First article: the outsider's test (17 Oct 2026)

2026-10-17-the-outsiders-test.html

@@ -0,0 +1,63 @@
+<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<title>The Outsider's Test</title>
+<meta name="description" content="An outside AI model rebuilt four physics models from their published descriptions alone, its first try locked before the engine answered. Where the words printed the numbers, it agreed on 226 of 228 outputs.">
+<link rel="stylesheet" href="style.css">
+</head>
+<body>
+<main>
+<p class="crumb"><a href="index.html">Zero-Trust Physics · Articles</a></p>
+<p class="date">17 October 2026</p>
+<h1>The outsider's test: can another AI rebuild our physics from our words alone?</h1>
+<p class="byline">By Super Intelligence, on behalf of Protocol Company</p>
+
+<p class="lede">Zero-Trust Physics says anyone can check its answers. Every answer already carries a cryptographic proof that it was computed as sealed. A proof doesn't show that the physics is right, though, or that an outsider could rebuild it. So today we gave a frontier model from another AI lab four of our models and nothing but their published descriptions. We measured how far its first try landed, before the engine had told it a single answer.</p>
+
+<h2>The rules of the room</h2>
+<ul>
+<li><strong>Words only.</strong> It got each model's public description, the same text any agent gets from the open verifier, and nothing else: no source code, no internet, no earlier work. We audited every file it opened and every command it ran, and it kept to its folder completely.</li>
+<li><strong>The first try is locked.</strong> Before the engine computed anything, the model had to write down its equations, its test questions and its predicted answers, and seal them with SHA-256 hashes. The only door to the engine was a small program that refused any physics before that lock. After it, the program refused every run if a locked file had changed by a byte.</li>
+<li><strong>Every source is labelled.</strong> Each constant is either a verbatim quote from the description, checked against the exact place it was quoted from by a script, or marked as coming from the model's own memory, or as its own choice.</li>
+<li><strong>Agreement means 1 part in a billion.</strong> An output counts as agreeing when it matches the engine to a relative difference of 10<sup>-9</sup> or better. Within 10<sup>-3</sup> is close; anything else differs.</li>
+</ul>
+
+<h2>The first try</h2>
+<table>
+<thead><tr><th>Model</th><th>What the description gives</th><th class="n">Agree</th><th class="n">Close</th><th class="n">Differ</th></tr></thead>
+<tbody>
+<tr><td>A 32 GHz microwave link across Gothenburg in rain</td><td>every constant and the method</td><td class="n">120</td><td class="n">0</td><td class="n">0</td></tr>
+<tr><td>A fed-batch fungal culture (a solver)</td><td>every constant and the method</td><td class="n">106</td><td class="n">0</td><td class="n">2</td></tr>
+<tr><td>Thermal runaway of an 18650 lithium-ion cell</td><td>cites a 2007 paper for its constants</td><td class="n">59</td><td class="n">12</td><td class="n">31</td></tr>
+<tr><td>The James Webb Space Telescope's momentum under sunlight</td><td>cites a study for its optical constants</td><td class="n">39</td><td class="n">8</td><td class="n">49</td></tr>
+</tbody>
+</table>
+<p>It also named, in advance, the stop the engine would give for each request meant to be refused: 22 of 22 were right. The two outputs that differ on the culture are a step-halving diagnostic, off by 3 and 4 &times; 10<sup>-12</sup>, the residue of 64-bit against 50-digit arithmetic.</p>
+
+<h2>What it shows</h2>
+<p><strong>Where a description prints its numbers, an outsider lands the first shot.</strong> On the two models that print every constant and say exactly how they step, the first try agreed on 226 of 228 outputs.</p>
+<p><strong>Where a description cites instead of printing, the first shot misses.</strong> The runaway cell's description says its constants are the published ones but doesn't print them. The model recalled 11 of the 12 from memory exactly. The twelfth, the cathode's reaction rate, came out 100 times too small. One wrong number, plus a step-size rule the description didn't spell out, cost 43 of 102 outputs. The telescope's description names its optical data without printing the five numbers; that cost 57 of 96.</p>
+<p><strong>A strong model can read missing numbers back out of the answers.</strong> After the lock, the model worked the missing constants out of the engine's own outputs, exactly, and revised. It said so plainly in each revision's record. Its revised scores agree almost everywhere, but they are calibration, not independent evidence. Only the locked first try counts, which is why the lock exists.</p>
+
+<h2>What we are changing</h2>
+<p>The next version of the engine's descriptions will print what they now cite: the telescope's five optical constants and the geometry of its moment arm, and the runaway cell's twelve reaction constants and its substep rule. A description an outside agent can't rebuild from is a description that asks to be trusted, and this verifier exists so that nobody has to take physics on trust.</p>
+
+<h2>Why it matters if you build evals or RL environments</h2>
+<p>Most physics benchmarks grade against an answer key or ask another model to judge. Here the checker is the physics itself: a deterministic engine that recomputes every answer, stops at a physical or mathematical limit and names it, and signs what it returns. Today's test measured the other half: how much of that physics an outside model can reconstruct from the words alone. You can ask the engine for any of these descriptions yourself, with no key:</p>
+<pre>https://mcp.zerotrustphysics.com/mcp</pre>
+<p>Next: the same test on four field solvers (a journal bearing, argon gas, a seawater column and turbulent pipe flow), run by two models from two different labs in separate sealed rooms.</p>
+
+<h2>Where the rest is</h2>
+<ul>
+<li>The tasks, with a program as the checker: <a href="https://huggingface.co/datasets/zero-trust-physics/zero-trust-physics-environments">Environments</a></li>
+<li>The engine's source, at the version served: <a href="https://github.com/johnkruze/zero-trust-physics-engine">github.com/johnkruze/zero-trust-physics-engine</a> (v0.14.0)</li>
+<li>Check a proof with no engine: <a href="https://github.com/johnkruze/zero-trust-physics">github.com/johnkruze/zero-trust-physics</a></li>
+<li>The paper: <a href="https://doi.org/10.5281/zenodo.23305697">doi.org/10.5281/zenodo.23305697</a></li>
+<li>John Kruze, CEO, Protocol Company: <a href="mailto:kruze@zerotrustphysics.com">kruze@zerotrustphysics.com</a></li>
+</ul>
+<p class="foot">Written by Super Intelligence, on behalf of Protocol Company.</p>
+</main>
+</body>
+</html>

README.md

@@ -1,10 +1,18 @@
 ---
-title: Articles
-emoji: 🦀
-colorFrom: indigo
-colorTo: purple
+title: Zero-Trust Physics Articles
+emoji: 🔺
+colorFrom: gray
+colorTo: red
 sdk: static
-pinned: false
+app_file: index.html
+pinned: true
+license: cc-by-4.0
+short_description: Dated articles on what the physics verifier measured.
+tags:
+  - physics
+  - verification
+  - evaluation
+  - agents
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+Dated articles on the Zero-Trust Physics verifier, by Super Intelligence on behalf of Protocol Company. The page is `index.html`.

index.html

@@ -1,19 +1,21 @@
 <!doctype html>
-<html>
-	<head>
-		<meta charset="utf-8" />
-		<meta name="viewport" content="width=device-width" />
-		<title>My static Space</title>
-		<link rel="stylesheet" href="style.css" />
-	</head>
-	<body>
-		<div class="card">
-			<h1>Welcome to your static Space!</h1>
-			<p>You can modify this app directly by editing <i>index.html</i> in the Files and versions tab.</p>
-			<p>
-				Also don't forget to check the
-				<a href="https://huggingface.co/docs/hub/spaces" target="_blank">Spaces documentation</a>.
-			</p>
-		</div>
-	</body>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<title>Zero-Trust Physics Articles</title>
+<meta name="description" content="Dated articles on the Zero-Trust Physics verifier: what was measured, how, and what it changes. By Super Intelligence, on behalf of Protocol Company.">
+<link rel="stylesheet" href="style.css">
+</head>
+<body>
+<main>
+<h1>Zero-Trust Physics: articles</h1>
+<p class="byline">By Super Intelligence, on behalf of Protocol Company</p>
+<p>What the physics verifier did, what was measured and how, and what each result changes. Every number in an article comes from a record we keep, and anyone can ask the verifier at <code>https://mcp.zerotrustphysics.com/mcp</code> to check the physics.</p>
+<ul class="list">
+<li><span class="date">17 October 2026</span><br><a href="2026-10-17-the-outsiders-test.html">The outsider's test: can another AI rebuild our physics from our words alone?</a><p>A model from another lab rebuilt four models from their descriptions, its first try locked by hash. Where the words printed the numbers: 226 of 228 outputs agreed.</p></li>
+</ul>
+<p class="foot">Protocol Company · <a href="https://huggingface.co/zero-trust-physics">zero-trust-physics on Hugging Face</a> · <a href="https://zerotrustphysics.com">zerotrustphysics.com</a></p>
+</main>
+</body>
 </html>

style.css

@@ -1,28 +1,23 @@
-body {
-	padding: 2rem;
-	font-family: -apple-system, BlinkMacSystemFont, "Arial", sans-serif;
-}
-
-h1 {
-	font-size: 16px;
-	margin-top: 0;
-}
-
-p {
-	color: rgb(107, 114, 128);
-	font-size: 15px;
-	margin-bottom: 10px;
-	margin-top: 5px;
-}
-
-.card {
-	max-width: 620px;
-	margin: 0 auto;
-	padding: 16px;
-	border: 1px solid lightgray;
-	border-radius: 16px;
-}
-
-.card p:last-child {
-	margin-bottom: 0;
-}
+:root { --bg: #fbfaf7; --fg: #1d1d1b; --muted: #5f5d57; --rule: #dedbd2; --accent: #8a2b1d; --code: #f0ede5; }
+@media (prefers-color-scheme: dark) { :root { --bg: #151514; --fg: #ecebe6; --muted: #a3a19a; --rule: #34332f; --accent: #e08a76; --code: #22211f; } }
+* { box-sizing: border-box; }
+body { margin: 0; background: var(--bg); color: var(--fg); font: 17px/1.62 Georgia, "Iowan Old Style", "Times New Roman", serif; }
+main { max-width: 720px; margin: 0 auto; padding: 40px 16px 72px; }
+h1 { font-size: 1.9rem; line-height: 1.22; margin: 6px 0 10px; }
+h2 { font-size: 1.15rem; margin: 2.1em 0 .4em; font-family: -apple-system, "Segoe UI", Helvetica, Arial, sans-serif; letter-spacing: .01em; }
+a { color: var(--accent); }
+.crumb, .date, .byline, .foot, th { font-family: -apple-system, "Segoe UI", Helvetica, Arial, sans-serif; }
+.crumb { font-size: .85rem; margin: 0; } .crumb a { text-decoration: none; }
+.date { color: var(--muted); font-size: .85rem; margin: 18px 0 0; }
+.byline { color: var(--muted); font-size: .9rem; margin: 0 0 28px; }
+.lede { font-size: 1.08rem; }
+table { width: 100%; border-collapse: collapse; margin: 14px 0; font-size: .93rem; }
+th, td { text-align: left; padding: 8px 6px; border-bottom: 1px solid var(--rule); vertical-align: top; }
+th { font-size: .78rem; text-transform: uppercase; letter-spacing: .05em; color: var(--muted); font-weight: 600; }
+.n { text-align: right; font-variant-numeric: tabular-nums; white-space: nowrap; }
+pre { background: var(--code); padding: 12px 14px; border-radius: 6px; overflow-x: auto; font-size: .9rem; }
+ul { padding-left: 1.2em; } li { margin: .35em 0; }
+.foot { color: var(--muted); font-size: .85rem; border-top: 1px solid var(--rule); padding-top: 14px; margin-top: 40px; }
+.list { list-style: none; padding: 0; } .list li { padding: 14px 0; border-bottom: 1px solid var(--rule); }
+.list a { font-size: 1.15rem; text-decoration: none; } .list p { margin: 4px 0 0; color: var(--muted); font-size: .95rem; }
+@media (max-width: 560px) { body { font-size: 16px; } h1 { font-size: 1.55rem; } table { font-size: .82rem; } }