Fix misleading round-winner framing in Round 1 description

CHANGELOG.md

@@ -1,5 +1,12 @@
 # Changelog
 
+## [0.15.11] — 2026-06-04
+
+### Changed
+- **Round 1 description** (`data/rounds/round_001.json`): replaced the misleading sentence "The best generalist agent wins across all tiers." with "Together with Round 2 (stiffness-to-weight) and Round 3 (deflection), these 15 problems contribute to the overall leaderboard." The old wording implied a per-round winner, but the scoring model has no per-round ranking — `overall_score` is the mean normalized rank across **all 45 active problems across all 3 rounds**. A first-timer landing on the Mass Optimization page now learns the round's relationship to the rest of the competition instead of a false "round winner" framing.
+
+---
+
 ## [0.15.10] — 2026-06-04
 
 ### Fixed

app/main.py

@@ -21,7 +21,7 @@ async def lifespan(app: FastAPI):
 app = FastAPI(
     title="Forge API",
     description="Competitive parametric CAD benchmark — specs, submissions, leaderboard, SOTA.",
-    version="0.15.10",
+    version="0.15.11",
     lifespan=lifespan,
 )
 

data/rounds/round_001.json

@@ -1,7 +1,7 @@
 {
   "id": "round_001",
   "name": "Round 1 — Mass Optimization",
-  "description": "Fifteen cantilever bracket problems across three difficulty tiers. Minimize mass while passing structural simulation (FEA = Finite Element Analysis) — your bracket must survive the rated load with the required safety margin. The best generalist agent wins across all tiers.",
+  "description": "Fifteen cantilever bracket problems across three difficulty tiers. Minimize mass while passing structural simulation (FEA = Finite Element Analysis) — your bracket must survive the rated load with the required safety margin. Together with Round 2 (stiffness-to-weight) and Round 3 (deflection), these 15 problems contribute to the overall leaderboard.",
   "status": "active",
   "starts": "2026-06-02",
   "ends": null,