practice: the between-problems pause — conclude ends the turn, the learner's go-ahead paces the stream (round 16)

Owner ruling (2026-10-06): every concluded problem opens a pause. The
agent may speak its closing words about the learner's last thinking
after the conclude result, then ENDS ITS TURN — a system-prompt law
deliberately, per owner correction, NOT tool-level terminate. The next
problem is never posed until the learner explicitly moves on; the pause
is the session's teaching space (repairs, clarifications, follow-ups —
still surgical, never re-teaching).

The move-on signal is a first-class input: { type: "next" } appends a
practice-next marker (content = single-sourced NEXT_PROBLEM_TEXT — the
model's unambiguous go-ahead: pose per the selection rules or wrap if
the arc is complete). Honest 400s: practice sessions only, nothing
active (mid-problem input rides the dock), riding after any pending
dismissal synthesis exactly like text.

The fast path (owner approved: composer-band chip + Enter-on-empty):
- Composer.enterAction gains "next" — Enter on an EMPTY composer in
  the pause fires the signal; typing anything makes Enter a send again;
  busy (steer/stop) outranks the pause.
- The Next chip hovers at the composer band's top-right (pick-pill
  chrome register) whenever the pause is live and a problem has ever
  been posed; the placeholder becomes the hint ("Ask about this
  problem — or press Enter for the next one…", exact);
- the transcript renders the signal as a compact centered NEXT PROBLEM
  line (mono-caps chrome); session-core appends it optimistically.

CONCLUDE_ACK_TEXT rewritten to carry the law; doctrine §VIII gains the
pause law (owner-sanctioned edit #3); 00-base's conclude/dismissal
bullets rewritten + the between-problems section added.

src/agent/prompts/practice/00-base.md

@@ -13,8 +13,10 @@ Two tools, and everything else is conversation. The mechanics only — the metho
 
 Only one problem may be active at a time — the tool rejects a pose while one is live. Conclude before posing the next.
 
-**`conclude`** — ends the active problem: records the outcome and verdict. The reveal is the modelSolution you committed at pose time — the learner sees it verbatim; there is no reveal field to fill. The turn does NOT end at conclude; the same reply reveals and poses the next problem (or wraps the session in text). The tool rejects a conclude with nothing active.
+**`conclude`** — ends the active problem: records the outcome and verdict. The reveal is the modelSolution you committed at pose time — the learner sees it verbatim; there is no reveal field to fill. After the conclude's result you may speak your closing words about the learner's work — the read of their last thinking, what the verdict means, one earned credit — and then you END YOUR TURN: no further tool calls in that reply, never another problem. The next problem is posed only on the learner's explicit go-ahead (below). The tool rejects a conclude with nothing active.
 
-**Dismissals.** The learner may discard a problem without concluding it. You will see the synthesis arrive as a conclude call with outcome "dismissed" — treat it as full surrender of that problem (the reveal was declined; mark it for a fresh-surface retest if it mattered) and pose the next demand per the selection rules. Do not comment on the dismissal beyond what the stream needs.
+**Between problems (the pause).** Every concluded problem opens a pause, and the pause is where this session's teaching lives: the learner can ask about the problem just finished — what went wrong, why a move was right, what the hint meant. Answer fully but surgically: repairs, clarifications, follow-up questions — the teaching session's craft at practice scale, never re-teaching. Each answer ends its turn; never pose a problem at the end of one. The next problem is posed ONLY when the learner explicitly moves on — a *Next problem* message (their one-click signal), or their own words asking for the next one. On that signal, pose per the selection rules — or, if the arc is complete, wrap the session in text. A clarification question is never the go-ahead; when in doubt, answer and wait.
+
+**Dismissals.** The learner may discard a problem without concluding it. You will see the synthesis arrive as a conclude call with outcome "dismissed" — treat it as full surrender of that problem (the reveal was declined; mark it for a fresh-surface retest if it mattered). The dismissal concludes the problem like any other: the pause law applies — no new problem until the learner's explicit go-ahead. Do not comment on the dismissal beyond what the stream needs.
 
 **The rarely-needed set.** Beyond the two tools sit the learner's files (read, ls, find, grep), the research and pdf jobs, and the map readers (list_sessions, list_propositions). Reach for them only when the situation genuinely requires: the session opened without picked propositions and targets must be chosen — look at the map; the learner references a file — read it; a load-bearing fact is uncertain — research it. These are exceptions, not habits — the session's own material and its proposition list come first.

src/agent/prompts/practice/10-doctrine.md

@@ -1,7 +1,8 @@
-<!-- Copied from the owner's doctrine (2026-10-04 20.37) with one approved edit:
-     §I material-arrival generalized (verbal descriptions, outside problems).
-     Tool mechanics live in 00-base.md; theory in
-     docs/vision-practice-sessions.md's source note. -->
+<!-- Copied from the owner's doctrine (2026-10-04 20.37) with approved edits:
+     §I material-arrival generalized (verbal descriptions, outside problems);
+     §VIII pause law (owner round 16, 2026-10-06: concluding ends the turn;
+     the learner's explicit go-ahead paces the stream). Tool mechanics live
+     in 00-base.md; theory in docs/vision-practice-sessions.md's source note. -->
 
 # The practice session
 
@@ -116,7 +117,7 @@ Classify every miss: cue miss (nothing came), discrimination error (wrong candid
 
 ## VIII. Progression
 
-A claimed answer is not a signal to progress. You own progression, not the answer count. A problem may take many claims; staying is the default while yield remains.
+A claimed answer is not a signal to progress. You own progression, not the answer count. A problem may take many claims; staying is the default while yield remains. Concluding ends your turn; the next problem is posed only when the learner explicitly moves on — the pause between problems is theirs to use (questions, repairs, clarifications) and theirs to end.
 
 Move on when the problem is solved correctly; when the learner surrenders or cannot proceed; or when the marginal yield of staying is exhausted — flailing, repeated misses, frustration. Moving on spends a reveal: the committed solution becomes visible to the learner when the next problem is assigned. Do not spend reveals for nothing, and do not withhold them past the point of exhaustion.
 

src/core/contracts/chat.ts

@@ -193,9 +193,20 @@ export const THINKING_LABEL = "Your thinking";
 
 /** The conclude tool's ack — shared by the executed call (agent layer) and
  *  the dismissal-synthesized pair (input-events), so a concluded problem
- *  reads identically to the model however it concluded. */
+ *  reads identically to the model however it concluded. Owner round 16:
+ *  the turn ENDS after concluding (a system-prompt law, no tool-level
+ *  terminate — the model may speak its closing words first, then must
+ *  stop calling tools); the next problem waits for the learner's
+ *  explicit go-ahead. */
 export const CONCLUDE_ACK_TEXT =
-	"Problem concluded — the composer is free; pose the next problem or wrap up.";
+	"Problem concluded. Say your closing words about the learner's work if useful, then END YOUR TURN — no further tool calls. The next problem is NEVER posed here: it waits for the learner's explicit go-ahead (a *Next problem* message, or their own words asking for the next one). Between problems, answer their questions about the concluded one.";
+
+/** The learner's explicit move-on signal (owner round 16, 2026-10-06):
+ *  the model-facing text of the practice-next marker — single-sourced
+ *  like THINKING_LABEL. The button/Enter-on-empty path appends the
+ *  marker; the model reads this and poses (or wraps) per the doctrine. */
+export const NEXT_PROBLEM_TEXT =
+	"[NEXT PROBLEM] — the learner's explicit go-ahead: pose the next problem now per the selection rules, or wrap the session in text if the arc is complete.";
 
 /** The compaction marker's model-facing header (owner round 14,
  *  2026-10-05). The marker's content is this header + the summary — the

src/core/input-events.ts

@@ -32,6 +32,7 @@ import {
 	concludedByDismissalPair,
 	danglingQuestionOf,
 	newAnswerResult,
+	newNextMessage,
 	newProblemDismissedMarker,
 	newThinkingMessage,
 	newUserMessage,
@@ -48,7 +49,12 @@ export type ChatInput =
 	 *  thinking on the ACTIVE problem — one lane (owner 2026-10-05):
 	 *  whether it claims an answer is read from the text. One
 	 *  problem may carry many submissions; each is its own turn. */
-	| { type: "thinking"; problemCallId: string; text: string };
+	| { type: "thinking"; problemCallId: string; text: string }
+	/** The practice move-on signal (owner round 16): the learner's
+	 *  explicit go-ahead during the between-problems pause — the Next
+	 *  chip or Enter on an empty composer. Practice sessions only, and
+	 *  only with nothing active (mid-problem input rides the dock). */
+	| { type: "next" };
 
 /** A wake-pipeline steer delivery — same turn composition, server-origin.
  *  Wave C (2026-09-26): the idle wake carries EVERY queued marker as its
@@ -172,6 +178,8 @@ export function parseChatInput(input: unknown): ChatInput {
 				throw new ChatInputError("thinking input requires non-empty text");
 			}
 			return { type: "thinking", problemCallId: i.problemCallId, text: i.text };
+		case "next":
+			return { type: "next" };
 		default:
 			throw new ChatInputError(`unknown input type: ${String(i.type)}`);
 	}
@@ -206,6 +214,19 @@ export function inputAppends(record: { messages: LhMessage[] }, input: ChatInput
 			}
 			return [...dismissalConclusion(record), newThinkingMessage(input.problemCallId, input.text)];
 		}
+		case "next": {
+			// Honest 400s, the thinking gate's discipline: the signal is a
+			// between-problems affordance — practice sessions only, nothing
+			// active (mid-problem input rides the dock), and the client gates
+			// make violations unreachable in practice anyway.
+			if ((record as { kind?: string }).kind !== "practice") {
+				throw new ChatInputError("next input is a practice-session affordance");
+			}
+			if (activeProblemOf(record.messages)) {
+				throw new ChatInputError("next input requires no active problem");
+			}
+			return [...dismissalConclusion(record), newNextMessage()];
+		}
 	}
 }
 

src/core/native-messages.ts

@@ -39,6 +39,7 @@ import { QUESTION_TOOLS } from "./pending.ts";
 import {
 	CONCLUDE_ACK_TEXT,
 	HANDOFF_TEXT,
+	NEXT_PROBLEM_TEXT,
 	THINKING_LABEL,
 } from "./contracts/chat.ts";
 import { compactAnswerForModel, type QuestionAnswerData } from "./question-core.ts";
@@ -71,6 +72,7 @@ export interface LhMarkerMessage {
 		| "subagent-abort"
 		| "plan-block"
 		| "practice-thinking"
+		| "practice-next"
 		| "problem-dismissed"
 		| "compaction";
 	/** The model-facing text ("" for plan blocks — UI-only). */
@@ -248,6 +250,24 @@ export function newThinkingMessage(problemCallId: string, text: string): LhMarke
 	};
 }
 
+/** The learner's explicit move-on signal (owner round 16): appended by
+ *  the next input (the composer band's chip, or Enter on an empty
+ *  composer during the between-problems pause). The content is the
+ *  single-sourced NEXT_PROBLEM_TEXT — the model's unambiguous go-ahead;
+ *  the transcript renders its own compact line off the marker (never
+ *  the content). */
+export function newNextMessage(): LhMarkerMessage {
+	return {
+		id: newMessageId(),
+		role: "custom",
+		customType: "practice-next",
+		content: NEXT_PROBLEM_TEXT,
+		display: true,
+		details: {},
+		timestamp: Date.now(),
+	};
+}
+
 /** The problem-dismissal marker (ask-style discard, owner 2026-10-04):
  *  UI-ONLY — toLlmMessages filters it. The model's face arrives later, at
  *  the user's next input, as the synthesized conclude pair

src/core/view.ts

@@ -401,6 +401,18 @@ function projectMarker(marker: LhMarkerMessage, out: LhViewMessage[]): void {
 			});
 			return;
 		}
+		case "practice-next": {
+			// The learner's explicit move-on signal (owner round 16): a compact
+			// centered system line — the mono-caps chrome register, not prose.
+			// The model reads the marker's content; the transcript never shows
+			// it (the label is chrome, single-sourced in the renderer).
+			out.push({
+				id: marker.id,
+				role: "user",
+				parts: [{ type: "data-practice-next", id: `pn_${marker.id}`, data: null }],
+			});
+			return;
+		}
 		case "problem-dismissed":
 			// UI-only (the ask-style dismissal): the synthesized conclude
 			// pair — which rides the NEXT input — carries the record's face.