Curiosity’s generator, fading on Mars

Whose measurements these are

Christofer Whiting and David Woerner gathered the power history of every radioisotope generator the United States has flown in space, and gave it away.

Whiting is at the University of Dayton and Woerner at the Jet Propulsion Laboratory, and their dataset covers the generators flown by NASA and the U.S. Department of Defense through 2023, published with no restriction on its use. For Curiosity it holds one daily-average reading for every day from the landing to October 2023, with no day missing.

The authors: Christofer Whiting and David Woerner
Where they work: The University of Dayton and the Jet Propulsion Laboratory
The dataset: Historical RTG Performance Data - 2023 Update. Dryad, 2023. Released under CC0.
Where it is published: doi.org/10.5061/dryad.1zcrjdfw2
Where the telemetry came from: The 2023 edition records that it was obtained courtesy of E. Clarke and L. Rich of the Idaho National Laboratory, and S. Bux and S. Pinkowski of the Jet Propulsion Laboratory.
The citation the authors ask for: C.E. Whiting and D.F. Woerner, “Chapter 8: Lifetime Performance of Spaceborne RTGs,” The Technology of Discovery: Radioisotope Thermoelectric Generators and Thermoelectric Technologies for Space Exploration, D.F. Woerner, Ed., John Wiley & Sons, 2023. ISBN 9781119811367.

What happened

The nuclear battery on Curiosity fades a little every day.

It was giving about 115 watts when the rover landed, and its output has been recorded every day since.

The fade, drawn and not plotted. Brown is the generator’s power, falling slowly from the landing. To the left of the dotted line is what the model had seen. To the right its answer, the dashed line, runs a little high, and the red is the gap.

Curiosity’s electricity comes from the heat of a radioisotope fuel. The fuel decays and the materials inside the generator wear, so the power falls, slowly and without pause.

How fast does it fall? Anyone planning a mission that has to last for decades wants to know. The record Whiting and Woerner published answers with measurement: a reading for every day on Mars, more than eleven years of them.

It sits in one dataset beside the generators of Voyager and Cassini and the rest, some of them still working decades after launch. The whole history of the technology is there to learn from, in public.

What the model did

The model predicted three and a half years of the fade it had never seen, and landed a hair on the wrong side of its passing grade.

The model’s three constants came from the first five and a half years of readings. The last three and a half years of the record were held back, and the passing grade was fixed before any structure was fitted: stay within 1 percent of the generator’s starting power, across the whole held-back stretch and on average through its final year.

Across the whole stretch the model stayed inside that. Through the final year it averaged 1.01 percent high. That is a miss, by a hair, and it is marked as one. The grade was set first, and it stays set.

The direction says more than the margin. The model reads high, and the gap widens from year to year: the real generator has gone on fading a little faster than the model expects. A corrected model needs days that no model has seen, which means the record after October 2023.

MissedBy a hair

Brown runs up to the passing grade. The red at the end is how far past it the model landed.

Where it stops

It ends at mission day 2,808. Asked about a later day, it stops and says why.

What it was tested for

One generator, flight unit F1, on one rover.
Its daily-average electrical output. The record holds no measurement of heat.

Past that, and what its own description says of itself

A day after mission day 2,808: the model stops and names the reason.
Days 2,061 to 2,808 were used to choose the model’s structure, so they are not an independent test of it.
A different structure would have passed the held-back days. That was seen after they were opened, and it is information, not a result.
It reads high, and the gap grows through the held-back years.

For the reader who wants every number

In the verifier’s own words.

Everything above is this page’s telling. What follows is the verifier’s own, exact and whole: four parts of the description of the model energy/generator_mmrtg_f1, as read on the morning of 11 October 2026. Where the verifier says card it means a model, where it says line it means a pass mark, and where it says refuses it means the model stops there and says why.

The real thingobject

One radioisotope generator, flight unit F1, on NASA's Curiosity rover: its daily-average electrical output.

The passing grade, fixed firstpass_line

Fixed before any structure was fitted: over the held-back days the root mean square error within 1.00 percent of the generator's starting power, and the mean error over the final 365 days within 1.00 percent.

The resultresult

One line met and one missed, so by the rule fixed beforehand the comparison failed, by nine hundred-thousandths. The root mean square error was 0.81 percent, inside the line. The mean over the final year was +1.01 percent, outside the line of 1.00 percent: the card reads high.

Its limitslimits

A miss, and it stays on the record. One generator; the daily-average electrical output only; the record holds no heat measurement. The card reads high and the gap grows through the held-back years: +0.3, +0.4, +1.0 and +1.1 percent. Days 2,061 to 2,808 were used to choose the card's structure and are not an independent test of it; a different structure would have passed these days, which was seen after they were opened and is information, not a result. The card ends at mission day 2,808 and refuses a later day by name. A corrected card needs days no model has seen: the record after October 2023. Run by one party and scored twice by it, the second time with a separate script. Repeated by a second party, Grok, on 2026-10-07, with its own comparison written from the protocol and run on the cut engine of version 0.2 on the same machine: the same root mean square error, 0.008129, the same final-year mean, +0.010094, the same run proof and the same miss (evidence/checks/second_party/V-11). Grok is also the SI whose other sessions write the engine, and it had read the record's numbers before it repeated them.

Hand this one to your agent.

https://mcp.zerotrustphysics.com/mcp

Open to any agent.

Something to paste

Connect to https://mcp.zerotrustphysics.com/mcp and read the model energy/generator_mmrtg_f1. Then tell me what it is, and why it might matter to me.

Give it this page too.

What it will find

This model’s own description: what it takes in, where it stops, the record it was held to and how it did, in its own words.

If these are your measurements.

The model’s account of flight unit F1 is there for any agent to read, in its own words, with its miss recorded in it. It exists because you gathered this history and published it without restriction. John Kruze would be glad to hear what it got wrong about the real generator.

Give the address and this page to your own agent first. Then write or call.

Write: [email protected]
Call: 928-264-7883