PIG/PIG-Demo
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

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

Browse Source 
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
This commit is contained in:
karti-ai
2026-08-28 15:39:03 -07:00
parent 5a9ff8dda9
commit a56f097f28
54 changed files with 8201 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/components/demo/DemoErrorBoundary.tsx
+90
View File
@@ -0,0 +1,90 @@
import { Component } from 'react';
import type { ErrorInfo, ReactNode } from 'react';
import { AlertTriangle, ExternalLink, RotateCcw } from 'lucide-react';
const REPO_URL = 'https://github.com/karti-ai/PIG-Demo';
export interface DemoErrorBoundaryProps {
children: ReactNode;
/** Named in the fallback copy, so the visitor knows what broke. */
demoTitle?: string;
/** Link to the exact source, if the caller knows it. Falls back to the repo. */
sourceHref?: string;
/** Called when the visitor asks to try again; use it to reset shell state. */
onReset?: () => void;
}
interface DemoErrorBoundaryState {
error: Error | null;
}
/**
* One broken demo must never take the site down.
*
* This is a class because there is still no hook for `componentDidCatch`; that
* is the entire reason for the exception to the function-component rule here.
*
* The fallback is deliberately calm and specific. A site whose pitch is
* "here are the receipts" cannot answer a crash with a shrug: it names the
* demo, links the source, and lets the visitor retry without a full reload.
*/
export class DemoErrorBoundary extends Component<DemoErrorBoundaryProps, DemoErrorBoundaryState> {
override state: DemoErrorBoundaryState = { error: null };
static getDerivedStateFromError(error: Error): DemoErrorBoundaryState {
return { error };
}
override componentDidCatch(error: Error, info: ErrorInfo) {
// No telemetry endpoint on a static site, and none is wanted. The console
// is the only place a maintainer can see this, so keep the component stack.
console.error('[pig-demo] a demo surface threw', error, info.componentStack);
}
private handleReset = () => {
this.setState({ error: null });
this.props.onReset?.();
};
override render() {
const { error } = this.state;
if (!error) return this.props.children;
const { demoTitle, sourceHref } = this.props;
return (
<div role="alert" className="card mx-auto my-10 max-w-xl p-6">
<div className="flex items-center gap-2 text-warning">
<AlertTriangle className="h-5 w-5" aria-hidden="true" />
<h2 className="text-base font-semibold">
{demoTitle ? `${demoTitle} failed to render` : 'This demo failed to render'}
</h2>
</div>
<p className="mt-3 text-sm leading-relaxed text-muted">
Something in this demo threw while drawing. The rest of the site is unaffected — every
other demo is a separate module. The environment and the recorded runs behind this page
are in the repository either way, and you can run them yourself.
</p>
<p className="mt-3 break-words rounded-lg bg-surface-2 px-3 py-2 font-mono text-xs text-muted">
{error.message || 'Unknown error'}
</p>
<div className="mt-5 flex flex-wrap gap-2">
<button
type="button"
onClick={this.handleReset}
className="tap inline-flex items-center gap-2 rounded-lg bg-primary px-4 py-2 text-sm font-medium text-primary-foreground transition-colors duration-2 ease-enter hover:bg-primary/90"
>
<RotateCcw className="h-4 w-4" aria-hidden="true" />
Try again
</button>
<a
href={sourceHref ?? REPO_URL}
className="tap inline-flex items-center gap-2 rounded-lg border border-border px-4 py-2 text-sm font-medium transition-colors duration-2 ease-enter hover:bg-surface-2"
>
Read the source
<ExternalLink className="h-4 w-4" aria-hidden="true" />
</a>
</div>
</div>
);
}
}