wordle-five: the engine, the reward, the solver and the probe that checks them

The Python is the source of truth; src/demos/wordle/engine.ts will be a port of it, and CI gates the two against a SHA-256 over all 21.2M (guess, answer) pattern pairs rather than a hand-picked vector file — a vector file only ever catches the cases somebody thought of. The reward is three weighted components, and the third one is the reason this demo is worth building. `solved` and `economy` pull toward winning. `consistency` pulls against them, because a player maximising information deliberately guesses words that cannot win — a word that splits the remaining candidates evenly teaches more than a word that might happen to be right. That is good play, and it costs consistency. The probe ladder proves the tension is real rather than asserted: inaction 0.0000 crude 0.0111 plausible 0.1224 candidate_only 0.8925 exhaustive 0.9031 oracle 0.9458 The two good policies are 0.05 apart and neither dominates — the entropy oracle takes 1.00 economy and 0.73 consistency, the candidate-only player takes 0.75 and 1.00. Which one wins is a decision about what you want, which is the whole argument the site exists to make. probe.py fails CI if either starts dominating. Two traps found by building it. `consistency` is scored over turns SPENT, not guesses accepted: counting only legal guesses hands a free 1.0 to a policy that plays one word and then jams the parser five times — one guess, no contradictions, perfect score. And `economy`'s denominator is the depth the SHIPPED solver reaches, not a depth-optimal search: entropy-greedy is not depth-optimal, so grading it against an exact optimum would make the oracle rung fail its own assertion on some seeds. The word lists are built from Wordnik (MIT) intersected with SCOWL, never from the original game's 2,315 answers. 4,603 answers makes this materially harder than the original, so the published SALET/3.4212 results are cited as belonging to that list and our own reference player's TARES/3.72 is measured here. verifiers is an optional extra. The engine, reward, solver and probe all run — and gate — without an RL stack resolvable. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_019mt6sHQHEnEYrJZvoMCJSB
2026-08-28 15:39:03 -07:00
parent 5a9ff8dda9
commit a56f097f28
54 changed files with 8201 additions and 0 deletions
@@ -0,0 +1,91 @@
+import type { ReactNode } from 'react';
+import { cn } from '@/lib/utils';
+
+export type StatTone = 'default' | 'positive' | 'warning' | 'danger' | 'info' | 'brand';
+
+export interface Stat {
+  /** Short. Two or three words; it sits above the number. */
+  label: string;
+  value: ReactNode;
+  /** One clarifying line, shown under the number at a smaller size. */
+  hint?: string;
+  tone?: StatTone;
+  /** Set when this number was derived under an edited reward, not recorded. */
+  edited?: boolean;
+}
+
+export interface StatStripProps {
+  stats: Stat[];
+  className?: string;
+  /** Announce changes as they happen. Off by default — the shell owns the
+   *  page's single live region and two competing ones talk over each other. */
+  live?: boolean;
+}
+
+const TONE: Record<StatTone, string> = {
+  default: 'text-fg',
+  positive: 'text-positive',
+  warning: 'text-warning',
+  danger: 'text-danger',
+  info: 'text-info',
+  brand: 'text-accent-fg',
+};
+
+/**
+ * A row of headline numbers. Scrolls horizontally on a phone rather than
+ * wrapping into a ragged grid: four stats reflowing to 2x2 at 390px puts the
+ * least important number in the most prominent corner.
+ */
+export function StatStrip({ stats, className, live = false }: StatStripProps) {
+  if (stats.length === 0) return null;
+  return (
+    <dl
+      className={cn(
+        'flex snap-x snap-mandatory gap-3 overflow-x-auto pb-1',
+        'sm:grid sm:snap-none sm:overflow-visible sm:pb-0',
+        stats.length <= 2 ? 'sm:grid-cols-2' : 'sm:grid-cols-3 lg:grid-cols-4',
+        className,
+      )}
+      {...(live ? { 'aria-live': 'polite' as const } : {})}
+    >
+      {stats.map((stat) => (
+        <div
+          key={stat.label}
+          className="card min-w-[9.5rem] flex-1 shrink-0 snap-start px-4 py-3"
+        >
+          <dt className="flex items-center gap-1.5 text-xs font-medium uppercase tracking-wide text-muted">
+            <span className="truncate">{stat.label}</span>
+            {stat.edited ? <EditedChip /> : null}
+          </dt>
+          <dd
+            className={cn(
+              'nums mt-1 text-2xl font-semibold leading-tight',
+              TONE[stat.tone ?? 'default'],
+            )}
+          >
+            {stat.value}
+          </dd>
+          {stat.hint ? <dd className="mt-0.5 text-xs text-muted">{stat.hint}</dd> : null}
+        </div>
+      ))}
+    </dl>
+  );
+}
+
+/**
+ * Marks a number the visitor caused rather than one we recorded. It appears on
+ * every derived value in the reward editor; without it, an edited ranking
+ * screenshots identically to a measured one.
+ */
+export function EditedChip({ className }: { className?: string }) {
+  return (
+    <span
+      className={cn(
+        'rounded-md bg-accent-subtle px-1.5 py-0.5 text-[10px] font-semibold uppercase tracking-wide text-accent-fg',
+        className,
+      )}
+    >
+      edited
+    </span>
+  );
+}