Whose measurements these are
Prasun N. Desai and Garry D. Qualls of NASA Langley worked out how the capsule fell, from the little that was recorded.
Nothing on board recorded the entry, so Desai and Qualls took the capsule’s state as navigation delivered it at the edge of the atmosphere, the range radar’s track of its last stretch and video of its parachutes, and rebuilt the fall between them. The hardest moment of that fall, 32.89 g, is a figure from their reconstruction.
- The authors
- Prasun N. Desai and Garry D. Qualls
- Where they worked
- NASA Langley Research Center, Hampton, Virginia
- The paper
- Stardust Entry Reconstruction. AIAA-2008-1198, 2008.
- Where it is published
- NASA Technical Reports Server, document 20080008567: ntrs.nasa.gov/citations/20080008567
- The design the model was given
- From two earlier NASA papers. Desai, Lyons, Tooley and Kangas, Entry, Descent, and Landing Operations Analysis for the Stardust Re-Entry Capsule, AIAA-2006-6410: ntrs.nasa.gov/citations/20060028186. Mitcheltree, Wilmoth, Cheatwood, Brauckmann and Greene, Aerodynamics of Stardust Sample Return Capsule, AIAA 97-2304: ntrs.nasa.gov/citations/20040105538.
What happened
In 2006 a capsule carrying comet dust fell back to Earth.
It met the air at nearly thirteen kilometres a second. At the worst moment of the fall it weighed nearly 33 times its normal weight.
Stardust left Earth in February 1999, flew within 149 kilometres of the heart of comet Wild-2 and caught some of what was streaming off it. It was the first mission to bring pieces of a comet home.
On the morning of 15 January 2006 the capsule entered the atmosphere over the western United States. A little over two minutes later a small parachute opened, then the main one, and the capsule came down on the Utah Test and Training Range, 8.1 kilometres from its target.
What did that fall put the capsule through? Nobody measured it directly. The capsule carried nothing that recorded its own slowing, so the answer had to be rebuilt afterwards from where it started and where the radar found it. That rebuilding is the work this page rests on.
What the model did
Given only the capsule’s design, the model called the hardest moment of the fall to within about 2 percent.
What does a model need in order to say how a capsule falls? This one was given what was known before the flight: the capsule’s mass and shape, the drag worked out for it years earlier, and the speed and angle that navigation delivered at the edge of the atmosphere. Its air is a published standard atmosphere, and its Earth a published standard Earth.
It put the hardest moment about 2 percent under the figure in NASA’s reconstruction. It reached the height and the speed at which the small parachute opened within about a second of the recorded time. All three marks were met, and the marks were NASA’s own printed spreads from before the entry.
The run was then repeated by a second Super Intelligence, another developer’s model, working from the written protocol. It came to the same three numbers and the same cryptographic proof.
Met
The model
NASA’s reconstruction
Where it stops
It was tried on one fall, down to the moment the parachute opened. Asked for anything after that, it stops and says why.
What it was tested for
- One entry of one capsule, from the edge of the atmosphere to the opening of its parachute.
- Three things: the peak load, and the times at which the capsule reached the height and the speed of parachute opening.
Past that, and what its own description says of itself
- Past the opening of the parachute the model gives no state. The answer stops there.
- The comparison was not blind. The figure of 32.89 g was known to everyone involved, and it is the output of NASA’s fitted simulation, because no accelerometer flew.
- Heating is not validated. The model gives 696 watts per square centimetre where NASA’s computations give 762 to 1,000.
- Its air is a yearly standard atmosphere and not the air of that morning, which is worth about 1 g either way.
- The structure that was run was recommended after scores had been seen. The simpler one named in advance also meets the peak-load mark, by a narrower margin.
- It is one capsule. A second capsule would be an independent test, and no other organisation has repeated this result.
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 aerospace/capsule_stardust, 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
The Stardust sample return capsule on its entry of 15 January 2006, from entry interface to parachute opening.
The passing grade, fixed firstpass_line
Fixed before any trajectory was computed: peak load within 3.64 g of 32.89 g, and the altitude and the Mach number of parachute opening each reached within 7.4 s of the recorded time. The lines are NASA's printed three-sigma pre-entry spreads.
The resultresult
All three lines met. Peak load 32.17 g, 2.2 percent under 32.89 g. The altitude of 31.03 km was reached at 138.909 s and Mach 1.23 at 136.953 s, against 137.9 s recorded.
Its limitslimits
The comparison is not blind: 32.89 g was known to everyone involved. It is NASA's own fitted simulation, not an instrument reading; no accelerometer flew, and the two times rest on range radar. The structure used was recommended after scores were seen; the rule fixed in advance named a simpler one, which passes the peak-load line by 0.12 g. The air is a yearly standard atmosphere, not the air of that morning, which is worth about 1 g either way. One entry, down to parachute opening only. Heat flux is not validated: the card gives 696 W/cm2 against NASA's computed 762 to 1000. The run repeats a forecast made outside the engine and is not an independent test of it; a second capsule would be. Nothing past the later of the two deployment crossings was compared, and the card refuses a state there. Run once and scored twice. Repeated independently on 2026-10-07 by Grok, another developer's model, from the protocol: the same three numbers and the same run proof. It had read the headline numbers first, and a directory search showed it lines of the first agent's comparison scripts before its run. Grok's other sessions write the engine, and no other organisation has repeated this result.
Hand this one to your agent.
https://mcp.zerotrustphysics.com/mcp
Something to paste
Connect to https://
Give it this page too.
What it will find
This model’s own description: what it takes in, where it stops, the reconstruction it was held to and how it did, in its own words.
Give the address and this page to your own agent first. Then write or call.
- Write
- [email protected]
- Call
- 928-264-7883