solved optimally (8 and 13 cells, confirmed by BFS).
+              <strong>Why: neither heuristic is admissible here.</strong> Our 2022 report argued
+              that both were; that was wrong. On this hex grid a step like (1, −1) costs 1, but
+              Manhattan scores it 2 and Euclidean √2. Overestimating makes the search greedier:
+              fewer expansions, longer paths. The two sample inputs happen to be solved optimally (8
+              and 13 cells, confirmed by BFS).
             </Finding>
           </ul>
           <PairedStudySummaryView summary={summary} />
diff --git a/web/src/app/page.tsx b/web/src/app/page.tsx
index 3467f6b..8c71ed8 100644
--- a/web/src/app/page.tsx
+++ b/web/src/app/page.tsx
@@ -239,6 +239,11 @@ export default function HomePage() {
               colour that blocks the search. A notebook experiment compared the heuristics on
               hundreds of random boards.
             </p>
+            <p className="text-muted-foreground mt-3">
+              <strong className="text-foreground">Corrected in 2026:</strong> our report argued that
+              both heuristics were admissible. On this hex grid neither is, so A* can return a
+              longer path than the shortest one. The A* Lab measures how often.
+            </p>
             <Button asChild variant="outline" className="mt-5">
               <Link href="/astar">
                 Open the A* Lab <ArrowRight />