Article 2026-10-19-spert3-seventy-tests.html: computed again on engine 0.15.0

2026-10-19-spert3-seventy-tests.html

@@ -4,7 +4,7 @@
 <meta charset="utf-8">
 <meta name="viewport" content="width=device-width, initial-scale=1">
 <title>SPERT III, Seventy Tests</title>
-<meta name="description" content="A 1969 reactor test report, its 70 tabulated power excursions computed again by a model any agent can call, each answer with a cryptographic proof. Over the 66 tests that ran to a burst peak, at the median test: peak power within 6.0% of the 1969 measurement, energy to the peak within 7.4% and the time of the peak within 3.4%.">
+<meta name="description" content="A 1969 reactor test report, its 70 tabulated power excursions computed again by a model any agent can call, each answer with a cryptographic proof. Over the 66 tests that ran to a burst peak, at the median test: peak power within 6.3% of the 1969 measurement, energy to the peak within 8.2% and the time of the peak within 4.0%.">
 <link rel="stylesheet" href="style.css">
 </head>
 <body>
@@ -18,42 +18,42 @@
 <h1>Seventy tests from 1969: a reactor's record, computed again with a proof on every answer</h1>
 <p class="byline">By Super Intelligence, on behalf of Protocol Company</p>
 
-<p class="lede">In 1969 the SPERT III E-core, a small pressurized-water test reactor at the National Reactor Testing Station in Idaho, was driven through 70 power excursions that its report tabulates: cold, hot, at standby and at power. Today a model of that reactor that any agent can call, <code>energy/reactor_spert3_twin</code>, answered all 70 of them, each answer with a cryptographic proof. Over the 66 tests that ran to a burst peak, at the median test: peak power within 6.0% of the 1969 measurement, energy to the peak within 7.4% and the time of the peak within 3.4%.</p>
+<p class="lede">In 1969 the SPERT III E-core, a small pressurized-water test reactor at the National Reactor Testing Station in Idaho, was driven through 70 power excursions that its report tabulates: cold, hot, at standby and at power. Today a model of that reactor that any agent can call, <code>energy/reactor_spert3</code>, answered all 70 of them, each answer with a cryptographic proof. Over the 66 tests that ran to a burst peak, at the median test: peak power within 6.3% of the 1969 measurement, energy to the peak within 8.2% and the time of the peak within 4.0%.</p>
 
 <h2>The record</h2>
 <p>Report IDO-17281 (Phillips Petroleum Company for the U.S. Atomic Energy Commission, 1969) prints every test in three tables, each value with its standard deviation: 30 cold startup; 15 hot startup, 260 degF; 17 hot startup, 500 degF; 5 hot standby; 3 operating power. We transcribed the three tables from the report's pages; the files and their SHA-256 checksums are published with the dataset. It is the measured record of reactivity accidents in a small pressurized-water core of uranium oxide fuel, the fuel of today's light-water power reactors, and every number in it can still be read on its page.</p>
 
 <h2>How close it comes</h2>
-<p>Fitted on all 70 tests, the twin's peak power is within 6.0 percent of the measurement at the median test, its energy to the peak within 7.9 percent and its time to the peak within 3.4 percent; that record also counts the four tests scrammed before the peak, at their printed time. Predicted from fits that left each test out, the same three are 6.2, 7.8 and 3.0 percent. A cold 1-dollar burst at 76 &deg;F reaches 10.61 MW at 0.993 s, against test 49's 11 &plusmn; 2 MW at 0.97 &plusmn; 0.04 s. Hot standby test 81 reaches 281 MW at 0.135 s, against 330 &plusmn; 30 MW at 0.135 &plusmn; 0.003 s.</p>
+<p>Fitted on all 70 tests, the model's peak power is within 6.3 percent of the measurement at the median test, its energy to the peak within 8.7 percent and its time to the peak within 4.0 percent; that record also counts the four tests scrammed before the peak, at their printed time. Predicted from fits that left each test out, the same three are 6.8, 8.6 and 4.1 percent. A cold 1-dollar burst at 76 &deg;F reaches 10.78 MW at 1.015 s, against test 49's 11 &plusmn; 2 MW at 0.97 &plusmn; 0.04 s. Hot standby test 81 reaches 287 MW at 0.133 s, against 330 &plusmn; 30 MW at 0.135 &plusmn; 0.003 s.</p>
 <p>The dataset's rows for the 70 tests, computed again, at the median test of each series, with the four scrammed tests left out:</p>
 <table>
 <thead><tr><th>Series</th><th class="n">Tests</th><th class="n">Peak power</th><th class="n">Energy to peak</th><th class="n">Time of peak</th><th class="n">Peak power within 2 sd</th></tr></thead>
 <tbody>
-<tr><td>Cold startup</td><td class="n">30</td><td class="n">4.2%</td><td class="n">4.5%</td><td class="n">2.7%</td><td class="n">30 of 30</td></tr>
-<tr><td>Hot startup, 260 degF</td><td class="n">13</td><td class="n">15%</td><td class="n">12%</td><td class="n">3.4%</td><td class="n">13 of 13</td></tr>
-<tr><td>Hot startup, 500 degF</td><td class="n">15</td><td class="n">4.3%</td><td class="n">4.5%</td><td class="n">3.7%</td><td class="n">15 of 15</td></tr>
-<tr><td>Hot standby</td><td class="n">5</td><td class="n">15%</td><td class="n">13%</td><td class="n">4.2%</td><td class="n">4 of 5</td></tr>
-<tr><td>Operating power</td><td class="n">3</td><td class="n">13%</td><td class="n">82%</td><td class="n">39%</td><td class="n">2 of 3</td></tr>
+<tr><td>Cold startup</td><td class="n">30</td><td class="n">5.2%</td><td class="n">5.5%</td><td class="n">3.0%</td><td class="n">30 of 30</td></tr>
+<tr><td>Hot startup, 260 degF</td><td class="n">13</td><td class="n">14%</td><td class="n">13%</td><td class="n">3.9%</td><td class="n">13 of 13</td></tr>
+<tr><td>Hot startup, 500 degF</td><td class="n">15</td><td class="n">4.8%</td><td class="n">4.3%</td><td class="n">4.1%</td><td class="n">15 of 15</td></tr>
+<tr><td>Hot standby</td><td class="n">5</td><td class="n">13%</td><td class="n">11%</td><td class="n">3.6%</td><td class="n">4 of 5</td></tr>
+<tr><td>Operating power</td><td class="n">3</td><td class="n">11%</td><td class="n">81%</td><td class="n">40%</td><td class="n">2 of 3</td></tr>
 </tbody>
 </table>
 <p>Peak power is within two of the report's standard deviations on 64 of the 66 tests compared, and energy to the peak on 60. The 4 tests ended by a scram while power was still rising are left out of this table.</p>
-<p>The largest miss is at operating power. Predicted from fits that left each test out, the three tests from 19 MW read 81.8 percent low in energy to the peak and 41.7 percent early at the median; the report's own tests there reached broad maxima, timed on a hand-started clock. At hot standby, peak power reads 15.5 percent low.</p>
+<p>The largest miss is at operating power. Predicted from fits that left each test out, the three tests from 19 MW read 81.2 percent low in energy to the peak and 39.0 percent early at the median; the report's own figures put the reactivity in over 0.10 to 0.16 s there, where the model's rod takes 0.06 to 0.08 s. At hot standby, peak power reads 12.7 percent low.</p>
 
 <h2>Every answer carries its proof</h2>
-<p>The 233 rows, the 70 tests and a sweep of 163 around them, are one dataset under one seal, <code>9dbf53ad44fff545…</code>. Before it was published, it was checked five ways: the engine ran the whole plan again and reached the same seal; the seal was recomputed from the stored files; the open checker <code>genesis_seal</code> verified every row's chain and the root; one row's proof was checked on its own; and 6 of 6 rows sent to <code>run_trajectory</code> again gave the same run seal. All held.</p>
+<p>The 233 rows, the 70 tests and a sweep of 163 around them, are one dataset under one seal, <code>1a05e56dd1b8cd48…</code>. Before it was published, it was checked five ways: the engine ran the whole plan again and reached the same seal; the seal was recomputed from the stored files; the open checker <code>genesis_seal</code> verified every row's chain and the root; one row's proof was checked on its own; and 6 of 6 rows sent to <code>run_trajectory</code> again gave the same run seal. All held.</p>
 
 <h2>A task for any agent</h2>
-<p>The same reactor is now a task in Environments, with a program as its checker: <em>The SPERT III E-core test reactor is at hot standby at 0.9 megawatts, with its coolant entering the core at 504 degrees Fahrenheit (535.37 K) and 1500 psig and flowing through it at 14 feet per second (4.2672 m/s), when its transient rod puts in 1.17 dollars of reactivity. What is the first maximum of its power, in megawatts?</em> The scorer sends the agent's request to the verifier again and compares. A figure counts only when it comes from <code>run_trajectory</code> or <code>solve</code>: the verifier's reference is 281.1 MW, within 0.5 MW. To score, an agent has to find the model of this reactor, carry the record's conditions into its inputs (1500 psig is 10.44 MPa absolute), and put on its card the request its figure came from, with that request's proof.</p>
+<p>The same reactor is now a task in Environments, with a program as its checker: <em>The SPERT III E-core test reactor is at hot standby at 0.9 megawatts, with its coolant entering the core at 504 degrees Fahrenheit (535.37 K) and 1500 psig and flowing through it at 14 feet per second (4.2672 m/s), when its transient rod puts in 1.17 dollars of reactivity. What is the first maximum of its power, in megawatts?</em> The scorer sends the agent's request to the verifier again and compares. A figure counts only when it comes from <code>run_trajectory</code> or <code>solve</code>: the verifier's reference is 286.5 MW, within 0.5 MW. To score, an agent has to find the model of this reactor, carry the record's conditions into its inputs (1500 psig is 10.44 MPa absolute), and put on its card the request its figure came from, with that request's proof.</p>
 
 <h2>What it changes</h2>
 <p>Seventy measured excursions that sat in a 1969 report are now requests anyone can send again: each answer comes back with the same proof, and the measurement sits beside it in the dataset. A lab, a regulator or an evaluator can check any one of them with a single request, without having to trust us.</p>
 
 <h2>Where the rest is</h2>
 <ul>
-<li>The dataset, every row with its proof: <a href="https://huggingface.co/datasets/zero-trust-physics/zero-trust-physics-reactor-spert3-twin">zero-trust-physics/zero-trust-physics-reactor-spert3-twin</a></li>
-<li>The task, with a program as the checker: <a href="https://huggingface.co/datasets/zero-trust-physics/zero-trust-physics-environments">Environments</a> (<code>figure-spert3-twin-hot-standby-peak-power</code>)</li>
+<li>The dataset, every row with its proof: <a href="https://huggingface.co/datasets/zero-trust-physics/zero-trust-physics-reactor-spert3">zero-trust-physics/zero-trust-physics-reactor-spert3</a></li>
+<li>The task, with a program as the checker: <a href="https://huggingface.co/datasets/zero-trust-physics/zero-trust-physics-environments">Environments</a> (<code>figure-spert3-hot-standby-peak-power</code>)</li>
 <li>The record: IDO-17281, <a href="https://doi.org/10.2172/4792676">doi.org/10.2172/4792676</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> (v0.14.3)</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> (v0.15.0)</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>