diff --git a/README.md b/README.md
index 48beb78..b2abae0 100644
--- a/README.md
+++ b/README.md
@@ -156,7 +156,7 @@ data/                           chats/ · graph/ (six ledgers + positions/source
 
 **The knowledge graph (two channels, one writer).** The write channel: the extraction pipeline — extractor (LLM: sees only the session, rendered as readable markdown by the shared `core/session-markdown.ts` engine — propositions as true statements, `follows`/`supports` edges with verbatim evidence; fully trusted, graph-blind) → auditor (LLM, an agentic loop that never sees the source: explores the existing graph through `find` / `list_sessions` / `list_propositions` tools, stamps each proposition with an identity verdict — link to an existing node (equivalence bias: paraphrases merge) or create — and may draw bridging edges between existing nodes and the new batch) → merger (`mergeFindings`, deterministic: applies the verdicts, link-appends phrasings/provenance, resolves and dedupes edges, checkpoints coverage) → layout settle. No quiz outcomes or understanding assessments live in the graph (2026-09-27 cleanup); practice (`/practice`) consumes the graph as feedback-only problem sets and writes nothing back. Triggered when a chat's **uncovered tokens** (message-view fingerprints, `coverage.jsonl`) cross `graph.extractionChunkTokens` (a settings knob, default 4096, floor 500 — read fresh per turn); failure checkpoints nothing, the next trigger re-runs. The same pipeline serves the **notes front door** (`POST /api/graph/import`: any `*.md`/`*.txt` becomes a source; re-imports are idempotent by path hash). The read channel: `GET /api/graph` derives the snapshot from the ledgers — a pure read; the home view is a hand-rolled Canvas 2D renderer (one rAF loop, dirty-flag idle, d3-zoom camera, zoom-gated labels, cluster hulls) over a d3-force layout, with server-persisted render settings and drag-committed positions. Clusters are **birth-session groupings**: a node joins the cluster of the session that birthed it (a chat, or an imported notes file as a file-kind session), derived at read — no embedding pass, no naming pass. The teacher reaches the graph on its own terms through the gated prober tools (`list_sessions` / `list_propositions`, the `agent.graphTools` setting) — there is no automatic context injection.
 
-**Practice (problem sets).** The graph's `follows`-edges become reasoning chains; the `/practice` creation dialog picks a birth-session cluster (the graph view's own groups) and freezes its propositions + chains into a self-contained `ProblemSet`. Three named-level axes steer generation (ladder rung 0–4, cue distance 0–2, Theory|Application — never a number to interpolate; defaults Direct exercise · Adjacent context · Theory). A partitioner agent covers every chain exactly once, one generator per partition writes problems against the owner's problem-engineering framework (commit-before-pose: statement + solution + rubric together), and grading is **strengthening, not assessment** — one attempt per problem with its own submit button, instant solution/rubric reveal, then a score 0–10 with brief reasoning landing live; nothing ever written back to the graph. Runs are registry-backed background jobs on the activity dock (no steer, never wakes the teacher); sets live in `data/practice`, crash-swept on boot.
+**Practice (problem sets).** The graph's `follows`-edges become reasoning chains; the `/practice` creation dialog picks a birth-session cluster (the graph view's own groups) and freezes its propositions + chains into a self-contained `ProblemSet`. Three named-level axes steer generation (ladder rung 0–4, cue distance 0–2, Theory|Application — never a number to interpolate; defaults Direct exercise · Adjacent context · Theory), plus an optional free-form note the agents weigh but need not follow. A partitioner agent covers every chain exactly once, one generator per partition writes problems against the owner's problem-engineering framework (commit-before-pose: statement + solution + rubric together), and grading is **strengthening, not assessment** — one attempt per problem with its own submit button, instant solution/rubric reveal, then a score 0–10 with brief reasoning landing live; nothing ever written back to the graph. Runs are registry-backed background jobs on the activity dock (no steer, never wakes the teacher); sets live in `data/practice`, crash-swept on boot.
 
 **Dictation.** Optional speech-to-text: the server mints short-lived Deepgram tokens; audio streams browser→Deepgram over WebSocket, never through the server.
 
diff --git a/app/api/practice/sets/route.ts b/app/api/practice/sets/route.ts
index 4ce615e..8985da4 100644
--- a/app/api/practice/sets/route.ts
+++ b/app/api/practice/sets/route.ts
@@ -1,8 +1,9 @@
 /**
  * POST /api/practice/sets — create a problem set (202 {id}; generation
  * runs in the background — the fire-and-forget pipeline). Body:
- *   {chatId, params}          — everything this session touched
- *   {nodeIds, params}         — a manual node selection
+ *   {nodeIds, params, title?, note?}  — a node selection, the cluster's
+ *                                     label, and the learner's optional
+ *                                     free-form note
  * Validation + typed-error mapping live in the service (server/practice).
  * GET — the set list (createdAt desc, the store's order).
  */
diff --git a/components/practice-dialog.tsx b/components/practice-dialog.tsx
index 1604c87..1133f6d 100644
--- a/components/practice-dialog.tsx
+++ b/components/practice-dialog.tsx
@@ -33,282 +33,364 @@ import type { GraphClusterGroup, PracticeParams } from "@/src/core";
 /** The creation defaults (owner, 2026-09-29): rung 2 Direct exercise,
  *  cue 1 Adjacent context, Theory. */
 export const DEFAULT_PRACTICE_PARAMS: PracticeParams = {
-	ladderRung: 2,
-	cueDistance: 1,
-	presentation: "theory",
+  ladderRung: 2,
+  cueDistance: 1,
+  presentation: "theory",
 };
 
 export interface PracticeFailure {
-	status: number;
-	error: string;
-	code?: string;
+  status: number;
+  error: string;
+  code?: string;
 }
 
 /** The dialog's pure body — exported for SSR pins (the delete-session /
  *  settings body pattern). The wrapper owns state + fetch + portal. */
 export function PracticeDialogBody({
-	params,
-	onParam,
-	clusters,
-	selectedKey,
-	onSelect,
-	busy,
-	failure,
-	onOpenSettings,
-	onClose,
-	onRun,
+  params,
+  onParam,
+  note,
+  onNote,
+  clusters,
+  selectedKey,
+  onSelect,
+  busy,
+  failure,
+  onOpenSettings,
+  onClose,
+  onRun,
 }: {
-	params: PracticeParams;
-	onParam: (patch: Partial<PracticeParams>) => void;
-	clusters: GraphClusterGroup[];
-	selectedKey: string | null;
-	onSelect: (key: string) => void;
-	busy: boolean;
-	failure: PracticeFailure | null;
-	onOpenSettings: () => void;
-	onClose: () => void;
-	onRun: () => void;
+  params: PracticeParams;
+  onParam: (patch: Partial<PracticeParams>) => void;
+  /** The learner's free-form note (owner, 2026-09-29) — optional,
+   *  typing-only; carried to the partitioner + generators as additional
+   *  input they weigh but need not follow. */
+  note: string;
+  onNote: (note: string) => void;
+  clusters: GraphClusterGroup[];
+  selectedKey: string | null;
+  onSelect: (key: string) => void;
+  busy: boolean;
+  failure: PracticeFailure | null;
+  onOpenSettings: () => void;
+  onClose: () => void;
+  onRun: () => void;
 }) {
-	const canRun = !busy && selectedKey !== null;
-	return (
-		<div
-			role="dialog"
-			aria-modal="true"
-			aria-label="New problem set"
-			data-testid="practice-dialog"
-			onClick={(e) => e.stopPropagation()}
-			className="flex max-h-[85vh] w-[32rem] max-w-[calc(100vw-2rem)] flex-col rounded-xl border bg-card p-5 shadow-xl"
-		>
-			<h2 className="font-title text-2xl tracking-tight">New problem set</h2>
+  const canRun = !busy && selectedKey !== null;
+  return (
+    <div
+      role="dialog"
+      aria-modal="true"
+      aria-label="New problem set"
+      data-testid="practice-dialog"
+      onClick={(e) => e.stopPropagation()}
+      className="flex max-h-[85vh] w-[32rem] max-w-[calc(100vw-2rem)] flex-col rounded-xl border bg-card p-5 shadow-xl"
+    >
+      <h2 className="font-title text-2xl tracking-tight">New problem set</h2>
 
-			{/* The cluster: a birth-session group of the graph, resolved to
+      {/* The cluster: a birth-session group of the graph, resolved to
 			    node ids on create (graph-domain, owner 2026-09-29). */}
-			<div className="mt-4 flex flex-col gap-1">
-				<span className="text-xs uppercase tracking-widest text-muted-foreground/70">Cluster</span>
-				<div className="max-h-40 overflow-y-auto rounded-lg border" data-testid="practice-dialog-clusters">
-					{clusters.map((cluster) => (
-						<button
-							type="button"
-							key={cluster.key}
-							data-testid={`practice-cluster-${cluster.key}`}
-							onClick={() => onSelect(cluster.key)}
-							className={`flex w-full items-center justify-between gap-3 px-3 py-2 text-left text-sm transition-colors hover:bg-accent ${
-								selectedKey === cluster.key ? "bg-accent" : ""
-							}`}
-						>
-							<span className="min-w-0 truncate">
-								{cluster.label ?? <span className="text-muted-foreground/60">unlabeled cluster</span>}
-							</span>
-							<span className="shrink-0 text-xs tabular-nums text-muted-foreground">
-								{cluster.kind === "notes" ? "notes · " : ""}
-								{cluster.nodeCount} propositions
-							</span>
-						</button>
-					))}
-					{clusters.length === 0 && (
-						<div className="px-3 py-2 text-sm text-muted-foreground">
-							No clusters yet — a cluster appears once a session births its first proposition.
-						</div>
-					)}
-				</div>
-			</div>
+      <div className="mt-4 flex flex-col gap-1">
+        <span className="text-xs uppercase tracking-widest text-muted-foreground/70">
+          Cluster
+        </span>
+        <div
+          className="max-h-40 overflow-y-auto rounded-lg border"
+          data-testid="practice-dialog-clusters"
+        >
+          {clusters.map((cluster) => (
+            <button
+              type="button"
+              key={cluster.key}
+              data-testid={`practice-cluster-${cluster.key}`}
+              onClick={() => onSelect(cluster.key)}
+              className={`flex w-full items-center justify-between gap-3 px-3 py-2 text-left text-sm transition-colors hover:bg-accent ${
+                selectedKey === cluster.key ? "bg-accent" : ""
+              }`}
+            >
+              <span className="min-w-0 truncate">
+                {cluster.label ?? (
+                  <span className="text-muted-foreground/60">
+                    unlabeled cluster
+                  </span>
+                )}
+              </span>
+              <span className="shrink-0 text-xs tabular-nums text-muted-foreground">
+                {cluster.kind === "notes" ? "notes · " : ""}
+                {cluster.nodeCount} propositions
+              </span>
+            </button>
+          ))}
+          {clusters.length === 0 && (
+            <div className="px-3 py-2 text-sm text-muted-foreground">
+              No clusters yet — a cluster appears once a session births its
+              first proposition.
+            </div>
+          )}
+        </div>
+      </div>
 
-			{/* The three parameters — named levels, never a number to
+      {/* The three parameters — named levels, never a number to
 			    interpolate (owner, 2026-09-29). Stepped sliders snap to the
 			    detents; each shows the level's NAME + its one-line
 			    definition; presentation is a two-segment toggle. */}
-			<div className="mt-4 flex flex-col gap-3">
-				<div className="flex flex-col gap-1">
-					<span className="flex items-center justify-between text-[11px] text-muted-foreground">
-						Ladder
-						<span className="text-muted-foreground/80">0 · 1 · 2 · 3 · 4</span>
-					</span>
-					<input
-						type="range"
-						min={0}
-						max={4}
-						step={1}
-						value={params.ladderRung}
-						data-testid="param-ladder"
-						onChange={(e) => onParam({ ladderRung: Number(e.target.value) as PracticeParams["ladderRung"] })}
-						className="h-1.5 w-full accent-primary"
-					/>
-					<span data-testid="param-ladder-level" className="text-xs">
-						<span className="font-medium">{LADDER_RUNGS[params.ladderRung].name}</span>
-						<span className="text-muted-foreground"> — {LADDER_RUNGS[params.ladderRung].definition}</span>
-					</span>
-				</div>
-				<div className="flex flex-col gap-1">
-					<span className="flex items-center justify-between text-[11px] text-muted-foreground">
-						Cue distance
-						<span className="text-muted-foreground/80">0 · 1 · 2</span>
-					</span>
-					<input
-						type="range"
-						min={0}
-						max={2}
-						step={1}
-						value={params.cueDistance}
-						data-testid="param-cue"
-						onChange={(e) => onParam({ cueDistance: Number(e.target.value) as PracticeParams["cueDistance"] })}
-						className="h-1.5 w-full accent-primary"
-					/>
-					<span data-testid="param-cue-level" className="text-xs">
-						<span className="font-medium">{CUE_LEVELS[params.cueDistance].name}</span>
-						<span className="text-muted-foreground"> — {CUE_LEVELS[params.cueDistance].definition}</span>
-					</span>
-				</div>
-				<div className="flex flex-col gap-1">
-					<span className="text-[11px] text-muted-foreground">Presentation</span>
-					<div className="flex gap-1" data-testid="param-presentation">
-						{(["theory", "application"] as const).map((pres) => (
-							<button
-								type="button"
-								key={pres}
-								data-testid={`param-presentation-${pres}`}
-								onClick={() => onParam({ presentation: pres })}
-								className={`flex-1 rounded-md border px-3 py-1.5 text-xs transition-colors ${
-									params.presentation === pres
-										? "border-primary bg-primary/10 font-medium text-foreground"
-										: "text-muted-foreground hover:bg-accent"
-								}`}
-							>
-								{PRESENTATIONS[pres].name}
-							</button>
-						))}
-					</div>
-					<span data-testid="param-presentation-level" className="text-xs text-muted-foreground">
-						{PRESENTATIONS[params.presentation].definition}
-					</span>
-				</div>
-			</div>
+      <div className="mt-4 flex flex-col gap-3">
+        <div className="flex flex-col gap-1">
+          <span className="flex items-center justify-between text-[11px] text-muted-foreground">
+            Ladder
+            <span className="text-muted-foreground/80">0 · 1 · 2 · 3 · 4</span>
+          </span>
+          <input
+            type="range"
+            min={0}
+            max={4}
+            step={1}
+            value={params.ladderRung}
+            data-testid="param-ladder"
+            onChange={(e) =>
+              onParam({
+                ladderRung: Number(
+                  e.target.value,
+                ) as PracticeParams["ladderRung"],
+              })
+            }
+            className="h-1.5 w-full accent-primary"
+          />
+          <span data-testid="param-ladder-level" className="text-xs">
+            <span className="font-medium">
+              {LADDER_RUNGS[params.ladderRung].name}
+            </span>
+            <span className="text-muted-foreground">
+              {" "}
+              — {LADDER_RUNGS[params.ladderRung].definition}
+            </span>
+          </span>
+        </div>
+        <div className="flex flex-col gap-1">
+          <span className="flex items-center justify-between text-[11px] text-muted-foreground">
+            Cue distance
+            <span className="text-muted-foreground/80">0 · 1 · 2</span>
+          </span>
+          <input
+            type="range"
+            min={0}
+            max={2}
+            step={1}
+            value={params.cueDistance}
+            data-testid="param-cue"
+            onChange={(e) =>
+              onParam({
+                cueDistance: Number(
+                  e.target.value,
+                ) as PracticeParams["cueDistance"],
+              })
+            }
+            className="h-1.5 w-full accent-primary"
+          />
+          <span data-testid="param-cue-level" className="text-xs">
+            <span className="font-medium">
+              {CUE_LEVELS[params.cueDistance].name}
+            </span>
+            <span className="text-muted-foreground">
+              {" "}
+              — {CUE_LEVELS[params.cueDistance].definition}
+            </span>
+          </span>
+        </div>
+        <div className="flex flex-col gap-1">
+          <span className="text-[11px] text-muted-foreground">
+            Presentation
+          </span>
+          <div className="flex gap-1" data-testid="param-presentation">
+            {(["theory", "application"] as const).map((pres) => (
+              <button
+                type="button"
+                key={pres}
+                data-testid={`param-presentation-${pres}`}
+                onClick={() => onParam({ presentation: pres })}
+                className={`flex-1 rounded-md border px-3 py-1.5 text-xs transition-colors ${
+                  params.presentation === pres
+                    ? "border-primary bg-primary/10 font-medium text-foreground"
+                    : "text-muted-foreground hover:bg-accent"
+                }`}
+              >
+                {PRESENTATIONS[pres].name}
+              </button>
+            ))}
+          </div>
+          <span
+            data-testid="param-presentation-level"
+            className="text-xs text-muted-foreground"
+          >
+            {PRESENTATIONS[params.presentation].definition}
+          </span>
+        </div>
+        {/* The learner's free-form note — the last word before create:
+			    wishes for the problems, weighed by the agents, nothing owed. */}
+        <div className="flex flex-col gap-1">
+          <span className="text-[11px] text-muted-foreground">
+            Anything else? (optional)
+          </span>
+          <textarea
+            value={note}
+            data-testid="practice-note"
+            onChange={(e) => onNote(e.target.value)}
+            rows={3}
+            maxLength={2000}
+            placeholder="Anything you want the problems to lean toward — situations, emphasis, context about you. The agents weigh it; nothing is promised."
+            className="w-full resize-y rounded-lg border bg-background px-3 py-2 text-sm outline-none focus-visible:ring-1 focus-visible:ring-ring"
+          />
+        </div>
+      </div>
 
-			{failure && (
-				<div data-testid="practice-dialog-error" className="mt-3 rounded-lg border border-destructive/40 bg-destructive/10 px-3 py-2 text-sm text-destructive">
-					{failure.code === "empty_selection" ? "Nothing practicable is left in that cluster — the graph changed." : failure.error}
-					{failure.status === 503 && (
-						<button
-							type="button"
-							data-testid="practice-dialog-open-settings"
-							onClick={onOpenSettings}
-							className="ml-2 underline underline-offset-2"
-						>
-							Open settings
-						</button>
-					)}
-				</div>
-			)}
+      {failure && (
+        <div
+          data-testid="practice-dialog-error"
+          className="mt-3 rounded-lg border border-destructive/40 bg-destructive/10 px-3 py-2 text-sm text-destructive"
+        >
+          {failure.code === "empty_selection"
+            ? "Nothing practicable is left in that cluster — the graph changed."
+            : failure.error}
+          {failure.status === 503 && (
+            <button
+              type="button"
+              data-testid="practice-dialog-open-settings"
+              onClick={onOpenSettings}
+              className="ml-2 underline underline-offset-2"
+            >
+              Open settings
+            </button>
+          )}
+        </div>
+      )}
 
-			<div className="mt-5 flex justify-end gap-2">
-				<button
-					type="button"
-					onClick={onClose}
-					disabled={busy}
-					className="rounded-md px-3 py-1.5 text-sm text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
-				>
-					Cancel
-				</button>
-				<button
-					type="button"
-					data-testid="practice-dialog-create"
-					onClick={onRun}
-					disabled={!canRun}
-					className="rounded-md bg-primary px-3 py-1.5 text-sm font-medium text-primary-foreground transition-colors hover:bg-primary/90 disabled:opacity-40"
-				>
-					{busy ? "Creating…" : "Create"}
-				</button>
-			</div>
-		</div>
-	);
+      <div className="mt-5 flex justify-end gap-2">
+        <button
+          type="button"
+          onClick={onClose}
+          disabled={busy}
+          className="rounded-md px-3 py-1.5 text-sm text-muted-foreground transition-colors hover:bg-accent hover:text-foreground"
+        >
+          Cancel
+        </button>
+        <button
+          type="button"
+          data-testid="practice-dialog-create"
+          onClick={onRun}
+          disabled={!canRun}
+          className="rounded-md bg-primary px-3 py-1.5 text-sm font-medium text-primary-foreground transition-colors hover:bg-primary/90 disabled:opacity-40"
+        >
+          {busy ? "Creating…" : "Create"}
+        </button>
+      </div>
+    </div>
+  );
 }
 
 /** The stateful wrapper: fetches the cluster list, runs the create POST
  *  (nodeIds + the cluster's label as title + params), navigates on 202. */
 export function PracticeDialog({ onClose }: { onClose: () => void }) {
-	const router = useRouter();
-	const { open: openSettings } = useSettingsDialog();
-	const [params, setParams] = useState<PracticeParams>({ ...DEFAULT_PRACTICE_PARAMS });
-	const [clusters, setClusters] = useState<GraphClusterGroup[] | null>(null);
-	const [selectedKey, setSelectedKey] = useState<string | null>(null);
-	const [busy, setBusy] = useState(false);
-	const [failure, setFailure] = useState<PracticeFailure | null>(null);
+  const router = useRouter();
+  const { open: openSettings } = useSettingsDialog();
+  const [params, setParams] = useState<PracticeParams>({
+    ...DEFAULT_PRACTICE_PARAMS,
+  });
+  const [note, setNote] = useState("");
+  const [clusters, setClusters] = useState<GraphClusterGroup[] | null>(null);
+  const [selectedKey, setSelectedKey] = useState<string | null>(null);
+  const [busy, setBusy] = useState(false);
+  const [failure, setFailure] = useState<PracticeFailure | null>(null);
 
-	// The picker's truth is the graph's clustering (fresh on open).
-	useEffect(() => {
-		let alive = true;
-		fetch("/api/graph/clusters", { cache: "no-store" })
-			.then((res) => res.json() as Promise<{ clusters: GraphClusterGroup[] }>)
-			.then((d) => {
-				if (alive) setClusters(d.clusters);
-			})
-			.catch(() => {
-				if (alive) setClusters([]);
-			});
-		return () => {
-			alive = false;
-		};
-	}, []);
+  // The picker's truth is the graph's clustering (fresh on open).
+  useEffect(() => {
+    let alive = true;
+    fetch("/api/graph/clusters", { cache: "no-store" })
+      .then((res) => res.json() as Promise<{ clusters: GraphClusterGroup[] }>)
+      .then((d) => {
+        if (alive) setClusters(d.clusters);
+      })
+      .catch(() => {
+        if (alive) setClusters([]);
+      });
+    return () => {
+      alive = false;
+    };
+  }, []);
 
-	// Esc closes (claimed, the delete-session discipline).
-	useEffect(() => {
-		const onKeyDown = (e: KeyboardEvent) => {
-			if (e.key === "Escape") {
-				e.stopPropagation();
-				onClose();
-			}
-		};
-		window.addEventListener("keydown", onKeyDown);
-		return () => window.removeEventListener("keydown", onKeyDown);
-	}, [onClose]);
+  // Esc closes (claimed, the delete-session discipline).
+  useEffect(() => {
+    const onKeyDown = (e: KeyboardEvent) => {
+      if (e.key === "Escape") {
+        e.stopPropagation();
+        onClose();
+      }
+    };
+    window.addEventListener("keydown", onKeyDown);
+    return () => window.removeEventListener("keydown", onKeyDown);
+  }, [onClose]);
 
-	const run = useCallback(async () => {
-		const picked = (clusters ?? []).find((c) => c.key === selectedKey);
-		if (!picked || busy) return;
-		setBusy(true);
-		setFailure(null);
-		try {
-			const res = await fetch("/api/practice/sets", {
-				method: "POST",
-				headers: { "content-type": "application/json" },
-				body: JSON.stringify({
-					nodeIds: picked.nodeIds,
-					params,
-					title: picked.label ?? undefined,
-				}),
-			});
-			const body = (await res.json()) as { id?: string; error?: string; code?: string };
-			if (!res.ok || !body.id) {
-				setFailure({ status: res.status, error: body.error ?? `Create failed (${res.status})`, code: body.code });
-				setBusy(false);
-				return;
-			}
-			router.push(`/practice/${body.id}`);
-		} catch {
-			setFailure({ status: 0, error: "Create failed — network error. Try again." });
-			setBusy(false);
-		}
-	}, [clusters, selectedKey, busy, params, router]);
+  const run = useCallback(async () => {
+    const picked = (clusters ?? []).find((c) => c.key === selectedKey);
+    if (!picked || busy) return;
+    setBusy(true);
+    setFailure(null);
+    try {
+      const res = await fetch("/api/practice/sets", {
+        method: "POST",
+        headers: { "content-type": "application/json" },
+        body: JSON.stringify({
+          nodeIds: picked.nodeIds,
+          params,
+          title: picked.label ?? undefined,
+          note: note.trim() || undefined,
+        }),
+      });
+      const body = (await res.json()) as {
+        id?: string;
+        error?: string;
+        code?: string;
+      };
+      if (!res.ok || !body.id) {
+        setFailure({
+          status: res.status,
+          error: body.error ?? `Create failed (${res.status})`,
+          code: body.code,
+        });
+        setBusy(false);
+        return;
+      }
+      router.push(`/practice/${body.id}`);
+    } catch {
+      setFailure({
+        status: 0,
+        error: "Create failed — network error. Try again.",
+      });
+      setBusy(false);
+    }
+  }, [clusters, selectedKey, busy, params, note, router]);
 
-	return createPortal(
-		<div
-			className="fixed inset-0 z-50 flex items-center justify-center bg-background/80"
-			onClick={() => {
-				if (!busy) onClose();
-			}}
-		>
-			<PracticeDialogBody
-				params={params}
-				onParam={(patch) => setParams((prev) => ({ ...prev, ...patch }))}
-				clusters={clusters ?? []}
-				selectedKey={selectedKey}
-				onSelect={setSelectedKey}
-				busy={busy}
-				failure={failure}
-				onOpenSettings={() => openSettings()}
-				onClose={onClose}
-				onRun={() => void run()}
-			/>
-		</div>,
-		document.body,
-	);
+  return createPortal(
+    <div
+      className="fixed inset-0 z-50 flex items-center justify-center bg-background/80"
+      onClick={() => {
+        if (!busy) onClose();
+      }}
+    >
+      <PracticeDialogBody
+        params={params}
+        onParam={(patch) => setParams((prev) => ({ ...prev, ...patch }))}
+        note={note}
+        onNote={setNote}
+        clusters={clusters ?? []}
+        selectedKey={selectedKey}
+        onSelect={setSelectedKey}
+        busy={busy}
+        failure={failure}
+        onOpenSettings={() => openSettings()}
+        onClose={onClose}
+        onRun={() => void run()}
+      />
+    </div>,
+    document.body,
+  );
 }
diff --git a/docs/architecture.md b/docs/architecture.md
index d5269b6..7d3ad83 100644
--- a/docs/architecture.md
+++ b/docs/architecture.md
@@ -382,7 +382,7 @@ Practice turns the graph into problem sets the learner reasons through out loud.
 
 **The jobs pattern — registry, not JobRecords.** Practice runs are NOT Workbench jobs: no `JobStore` records, no steer messages, nothing ever wakes the teacher — a practice run isn't chat-scoped (a set outlives its birth session and has no lesson to re-enter; its result is itself, persisted in its own store and read by its own page). What it shares with the Workbench is the observability surface: a `globalThis` run registry (the module-instance idiom — `__lhPracticeRuns`) with PER-PROBLEM ROWS (owner ruling, 2026-09-29 — matching the extraction per-chunk pattern): `setId:partition` while the partitioner runs ("Partitioning — title"), then one `setId:gen:problemId` per problem ("queued"→"generating", label "Problem i/N — title"), and one `setId:grade:problemId` per graded answer ("queued"→"grading", "Grading i/N — title") — each row its OWN abort controller and its OWN usage accumulation; the row key IS the feed id (the dock dup-key law) and the abort target. ABORT ROUTES BY ROW SHAPE: a gen-row abort kills THAT problem only (siblings continue; the set still settles `ready` with one errored problem), a partition-row abort ends the whole run, a grade-row abort drops that grade (answer kept — one attempt, the problem stays done); set-level abort (delete) composes from row aborts plus the synchronous persisted marks. Model calls — partitioner included — run under the SHARED background-LLM semaphore (`bg-llm-semaphore.ts`): one process-wide pool of 3 permits for ALL background LLM jobs, extraction chunks included (total background LLM concurrency bounded at 3, whoever calls). A boot sweep marks orphaned `partitioning`/`generating` sets as `error "Interrupted (server restarted)"` — once per process, and only when the set's `updatedAt` heartbeat is >120s stale (a live run's 30s FILE watchdog keeps the heartbeat fresh even across dev's per-worker processes — the registry is not shared between them; the guard trio closed a live-stamping bug from the first live test) — the job runner's crash-honesty at set grain. Store writes serialize behind a per-set mutex; the pipeline is the sole writer of `data/practice`.
 
-**API + UI.** Creation is GRAPH-DOMAIN, not chat-domain (owner ruling 2026-09-29: "problem sets always take in nodes — graph objects"): `POST /api/practice/sets` takes `{nodeIds, params, title?}` ONLY — no chatId, no clusterId; the UI resolves a picked cluster to its member node ids and passes the cluster's label as the title. `GET /api/graph/clusters` serves the picker: the birth-session partition itself (`birthSessionPartition`, same ordering — size desc, ties by id) — EVERY group regardless of size (owner ruling, same evening: MIN_CLUSTER_SIZE is a hull-drawing rule; practice availability must not inherit a visual threshold), each as `{key, kind: "chat"|"notes", label, nodeIds, nodeCount}`; a cluster whose naming record is gone keeps `label: null` — honest, never guessed. 202 `{id}` with generation fire-and-forget; 400 empty selection (empty intersection with live nodes — the vanished-nodes edge) / bad params; 503 model/key refusal. `GET /api/practice/sets` · `GET/DELETE /api/practice/sets/[id]` (delete aborts live runs first) · `POST /api/practice/sets/[id]/answers` (202; 409 when not ready; 400 re-answering a submitted problem — one attempt). Surfaces: `/practice` (the list — the chat-pattern top bar with the sidebar reopen toggle and the New action at the right edge, status/counts rows, delete confirm, the new-set dialog with the named-level controls (a 5-detent ladder slider, a 3-detent cue slider, a Theory|Application toggle — each showing the level's name + definition) and the cluster picker — the ONLY creation entry for now; the graph-view cluster entry comes later with the free-form selection UI) and `/practice/[id]` (the exam page — the chat-pattern top bar, a levels + status line, numbered problems, markdown statements with math healing at the display seam, a textarea + focus-gated dictation bar per problem with its OWN Submit button — one attempt per problem, instant solution/rubric reveal, the score landing live). Entry: the sidebar's Practice nav row.
+**API + UI.** Creation is GRAPH-DOMAIN, not chat-domain (owner ruling 2026-09-29: "problem sets always take in nodes — graph objects"): `POST /api/practice/sets` takes `{nodeIds, params, title?, note?}` ONLY — no chatId, no clusterId; the UI resolves a picked cluster to its member node ids and passes the cluster's label as the title. The `note` is the learner's optional free-form creation input (owner, same evening: more learner control, no other purpose): trimmed, capped at 2000 characters, persisted on the set, and carried to the partitioner and every generator as `THE LEARNER'S NOTE` — an additional input the agents WEIGH but are never obliged to follow (the doctrine pins the stance: it never moves a problem off its rung or out of its band, and the partition laws always win). `GET /api/graph/clusters` serves the picker: the birth-session partition itself (`birthSessionPartition`, same ordering — size desc, ties by id) — EVERY group regardless of size (owner ruling, same evening: MIN_CLUSTER_SIZE is a hull-drawing rule; practice availability must not inherit a visual threshold), each as `{key, kind: "chat"|"notes", label, nodeIds, nodeCount}`; a cluster whose naming record is gone keeps `label: null` — honest, never guessed. 202 `{id}` with generation fire-and-forget; 400 empty selection (empty intersection with live nodes — the vanished-nodes edge) / bad params; 503 model/key refusal. `GET /api/practice/sets` · `GET/DELETE /api/practice/sets/[id]` (delete aborts live runs first) · `POST /api/practice/sets/[id]/answers` (202; 409 when not ready; 400 re-answering a submitted problem — one attempt). Surfaces: `/practice` (the list — the chat-pattern top bar with the sidebar reopen toggle and the New action at the right edge, status/counts rows, delete confirm, the new-set dialog with the named-level controls (a 5-detent ladder slider, a 3-detent cue slider, a Theory|Application toggle — each showing the level's name + definition), the cluster picker, and the optional free-form note textarea (typing-only) — the ONLY creation entry for now; the graph-view cluster entry comes later with the free-form selection UI) and `/practice/[id]` (the exam page — the chat-pattern top bar, a levels + status line, numbered problems, markdown statements with math healing at the display seam, a textarea + focus-gated dictation bar per problem with its OWN Submit button — one attempt per problem, instant solution/rubric reveal, the score landing live). Entry: the sidebar's Practice nav row.
 
 **Model wiring.** Three roles — `partitioner`, `problemGenerator`, `grader` — all pinned to the **strong** slot (quality-critical, one settings edit from cheap) and in `BACKGROUND_MODEL_ROLES` (per-role overrides). The stage agents are single-submit loops on the shared `runSubmitAgent` seam (`submit-agent.ts`, extracted from extraction for exactly this): a `validate` hook turns a bad submit into a tool-input error the model retries against, within `maxTurns 4`; exhaustion fails the stage honestly (the set says so, the chains re-run cleanly on a new set).
 
diff --git a/next-env.d.ts b/next-env.d.ts
index a419cbe..ce4e94a 100644
--- a/next-env.d.ts
+++ b/next-env.d.ts
@@ -1,7 +1,7 @@
 /// <reference types="next" />
 /// <reference types="next/image-types/global" />
-import "./.next/dev/types/routes.d.ts";
-import "./.next/dev/types/root-params.d.ts";
+import "./.next/types/routes.d.ts";
+import "./.next/types/root-params.d.ts";
 
 // NOTE: This file should not be edited
 // see https://nextjs.org/docs/app/api-reference/config/typescript for more information.
diff --git a/src/agent/subagents/practice-pipeline.ts b/src/agent/subagents/practice-pipeline.ts
index f080aea..74fa988 100644
--- a/src/agent/subagents/practice-pipeline.ts
+++ b/src/agent/subagents/practice-pipeline.ts
@@ -275,6 +275,7 @@ export function createPracticePipeline(deps: PracticePipelineDeps) {
 		set: ProblemSet,
 		previousProblems: PreviousProblemRef[],
 		partitionRow: RunningPractice,
+		note?: string,
 	): Promise<void> {
 		const controller = partitionRow.controller;
 
@@ -308,6 +309,7 @@ export function createPracticePipeline(deps: PracticePipelineDeps) {
 					propositions: set.propositions,
 					params: set.params,
 					previousProblems,
+					note,
 					signal: controller.signal,
 				}),
 			);
@@ -370,6 +372,7 @@ export function createPracticePipeline(deps: PracticePipelineDeps) {
 							propositions: set.propositions,
 							params: set.params,
 							previousProblems,
+							note,
 							signal: row.controller.signal,
 						});
 						row.usage = generated.log;
@@ -431,6 +434,7 @@ export function createPracticePipeline(deps: PracticePipelineDeps) {
 		nodeIds: string[],
 		params: ProblemSet["params"],
 		titleHint?: string,
+		note?: string,
 	): Promise<{ set: ProblemSet; done: Promise<void> }> {
 		agents.refuseIfStagesUnready();
 
@@ -458,11 +462,12 @@ export function createPracticePipeline(deps: PracticePipelineDeps) {
 			chains,
 			status: "partitioning",
 			problems: [],
+			...(note ? { note } : {}),
 		});
 
 		const previousProblems = dedupInput(deps.practiceStore, selectedIds);
 		const row = registerRow(`${set.id}:partition`, set.id, "generation", "partitioning", `Partitioning — ${title}`);
-		const done = generateSet(set, previousProblems, row);
+		const done = generateSet(set, previousProblems, row, note);
 		return { set, done };
 	}
 
diff --git a/src/agent/subagents/practice.ts b/src/agent/subagents/practice.ts
index 96d51b6..9a76a73 100644
--- a/src/agent/subagents/practice.ts
+++ b/src/agent/subagents/practice.ts
@@ -188,6 +188,7 @@ INPUTS (in the user message):
   THE CUE DISTANCE: 0 Lesson context · 1 Adjacent context · 2 Far context — how far the situation sits from the cues the learner has already met: the lesson's own examples AND every previous problem's situation. Far context is the edge of the recognizable band — the proposition must still genuinely apply, and the learner must be able to find it by looking.
   THE PRESENTATION: Theory (the subject's own terms — symbols, formal objects, native vocabulary) or Application (a concrete worldly affair in which the proposition's structure appears in the wild). It dresses the situation; it never changes the rung or the cue distance.
   Apply Part II step 2: degrade honestly if the target cannot support them — the degradation shows in the problem's own shape, never in meta-commentary.
+- THE LEARNER'S NOTE (optional): free-form text the learner attached when creating the set — wishes, emphasis, context about themselves. Weigh it as preference, not instruction: take what serves the target and the parameters, ignore what conflicts with them, and never let it move a problem off its rung or out of its band. Its absence means nothing.
 - FORMATTING: your statement, solution, and rubric are rendered as full Markdown — use everything it offers. Mathematics is always LaTeX: inline as \\(…\\), display as \\[…\\] on its own lines — a bare $ is literal currency and never starts math. Inside math, use LaTeX commands for symbols, not unicode stand-ins (they fail to render). Code, pseudocode, and formal systems live in fenced code blocks with a language tag; identifiers and symbols inline as \`code\`. Use tables for parallel structure and comparisons, lists for enumerations. Skip headers — at this size they are pointless decoration; structure with paragraphs, bold labels, and lists (a genuinely long derivation may earn one). Prose-only renderings of mathematical or structured content are a defect: if a sentence carries an equation, a derivation, or a piece of code, format it as what it is.
 - PREVIOUS PROBLEMS: problems the learner has already answered, with the chains they exercised. Do not rebuild one; keep your situation distant from theirs — cue distance counts previous problems' situations (Part I, "Cue distance").
 
@@ -208,6 +209,7 @@ PARTITION DOCTRINE:
 - READ THE MIX OFF THE SHAPE: many singletons and few multi-premise chains suggest recall + discrimination problems; long premise chains suggest composition and path-finding. Your partition is where that judgment lives.
 - RESPECT THE PARAMETERS: the request names the ladder rung, the cue distance, and the presentation. Low rungs (0–2) mean many small assignments — single chains standing alone; high rungs (3–4) mean larger assignments — composition needs material to chain, path-finding needs distractor material to conceal the path among. Far cue distance makes grouping confusables together MORE important: the discrimination must travel into the far situation. Presentation does not shape your partition.
 - DO NOT DUPLICATE: previous problems the learner has already seen are listed with their chains. If an existing problem already covers a group of chains, re-partition so the new set differs in what it groups and how.
+- THE LEARNER'S NOTE (optional): free-form text the learner attached at creation — it may wish for grouping, count, or emphasis. Weigh it as preference; the partition laws above always win, and when the note conflicts with them, follow the laws silently.
 
 OUTPUT — your entire deliverable is ONE submit_partition call: "partitions", a list of { chains: [ids] }, one entry per problem, together containing every input chain exactly once. No empty entries; no unknown ids; no problem without a real target relationship.`;
 
@@ -324,6 +326,8 @@ export interface PartitionerInput {
 	propositions: PropositionSnapshot;
 	params: PracticeParams;
 	previousProblems: PreviousProblemRef[];
+	/** The learner's creation note — optional; weighed, never obeyed. */
+	note?: string;
 	signal?: AbortSignal;
 }
 export interface PartitionerOutput {
@@ -339,6 +343,8 @@ export interface GeneratorInput {
 	propositions: PropositionSnapshot;
 	params: PracticeParams;
 	previousProblems: PreviousProblemRef[];
+	/** The learner's creation note — optional; weighed, never obeyed. */
+	note?: string;
 	signal?: AbortSignal;
 }
 export interface GeneratedProblem {
@@ -394,6 +400,14 @@ function validateProblem(value: ProblemSubmit, target: string[]): string | null
 	return null;
 }
 
+/** The learner's creation note as a prompt block — present ONLY when a
+ *  note exists (owner, 2026-09-29: more learner control, nothing owed). */
+function learnerNoteBlock(note: string | undefined): string | null {
+	return note
+		? `THE LEARNER'S NOTE (free-form additional input — weigh it; you are not obliged to follow it)\n${note}`
+		: null;
+}
+
 export function createPracticeAgents(deps: PracticeModels) {
 	const stageEngine = (role: ModelRole, override?: PiEngine): PiEngine | undefined =>
 		override ?? deps.engineFor?.(role);
@@ -433,11 +447,13 @@ export function createPracticeAgents(deps: PracticeModels) {
 		const log: PracticeStepLog = { model: null, steps: [] };
 		const prompt = [
 			renderChains(input.chains, input.propositions),
-			"THE PARAMETERS",
 			renderParams(input.params),
+			learnerNoteBlock(input.note),
 			"PREVIOUS PROBLEMS (already seen — do not duplicate)",
 			renderPreviousProblems(input.previousProblems),
-		].join("\n\n");
+		]
+			.filter((block): block is string => block !== null)
+			.join("\n\n");
 		const submitted = await runSubmitAgent({
 			role: "partitioner",
 			toolName: "submit_partition",
@@ -461,11 +477,13 @@ export function createPracticeAgents(deps: PracticeModels) {
 		const prompt = [
 			`YOUR TARGET (the chains this problem is built to strengthen)\n${targetChains.map(chainLine).join("\n")}`,
 			`THE FULL SET — propositions and chains of the entire selection (supporting material: minimal use, maximal focus on the target)\n${renderChains(input.fullChains, input.propositions)}`,
-			"THE PARAMETERS",
 			renderParams(input.params),
+			learnerNoteBlock(input.note),
 			"PREVIOUS PROBLEMS (already answered — do not rebuild one; keep your situation distant from theirs)",
 			renderPreviousProblems(input.previousProblems),
-		].join("\n\n");
+		]
+			.filter((block): block is string => block !== null)
+			.join("\n\n");
 		const problem = await runSubmitAgent({
 			role: "problemGenerator",
 			toolName: "submit_problem",
diff --git a/src/core/contracts/practice.ts b/src/core/contracts/practice.ts
index d17c52a..c8240c3 100644
--- a/src/core/contracts/practice.ts
+++ b/src/core/contracts/practice.ts
@@ -187,6 +187,13 @@ export interface ProblemSet {
 	status: ProblemSetStatus;
 	/** Set when the partition itself (or the run as a whole) failed. */
 	error?: string;
+	/** The learner's free-form note from the creation dialog (owner,
+	 *  2026-09-29: more learner control, no other purpose). Carried to the
+	 *  partitioner and every generator as ADDITIONAL INPUT — weighed as
+	 *  preference, never an obligation, never overriding the laws. Trimmed,
+	 *  capped (2000 chars at the API); persisted so the set stays a
+	 *  self-contained document of what was asked for. */
+	note?: string;
 	/** Problem stubs appear here the moment the partitioner submits —
 	 *  numbered by position, "generating", propositions invisible. */
 	problems: Problem[];
diff --git a/src/server/practice.ts b/src/server/practice.ts
index 7b639da..9dadc32 100644
--- a/src/server/practice.ts
+++ b/src/server/practice.ts
@@ -81,11 +81,16 @@ export function practiceErrorResponse(err: unknown): Response {
 	return Response.json({ error: "Internal error" }, { status: 500 });
 }
 
+/** The learner-note cap (creation dialog; the textarea's own maxLength
+ *  matches — server-side is the law). */
+const NOTE_MAX = 2000;
+
 /** POST /api/practice/sets — 202 {id}; the set generates in the
  *  background (fire-and-forget pipeline). Body is {nodeIds, params,
- *  title?} ONLY (owner ruling 2026-09-29: problem sets take NODES —
+ *  title?, note?} ONLY (owner ruling 2026-09-29: problem sets take NODES —
  *  graph objects; the UI resolves a picked cluster to its member ids,
- *  and may pass the cluster label as the title). */
+ *  and may pass the cluster label as the title; note is the learner's
+ *  free-form creation input, optional). */
 export async function createPracticeSetResponse(body: unknown): Promise<Response> {
 	const b = (body ?? {}) as Record<string, unknown>;
 	const params = parseParams(b.params);
@@ -97,17 +102,28 @@ export async function createPracticeSetResponse(body: unknown): Promise<Response
 	}
 	const titleHint =
 		typeof b.title === "string" && b.title.trim() !== "" ? b.title.trim() : undefined;
+	let note: string | undefined;
+	if (b.note !== undefined) {
+		if (typeof b.note !== "string") {
+			return Response.json({ error: "note must be a string" }, { status: 400 });
+		}
+		const trimmed = b.note.trim();
+		if (trimmed.length > NOTE_MAX) {
+			return Response.json({ error: `note is over ${NOTE_MAX} characters` }, { status: 400 });
+		}
+		note = trimmed || undefined; // whitespace-only = no note
+	}
 	if (
 		!Array.isArray(b.nodeIds) ||
 		!b.nodeIds.every((id): id is string => typeof id === "string" && ID_OK.test(id))
 	) {
-		return Response.json({ error: "Body must be {nodeIds, params, title?}" }, { status: 400 });
+		return Response.json({ error: "Body must be {nodeIds, params, title?, note?}" }, { status: 400 });
 	}
 	if (b.nodeIds.length === 0) {
 		return Response.json({ error: "nodeIds is empty", code: "empty_selection" }, { status: 400 });
 	}
 	try {
-		const { set } = await practice.createProblemSet(b.nodeIds, params, titleHint);
+		const { set } = await practice.createProblemSet(b.nodeIds, params, titleHint, note);
 		return Response.json({ id: set.id }, { status: 202 });
 	} catch (err) {
 		return practiceErrorResponse(err);
diff --git a/src/store/practice.ts b/src/store/practice.ts
index a66e7c3..fbbcfd9 100644
--- a/src/store/practice.ts
+++ b/src/store/practice.ts
@@ -111,6 +111,7 @@ export function isProblemSet(value: unknown): value is ProblemSet {
 		typeof v.id === "string" &&
 		v.id.startsWith("ps_") &&
 		typeof v.title === "string" &&
+		(v.note === undefined || typeof v.note === "string") &&
 		typeof v.createdAt === "string" &&
 		isParams(v.params) &&
 		isStringArray(v.nodeIds) &&
diff --git a/tests/practice-agents.test.ts b/tests/practice-agents.test.ts
index 35eec06..152b938 100644
--- a/tests/practice-agents.test.ts
+++ b/tests/practice-agents.test.ts
@@ -72,6 +72,25 @@ function silentModel(): PiEngine {
 	return fauxEngine([fauxAssistantMessage("I refuse to use tools"), fauxAssistantMessage("…")]);
 }
 
+/** A one-shot submitter that first records the LAST USER MESSAGE it saw —
+ *  for pinning what actually rode the prompt. */
+function capturingSubmitModel(name: string, input: unknown, seen: string[]): PiEngine {
+	return fauxEngine([
+		(context) => {
+			const messages = (context as { messages: Array<{ role: string; content: unknown }> }).messages;
+			for (let i = messages.length - 1; i >= 0; i--) {
+				const m = messages[i]!;
+				if (m.role !== "user") continue;
+				seen.push(typeof m.content === "string" ? m.content : JSON.stringify(m.content));
+				break;
+			}
+			return fauxAssistantMessage([fauxToolCall(name, input as ToolCall["arguments"], { id: `${name}-call` })], {
+				stopReason: "toolUse",
+			});
+		},
+	]);
+}
+
 // ─── rendering ───────────────────────────────────────────────────────────────
 
 describe("renderParams — named levels, never a number", () => {
@@ -164,6 +183,30 @@ describe("runPartitioner", () => {
 			/the partitioner never submitted a partition/,
 		);
 	});
+
+	it("the learner's note: rides the prompt as its own block when present, absent when unset", async () => {
+		const seen: string[] = [];
+		const agents = createPracticeAgents({
+			partitioner: capturingSubmitModel("submit_partition", { partitions: [{ chains: ["n3", "n2"] }] }, seen),
+		});
+		await agents.runPartitioner({
+			chains,
+			propositions,
+			params,
+			previousProblems: [],
+			note: "one big problem, please — and lean on pricing",
+		});
+		assert.equal(seen.length, 1);
+		assert.ok(seen[0]!.includes("THE LEARNER'S NOTE"), "the block title");
+		assert.ok(seen[0]!.includes("one big problem, please — and lean on pricing"), "the note text");
+
+		const seenBare: string[] = [];
+		const bare = createPracticeAgents({
+			partitioner: capturingSubmitModel("submit_partition", { partitions: [{ chains: ["n3", "n2"] }] }, seenBare),
+		});
+		await bare.runPartitioner({ chains, propositions, params, previousProblems: [] });
+		assert.ok(!seenBare[0]!.includes("THE LEARNER'S NOTE"), "no block when no note");
+	});
 });
 
 // ─── the problem generator ───────────────────────────────────────────────────
@@ -218,6 +261,31 @@ describe("runProblemGenerator", () => {
 		const agents = createPracticeAgents({ problemGenerator: silentModel() });
 		await assert.rejects(agents.runProblemGenerator(generatorInput), /the problem generator never submitted a problem/);
 	});
+
+	it("the learner's note: rides the prompt when present, absent when unset", async () => {
+		const seen: string[] = [];
+		const agents = createPracticeAgents({
+			problemGenerator: capturingSubmitModel(
+				"submit_problem",
+				{ chains: ["n3"], statement: "…", solution: "…", rubric: "…" },
+				seen,
+			),
+		});
+		await agents.runProblemGenerator({ ...generatorInput, note: "situations from central banking, if you can" });
+		assert.ok(seen[0]!.includes("THE LEARNER'S NOTE"));
+		assert.ok(seen[0]!.includes("situations from central banking, if you can"));
+
+		const seenBare: string[] = [];
+		const bare = createPracticeAgents({
+			problemGenerator: capturingSubmitModel(
+				"submit_problem",
+				{ chains: ["n3"], statement: "…", solution: "…", rubric: "…" },
+				seenBare,
+			),
+		});
+		await bare.runProblemGenerator(generatorInput);
+		assert.ok(!seenBare[0]!.includes("THE LEARNER'S NOTE"), "no block when no note");
+	});
 });
 
 // ─── the grader ──────────────────────────────────────────────────────────────
@@ -291,6 +359,13 @@ describe("refusal gates (BYOK honest refusal, extraction's pattern)", () => {
 // ─── formatting doctrine (owner-approved 2026-09-29) ─────────────────────────
 
 describe("formatting doctrine pins", () => {
+	it("GENERATOR_DOCTRINE pins the learner-note stance (weighed, never owed)", () => {
+		assert.ok(
+			GENERATOR_DOCTRINE.includes("- THE LEARNER'S NOTE (optional): free-form text the learner attached when creating the set"),
+		);
+		assert.ok(GENERATOR_DOCTRINE.includes("never let it move a problem off its rung or out of its band"));
+	});
+
 	it("GENERATOR_DOCTRINE carries the FORMATTING bullet verbatim (Markdown + LaTeX + fenced code + tables/lists)", () => {
 		assert.ok(
 			GENERATOR_DOCTRINE.includes(
diff --git a/tests/practice-pipeline.test.ts b/tests/practice-pipeline.test.ts
index c3238db..09aff23 100644
--- a/tests/practice-pipeline.test.ts
+++ b/tests/practice-pipeline.test.ts
@@ -183,6 +183,45 @@ function partitionModel(partitions: Array<{ chains: string[] }>): PiEngine {
 	]);
 }
 
+/** A partitioner that RECORDS the prompt it saw before submitting — the
+ *  learner-note thread pin. */
+function capturingPartitioner(partitions: Array<{ chains: string[] }>, seen: string[]): PiEngine {
+	return fauxEngine([
+		(context) => {
+			seen.push(lastUserText(context as never));
+			return fauxAssistantMessage(
+				[fauxToolCall("submit_partition", { partitions } as ToolCall["arguments"], { id: "p-call" })],
+				{ stopReason: "toolUse" },
+			);
+		},
+	]);
+}
+
+/** An echo generator that RECORDS each prompt before submitting. */
+function capturingEchoGenerator(seen: string[], count: number): PiEngine {
+	const factory: FauxResponseStep = (context) => {
+		const prompt = lastUserText(context as never);
+		seen.push(prompt);
+		const target = targetOf(prompt);
+		return fauxAssistantMessage(
+			[
+				fauxToolCall(
+				"submit_problem",
+				{
+					chains: target,
+					statement: `Problem on ${target.join(", ")}`,
+					solution: "The model solution.",
+					rubric: "- names it",
+				} as ToolCall["arguments"],
+					{ id: "g-call" },
+				),
+			],
+			{ stopReason: "toolUse" },
+		);
+	};
+	return fauxEngine(Array.from({ length: count }, () => factory));
+}
+
 /** Hangs until the call's signal aborts (then resolves no submit call) —
  *  the extraction.test.ts cancellation-mimicking idiom. Queue one step
  *  per concurrent call (faux queues are finite). */
@@ -360,6 +399,43 @@ describe("createProblemSet (generation)", () => {
 		assert.equal(practiceStatuses().length, 0);
 	});
 
+	it("the learner's note threads to the partitioner + every generator and persists on the set (owner, 2026-09-29)", async () => {
+		const partSeen: string[] = [];
+		const genSeen: string[] = [];
+		const f = fixture({
+			partitioner: capturingPartitioner([{ chains: ["b"] }, { chains: ["e"] }], partSeen),
+			problemGenerator: capturingEchoGenerator(genSeen, 2),
+			grader: echoGrader(1),
+		});
+		const pipeline = f.make();
+		const { set, done } = await pipeline.createProblemSet(["a", "b", "e"], PARAMS, undefined, "keep it geometric");
+		await done;
+
+		assert.equal(set.note, "keep it geometric", "persisted at creation");
+		assert.equal(f.practice.store.getSet(set.id)!.note, "keep it geometric", "survives the store round-trip");
+		assert.ok(partSeen[0]!.includes("THE LEARNER'S NOTE"), "rode the partitioner prompt");
+		assert.ok(partSeen[0]!.includes("keep it geometric"));
+		assert.equal(genSeen.length, 2);
+		for (const prompt of genSeen) {
+			assert.ok(prompt.includes("THE LEARNER'S NOTE"), "rode every generator prompt");
+			assert.ok(prompt.includes("keep it geometric"));
+		}
+
+		// No note → no block, anywhere (optional in every sense).
+		const partSeenBare: string[] = [];
+		const genSeenBare: string[] = [];
+		const f2 = fixture({
+			partitioner: capturingPartitioner([{ chains: ["b", "e"] }], partSeenBare),
+			problemGenerator: capturingEchoGenerator(genSeenBare, 1),
+			grader: echoGrader(1),
+		});
+		const s2 = await f2.make().createProblemSet(["a", "b", "e"], PARAMS);
+		await s2.done;
+		assert.equal(s2.set.note, undefined);
+		assert.ok(!partSeenBare[0]!.includes("THE LEARNER'S NOTE"));
+		assert.ok(!genSeenBare[0]!.includes("THE LEARNER'S NOTE"));
+	});
+
 	it("per-problem failure isolation: one generator fails, the set still completes", async () => {
 		const f = fixture({ partitioner: partitionModel([{ chains: ["b"] }, { chains: ["e"] }]), problemGenerator: echoGenerator(8, ["e"]), grader: echoGrader(1) });
 		const pipeline = f.make();
diff --git a/tests/practice-service.test.ts b/tests/practice-service.test.ts
index b986e5f..dafbc0a 100644
--- a/tests/practice-service.test.ts
+++ b/tests/practice-service.test.ts
@@ -41,3 +41,29 @@ describe("createPracticeSetResponse — strict named-level params", () => {
 		assert.equal(await statusOf("high"), 400);
 	});
 });
+
+describe("createPracticeSetResponse — the learner's note (owner, 2026-09-29)", () => {
+	const PARAMS = { ladderRung: 2, cueDistance: 1, presentation: "theory" } as const;
+
+	it("non-string note → 400", async () => {
+		const res = await createPracticeSetResponse({ ...NODE_IDS, params: PARAMS, note: 42 });
+		assert.equal(res.status, 400);
+	});
+
+	it("over 2000 characters → 400 (the API is the law; the textarea's maxLength mirrors it)", async () => {
+		const res = await createPracticeSetResponse({ ...NODE_IDS, params: PARAMS, note: "x".repeat(2001) });
+		assert.equal(res.status, 400);
+		const body = (await res.json()) as { error: string };
+		assert.match(body.error, /over 2000 characters/);
+	});
+
+	it("exactly 2000 passes the note check (boundary — the next check, not the note, refuses)", async () => {
+		// Malformed nodeIds AFTER a max-length note: reaching the nodeIds
+		// error proves the note check passed (all checks are rejecting-path;
+		// the pipeline is never reached — this suite's doctrine).
+		const res = await createPracticeSetResponse({ nodeIds: 5, params: PARAMS, note: "x".repeat(2000) });
+		assert.equal(res.status, 400);
+		const body = (await res.json()) as { error: string };
+		assert.match(body.error, /Body must be/);
+	});
+});
diff --git a/tests/practice-store.test.ts b/tests/practice-store.test.ts
index 1401a63..6f09db3 100644
--- a/tests/practice-store.test.ts
+++ b/tests/practice-store.test.ts
@@ -152,6 +152,13 @@ describe("isProblemSet", () => {
 		assert.ok(!isProblemSet({ ...init(), status: "cooked" }));
 	});
 
+	it("the learner's note: optional string, round-trips; non-string refused (owner, 2026-09-29)", () => {
+		assert.ok(isProblemSet({ ...init({ note: "keep it geometric" }), id: "ps_noted" }));
+		assert.ok(!isProblemSet({ ...init({ note: 42 as unknown as string }), id: "ps_badnote" }));
+		const created = practiceStore.createSet(init({ note: "lean toward pricing" }));
+		assert.equal(practiceStore.getSet(created.id)!.note, "lean toward pricing");
+	});
+
 	it("tolerates optional fields — a minimal stub (no error/answer/grade) reads", () => {
 		const minimal = { ...init({ problems: [{ id: "pp_1", chains: [], statement: "", solution: "", rubric: "", status: "generating" }] }), id: "ps_valid" };
 		assert.ok(isProblemSet(minimal));
diff --git a/tests/render-practice.test.tsx b/tests/render-practice.test.tsx
index 5ecafbe..c317e11 100644
--- a/tests/render-practice.test.tsx
+++ b/tests/render-practice.test.tsx
@@ -87,6 +87,8 @@ describe("PracticeDialogBody — creation (cluster picker)", () => {
 	const base = {
 		params: { ...DEFAULT_PRACTICE_PARAMS },
 		onParam: () => {},
+		note: "",
+		onNote: () => {},
 		clusters: clusters(),
 		selectedKey: null,
 		onSelect: () => {},
@@ -126,6 +128,17 @@ describe("PracticeDialogBody — creation (cluster picker)", () => {
 		assert.match(moved, /param-presentation-application[^]*border-primary/, "application segment active");
 	});
 
+	it("the learner's note: optional textarea under the params, carries text, capped client-side", () => {
+		const html = renderToStaticMarkup(<PracticeDialogBody {...base} note="lean toward pricing applications" />);
+		assert.ok(html.includes('data-testid="practice-note"'));
+		assert.ok(html.includes("Anything else? (optional)"), "the label");
+		assert.ok(html.includes("lean toward pricing applications"), "controlled value renders");
+		assert.match(html, /maxlength="2000"/i, "client-side cap mirrors the API law");
+		// Empty default: present but empty — optional in every sense.
+		const bare = renderToStaticMarkup(<PracticeDialogBody {...base} />);
+		assert.match(bare, /data-testid="practice-note"[^>]*><\/textarea>/, "no value when unset");
+	});
+
 	it("cluster picker: label + proposition-count rows, notes kind marked, Create gated until picked", () => {
 		const html = renderToStaticMarkup(<PracticeDialogBody {...base} />);
 		assert.ok(html.includes('data-testid="practice-dialog-clusters"'));