tidy landing layout and docs copy

Author: atharvas

src/components/pages/DocsPage.svelte

@@ -3,7 +3,7 @@
 		{
 			name: "fc-eval",
 			description:
-				"Run frontier LLM agents against the FormulaCode benchmark. Spins up reproducible Docker environments, replays the unit-test suite, and computes per-workload speedup, advantage, and stratified scores. Bring your own Terminus or OpenHands agent — `fceval run -d formulacode -a <your-agent>` is all it takes.",
+				"Run frontier LLM agents against the FormulaCode benchmark. Spins up reproducible Docker environments, verifies correctness against the unit-test suite, and computes per-workload speedup, advantage, and stratified scores.",
 			docsUrl: "/docs/eval/",
 			repoUrl: "https://github.com/formula-code/fc-eval",
 			image: "/assets/og/fc-eval.png",
@@ -12,7 +12,7 @@
 		{
 			name: "datasmith",
 			description:
-				"The four-stage pipeline that mines FormulaCode's tasks from real GitHub repositories: scraping high-quality performance PRs, attribute-filtering them with LLM judges, synthesising reproducible build environments, and running the statistical-significance tests that admit a candidate into the benchmark.",
+				"The pipeline for curating FormulaCode's tasks from real GitHub repositories. The code for scraping, filtering, building, and verifying high-quality performance PRs is maintained here.",
 			docsUrl: "/docs/data/",
 			repoUrl: "https://github.com/formula-code/datasmith",
 			image: "/assets/og/datasmith.png",
@@ -53,9 +53,9 @@
 			</div>
 			<h1 class="page-title">Build, evaluate, and explore FormulaCode</h1>
 			<p class="page-desc">
-				Three open-source repositories power the benchmark. Pick the docs you
-				need — run agents against the benchmark, see how tasks are mined, or
-				query the live data behind the leaderboard.
+				FormulaCode consists of two parts: a pipeline to construct performance
+				optimization tasks, and an execution harness that connects a language
+				model to our terminal sandbox.
 			</p>
 		</div>
 	</header>
@@ -99,17 +99,22 @@
 			<div class="endpoints-head">
 				<h2 class="section-title">Live data endpoints</h2>
 				<p class="section-subtitle">
-					Two subdomains expose the live task and run database. <strong>Uptime
-					is not guaranteed</strong> — these are research endpoints, sometimes
-					rebuilt mid-week. For reproducible evaluation, prefer the static
-					CSV that ships with this site.
+					Two subdomains expose the live task and run database. <strong
+						>Uptime is not guaranteed</strong
+					> — these are research endpoints, sometimes rebuilt mid-week. For reproducible
+					evaluation, prefer the static CSV that ships with this site.
 				</p>
 			</div>
 
 			<ul class="endpoints-list">
 				{#each liveEndpoints as e}
 					<li class="endpoint">
-						<a class="endpoint-host" href={e.href} target="_blank" rel="noopener noreferrer">
+						<a
+							class="endpoint-host"
+							href={e.href}
+							target="_blank"
+							rel="noopener noreferrer"
+						>
 							{e.host} ↗
 						</a>
 						<span class="endpoint-label">{e.label}</span>

src/components/pages/OverviewPage.svelte

@@ -13,14 +13,14 @@
 <div class="overview-page">
 	<OverviewHeader />
 
-	<DatasetPipelineViz />
 	<Abstract />
+	<DatasetPipelineViz />
 	<DatasetStatistics />
 	<KeyFindingsScroll />
 
 	<ResultsHighlights
 		title="Leaderboard at a glance"
-		subtitle="The six findings above roll up into the per-agent advantage scores below. Positive bars beat the human expert; negative bars trail."
+		subtitle="The findings above roll up into the per-agent advantage scores below. Positive bars beat the human expert; negative bars trail."
 		stratified={leaderboardData.stratified}
 	/>
 

src/components/sections/ResultsHighlights.svelte

@@ -16,7 +16,6 @@
 
 	const TABS = [
 		{ key: "advantage", label: "Overall", desc: "Overall" },
-		{ key: "level1", label: "L1 — Params", desc: "Parameter level" },
 		{ key: "level2", label: "L2 — Function", desc: "Function level" },
 		{ key: "level3", label: "L3 — Class", desc: "Class level" },
 		{ key: "level4", label: "L4 — Module", desc: "Module level" }
@@ -43,25 +42,6 @@
 		return { abs: Math.max(Math.abs(max), Math.abs(min), 0.02) };
 	})();
 
-	$: leader = sortedRows[0];
-	$: laggard = sortedRows[sortedRows.length - 1];
-
-	$: takeaway = (() => {
-		if (!leader || !laggard) return "";
-		const tab = TABS.find((t) => t.key === activeTab);
-		const ctx = tab?.desc ?? "Overall";
-		const ln = `${leader.agent} · ${leader.model}`;
-		const lln = `${laggard.agent} · ${laggard.model}`;
-		const fmt = (v) => `${v >= 0 ? "+" : ""}${v.toFixed(4)}`;
-		if (leader.value > 0 && laggard.value < 0) {
-			return `${ctx}: <strong>${ln}</strong> leads at <code>${fmt(leader.value)}</code>, slightly above the human expert. <strong>${lln}</strong> trails at <code>${fmt(laggard.value)}</code>.`;
-		}
-		if (leader.value > 0) {
-			return `${ctx}: <strong>${ln}</strong> edges out the expert at <code>${fmt(leader.value)}</code>; the rest sit at or below the human baseline.`;
-		}
-		return `${ctx}: every agent underperforms the expert. <strong>${ln}</strong> is closest at <code>${fmt(leader.value)}</code>; <strong>${lln}</strong> is furthest behind at <code>${fmt(laggard.value)}</code>.`;
-	})();
-
 	function widthFor(v) {
 		if (!Number.isFinite(v) || scale.abs === 0) return 0;
 		return (Math.abs(v) / scale.abs) * 50;
@@ -101,13 +81,6 @@
 		</div>
 
 		<div class="results-panel">
-			<div class="results-panel-title">
-				{TABS.find((t) => t.key === activeTab)?.desc} advantage
-			</div>
-			<div class="results-kicker">
-				Ranked by Σ(oracle speedup − agent speedup) / N. Bars symmetric around 0.
-			</div>
-
 			<div class="bar-chart" class:revealed>
 				{#each sortedRows as row, i (row.agent + row.model + activeTab)}
 					<div class="bar-row" style="--i: {i}">
@@ -139,8 +112,6 @@
 					</div>
 				{/each}
 			</div>
-
-			<p class="results-caption">{@html takeaway}</p>
 		</div>
 	</div>
 </section>
@@ -200,20 +171,6 @@
 		box-shadow: var(--shadow);
 	}
 
-	.results-panel-title {
-		font-family: var(--sans);
-		font-size: 0.875rem;
-		font-weight: 700;
-		color: var(--text-primary);
-		margin-bottom: var(--space-xs);
-	}
-
-	.results-kicker {
-		font-size: 0.75rem;
-		color: var(--text-muted);
-		margin-bottom: var(--space-md);
-	}
-
 	.bar-chart {
 		display: grid;
 		gap: 10px;
@@ -322,29 +279,6 @@
 		color: var(--score-bad);
 	}
 
-	.results-caption {
-		margin-top: var(--space-md);
-		padding-top: var(--space-sm);
-		border-top: 1px dashed var(--border-primary);
-		font-size: 0.8125rem;
-		line-height: 1.65;
-		color: var(--text-muted);
-	}
-
-	.results-caption :global(strong) {
-		color: var(--text-primary);
-		font-weight: 600;
-	}
-
-	.results-caption :global(code) {
-		font-family: var(--mono);
-		font-size: 0.85em;
-		padding: 1px 5px;
-		background: var(--bg-secondary);
-		border-radius: 3px;
-		color: var(--text-primary);
-	}
-
 	@media (max-width: 720px) {
 		.bar-row {
 			grid-template-columns: 1fr;