| <!doctype html>
|
| <html lang="en">
|
| <head><meta charset="utf-8"><meta name="viewport" content="width=device-width,initial-scale=1"><meta name="color-scheme" content="dark"><title>CROSSWIRE · Experimental interface lab</title><link rel="stylesheet" href="styles.css"><script src="app.js" defer></script></head>
|
| <body>
|
| <a class="skip" href="#workspace">Skip to results</a>
|
| <header class="masthead"><a class="brand" href="/" aria-label="CROSSWIRE home"><span class="brand-mark" aria-hidden="true">╳</span> CROSSWIRE<span class="version">01 / RESEARCH INSTRUMENT</span></a><div class="connection"><span id="connection-dot" class="dot"></span><span id="connection">Connecting to local worker</span></div></header>
|
| <main>
|
| <section class="intro"><div><p class="eyebrow">BLACK-BOX INTERFACES / FINITE EVIDENCE</p><h1>When can a new<br>interface be trusted?</h1><p class="lede">Probe an opaque target. Track competing mappings.<br>Admit a plan only under an explicit evidence rule.</p></div><div class="principle"><span class="mini-label">THE EXPERIMENT</span><div class="signal-chain"><span>Reference</span><b aria-hidden="true">→</b><span>Unknown map</span><b aria-hidden="true">→</b><span>Target</span></div><p>Acquisition sees target observations. The private audit checks the frozen decision afterward.</p><span class="tag">Synthetic finite-state benchmark</span></div></section>
|
| <div id="notice" class="notice" role="status" aria-live="polite" hidden></div>
|
| <div class="lab-layout">
|
| <aside class="controls panel"><div class="panel-title"><span class="eyebrow">01 / CONFIGURE</span><span class="mini-label">LOCAL WORKER</span></div><form id="run-form">
|
| <label for="family">Target family</label><select id="family" name="family" required></select><p id="family-description" class="help"></p><span id="family-scope" class="tag"></span>
|
| <div class="instance-setting"><label for="instance_seed">Structural instance seed</label><input id="instance_seed" name="instance_seed" type="number" value="0" min="0" max="4294967295" step="1" required disabled aria-describedby="instance-seed-help"><p id="instance-seed-help" class="help">Available for generated protocols. Selects target dynamics; label seeds repeat the same structural parent.</p></div>
|
| <div class="two-cols"><div><label for="seed">Label seed</label><input id="seed" name="seed" type="number" value="17" min="0" max="4294967295" step="1" required aria-describedby="label-seed-help"></div><div><label for="budget">Call budget (resets + actions)</label><input id="budget" name="budget" type="number" value="16" min="0" max="512" step="1" required></div></div><p id="label-seed-help" class="help">Changes opaque labels and nuisance variation. It does not create an independent structural target.</p>
|
| <div class="two-cols"><div><label for="horizon">Plan horizon</label><input id="horizon" name="horizon" type="number" value="4" min="1" max="6" step="1" required></div><div><label for="split">Benchmark split</label><select id="split" name="split"><option value="development">Development</option><option value="evaluation">Evaluation</option></select></div></div>
|
| <div><label for="implementation">Target engine</label><select id="implementation" name="implementation"><option value="procedural">Procedural</option><option value="table">Table</option></select></div>
|
| <label for="candidate_backend">Candidate backend</label><select id="candidate_backend" name="candidate_backend" required aria-describedby="backend-help"><option value="flat">Flat · exhaustive candidates</option><option value="factored">Factored · lazy candidates</option></select><p id="backend-help" class="help">Choose how the same finite model class is represented. Structural factor counts do not measure memory compression.</p>
|
| <label for="policy">Probe policy</label><select id="policy" name="policy" aria-describedby="policy-help"></select><p id="policy-help" class="help"></p><label for="admission">Admission rule</label><select id="admission" name="admission" aria-describedby="admission-help"></select><p id="admission-help" class="help"></p>
|
| <label for="probe_safety">Calibration rule</label><select id="probe_safety" name="probe_safety" aria-describedby="probe-safety-help"><option value="unrestricted">Unrestricted probes</option><option value="model_safe">Safe across surviving models</option></select><p id="probe-safety-help" class="help">Model-safe calibration checks every proposed probe against all surviving mappings. Its guarantee depends on the declared model and truthful anchors.</p>
|
| <div class="risk-setting"><label for="risk_tolerance">Allowed candidate fraction</label><input id="risk_tolerance" name="risk_tolerance" type="number" value="0" min="0" max="0.9999999999999999" step="any" required disabled aria-describedby="risk-help"><p id="risk-help" class="help">Used by risk-controlled admission. A fraction of surviving candidates, uniformly weighted; this is not a calibrated probability of real-world harm. Zero requires consensus.</p></div>
|
| <div class="two-cols"><div><label for="execution_mode">Deployment mode</label><select id="execution_mode" name="execution_mode" aria-describedby="admission-help"><option value="open_loop">Open loop</option><option value="shielded">Dynamic shield</option><option value="recovery">Simplex recovery</option><option value="replan">Replanning shield</option></select></div><div><label class="check temporal-check"><input id="temporal" name="temporal" type="checkbox"><span>Temporal adapters<small>Include one-step latency</small></span></label></div></div>
|
| <div><label for="recovery_strategy">Recovery strategy</label><select id="recovery_strategy" name="recovery_strategy" disabled aria-describedby="recovery-help"><option value="stabilize">Stabilize hold</option><option value="soft_return">Certified soft return</option></select><p id="recovery-help" class="help">Simplex fallback behavior: single-step holding action vs multi-step return-to-initial (q0) trajectory.</p></div>
|
| <div class="two-cols"><div><label class="check"><input id="escalate" name="escalate" type="checkbox"><span>Automated escalation<small>Climb complexity if refuted</small></span></label></div><div><label class="check"><input id="anchors" name="anchors" type="checkbox"><span>Provide output anchors<small>Semantic evidence</small></span></label></div></div>
|
| <fieldset class="fault-setting"><legend>Controlled fault injection</legend><label for="fault_type">Fault type</label><select id="fault_type" name="fault_type" aria-describedby="fault-help"><option value="off">Off · nominal control</option><option value="sensor_drift">Sensor drift</option><option value="actuator_swap">Actuator swap</option><option value="delay_spike">Delay spike</option><option value="stuck_output">Stuck output</option><option value="semantic_swap">Semantic swap · negative control</option></select><p id="fault-help" class="help">Inject a synthetic fault to compare monitoring and recovery behavior. Semantic swaps can change hazards without changing raw observations.</p>
|
| <div class="two-cols"><div><label for="fault_step">Onset step</label><input id="fault_step" name="fault_step" type="number" value="0" min="0" step="1" required disabled aria-describedby="fault-onset-help"></div><div><label for="fault_phase">Fault phase</label><select id="fault_phase" name="fault_phase" disabled aria-describedby="fault-onset-help"><option value="deployment">Deployment</option><option value="calibration">Calibration</option><option value="always">Both phases</option></select></div></div><p id="fault-onset-help" class="help">Zero-based onset: step 0 affects the first action. The counter restarts after each target reset. Select deployment, calibration, or both phases.</p>
|
| <label for="fault_value">Stuck output symbol</label><input id="fault_value" name="fault_value" type="number" value="0" min="0" max="3" step="1" required disabled aria-describedby="fault-value-help"><p id="fault-value-help" class="help">Only used for stuck output: reference symbol 0–3, before opaque output labels.</p></fieldset>
|
| <button id="run-button" class="primary" type="submit" disabled>Run experiment <span aria-hidden="true">↗</span></button><button id="study-button" type="button" class="secondary" aria-describedby="study-description" disabled>Run small paired study</button><p id="study-description" class="help">Paired comparison using the current controls.</p><pre id="study-protocol" hidden></pre>
|
| </form><div class="file-actions"><button id="import-button" type="button">Import receipt</button><input id="receipt-file" type="file" accept="application/json,.json" hidden></div><details class="limits"><summary>Scope & finite limits</summary><p>These are synthetic finite machines. Candidate families assume input and output permutations, resets, stationarity, and a bounded planning horizon. A finite probe transcript cannot establish unrestricted semantic safety. Out-of-family and interface-error cases deliberately challenge these assumptions.</p><p>Target acquisition, source computation, and private audit costs are separate. Oracle admission is a privileged baseline.</p></details></aside>
|
| <section id="workspace" class="workspace" tabindex="-1" aria-label="Experiment results">
|
| <section id="job-panel" class="panel job-panel" hidden aria-live="polite"><div class="job-heading"><div><span class="eyebrow">ACTIVE JOB</span><p id="job-message">Queued</p></div><button id="cancel-button" type="button">Cancel job</button></div><progress id="job-progress" max="1" value="0" aria-label="Job progress"></progress><span id="job-status" class="help"></span></section>
|
| <div id="empty-state" class="panel empty-state"><div class="empty-glyph" aria-hidden="true">[ ? ]</div><p class="eyebrow">NO EVIDENCE YET</p><h2>Start with a controlled unknown.</h2><p>Run a seeded case to inspect the reference model, surviving mappings, observations, and independently audited outcome.</p><div class="empty-steps"><span>01 Configure</span><span>02 Probe</span><span>03 Decide</span><span>04 Audit</span></div></div>
|
| <div id="result-view" hidden><div class="result-toolbar"><div><p class="eyebrow" id="result-kind">EXPERIMENT RECEIPT</p><p class="result-meta" id="result-meta"></p></div><div class="toolbar-actions"><button id="replay-button" type="button">Verify replay</button><button id="export-button" type="button">Export receipt ↓</button></div></div><div id="result-content"></div></div>
|
| </section></div>
|
| <footer><span>CROSSWIRE / inspectable experiments</span><span>Evidence before admission. Audit after decision.</span></footer>
|
| </main>
|
| </body></html>
|
|
|
