I have submitted the prompt below to a number of AIs. I will post links to the responses for Grok and ChatGPT and then also have their probability estimates that this is the dominant theory in 20 years. I will do this again some years later when there is better data and better AIs and see what the new predictions are.
The prompt:
The planetary scientists are confident that there are CO2 cornice avalanches on Mars but
so far don't seem to accept the hypothesis of H2O cornice avalanches on Mars.
This is not liquid water but solid H2O so I think it counts as "dry granular" avalanche.
I think the H₂O cornice theory elegantly explains:
1) Hydrated salts explain very visible RSL trail that then fades (can hydrate during sublimation)
2) Hydrated salt signatures detected from orbit at RSL sites
3) Seasonal timing (cornices collect frost during winter and fail/avalanche during warmer summer)
4) Recurrent nature (annual frost accumulation to cornices can go on for millions of years)
5) Fade timelines match (hydrated salts persist weeks-months then fade)
6) Many RSL sites do show bright patches above them (future cameras with higher resolution could show more)
7) Many RSL and occur below areas suitable for wind-driven frost accumulation.
8) RSL trails match "dry granular flows" and cornice avalanches qualify (some RSL trails cross with liquid won't do)
9) No accumulation of material at the bottom of RSLs (H2O gone by end of trail)
There is spectral evidence of H2O hydrated salts at some of the "Warm Seasonal Flows" or "Recurring Slope Lineae (RSL)" sites.
Vince Cate has shown that ice/snowball passing over salt/iron-sulfate can in a vacuum can make hydrated salts
(https://www.youtube.com/watch?v=LDoOZyHfBvY). So a collection of ice/snow/frost sliding down a cliff
from a H2O cornice avalanche on Mars could make hydrates that take a long time to go away.
There are pictures from the surface of H2O frost near a lander or rover, so we know there is some
amount of frost on Mars. Wind could collect small amounts of frost from a large area into a cornice.
In the warm season a cornice could come lose and make an avalanche. The frozen H2O avalanche could make some
temporary hydrated salts leaving a visible trail. Seems like H2O cornice avalanches on Mars fit the
evidence, why is this not the preferred theory?
In some pictures it seems there are light or white areas above the RSLs that I think could be H2O cornices, like
https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh6-nhvoSmKpBqEIKAhYg3563GjLB0LmFwV_0Yn9No_3CiLRNjnaguCviu56YriuKfBFOFomCzU1aPHnwmEdoMlvNVlF5HHjRComz4mWcsAcUg3plTnN2N77K_W19zwqW9Z_h5o9arQilo/s1600/6+cornices+on+mars.jpg
It seems detectable white/light areas are found above around 1/3rd of the RSL sites, and with future higher resolution cameras perhaps all sites.
How many kg of H2O frost would it take to darken/hydrate the surface of soil the size of typical RSL trails?
Over a "winter" could wind on Mars collect enough frost to get this many Kg of H2O together into a cornice?
Could this cornice last through the winter and then collapse in the summer?
If the location of the cornice was partly shaded by rock outcrops would that make the collection and lasting time easier?
Should cameras currently orbiting Mars see a cornice of this many Kg?
How many Kg of H2O has to be in the hydrated salts of an RSL tail before it can be detected from space? Or does it need multiple trails typically to detect H2O?
Do the RSL trails have a range of steepness that a collection of frost on Mars (say 5 to 100 Kg) could slide down once it collapsed?
Do RSLs happen below locations that would be good for H2O cornice accumulation?
Are there wind swept plains above the RSL sites that H2O frost could be forming on and collected from?
When an article says "RSL are better fit to dry flow processes" they mean that it does not show evidence liquid water but cornice of
collected H2O frost making a tiny avalanche down the cliff would be a "dry flow process", right?
So really any evidence supporting "dry granular flows" also supports a H2O cornice theory?
If the H2O cornice avalanche made some hydrated salts for a visible RSL trail how long would we expect it would take for the hydrates to go away"?
How long are RSL trails observed to take to fade out?
Do spectral analyses of light patches above RSL show H2O ice signatures?
Are RSL more common on shaded, wind-exposed slopes vs. exposed ones?
How do RSL timing align with modeled frost accumulation/sublimation cycles?
Can AI mapping correlate white areas with RSL initiation rates?
What subsurface ice depth (via radar) exists near RSL sites?
Are the temperatures at RSL sites reasonable for CO2 ice?
Could a cornice with a mixture of CO2 and H2O better fit the evidence?
Do RSL sites show aeolian features indicating frost transport?
Are there coming Mars missions either in orbit or on the ground that should give us more details on the RSLs?
Some say it sounds good and fits the evidence well but "the numbers work against H2O cornice avalanches" but could the numbers be wrong somehow?
What other questions could help to evaluate this H2O cornice avalanche theory?
Is there any experiment that we could do here on Earth that might support or contradict the theory?
The competing theories seem to be "Dry Granular Flows", "Transient Brine" or "Hygroscopic Soil".
I think the Dry Granular Flows (not counting frozen H2O) does not provide a good explanation of the tracts seen that
fade away or the hydrated salts. If it were dry sand scraping a surface layer to expose lower darker rock the path it would
not be such a sharp distinction on the edges of the path and it would not completely go away in such a short time and there
would have to be far more material at the bottom. Any liquid brine would boil away in a short time even if it were
underground. For all 3 of these it seems that
after thousands of years these would not still be going on year after year in
large enough amounts to fit the recurrent nature of what is observed. If it was sand in dry granular flows there
should be deposits at the bottom after thousands of years. Only the wind collecting frost into a cornice seems capable of
producing yearly recurrent activity and leaving a hydrated trail. Only the H2O cornice theory reconciles the vanishing act
of RSLs in weeks to months with their chemical fingerprints. As Sherlock Holmes said, "When you have eliminated all which is
impossible, then whatever remains, however improbable, must be the truth." While some current understanding makes some
people think H2O cornices are improbable, I think they must be the truth as the other theories are impossible.
Please make some percentage estimates for the chances of each of the RSL theories being the dominant one 20 years from now.
So AI results as of Nov 2, 2025:
The estimate of chances that H2O Cornice Avalanches on Mars is the dominant theory for RSLs in 20 years:
- ChatGPT (chatgpt.com) 25%
- Grok (grok.com) 10%
- Claude Sonnet 4.5 (claude.ai) 10%
- Deepseek (chat.deepseek.com) 35%
- Gemini 2.5 Pro (gemini.google.com) 5%
- Gemini 2.4 Pro (openrouter.ai) 5%
- Claude Sonnet 4 (openrouter.ai) 25-35%
- Grok 4 (openrouter.ai) 5%
- GPT-5 (openrouter.ai) 3%
- R1 0528 (openrouter.ai) 45%
- Amazon Nova Premier 1.0 (openrouter.ai) 20%
- Deep Cogito: Cogito V2 Preview Llama 405B (openrouter.ai) 40%
- Perpexity Sonar Pro Search (openrouter.ai) 15%
- OpenAI GPT-5 Pro (openrouter.ai) 3%
- Mistral Magistral Medium 2506 (thinking) (openrouter.ai) 50%
- deepseek.v3.1.671b.q8 (ollama.ai) 20% that H2O and 15% that H2O and CO2
- deepseekR1 (ollama.ai) 35%
- gpt-oss (ollama.ai) 2%
- llava (ollama.ai) 5-15%
- llama4m (ollama.ai) 40%
- qwen3.235b (ollama.ai) 20-35%
- gemma3 (ollama.ai) 40%
- cagito (ollama.ai) 40%
The (1) Grok is the minimal paid account doing "expert" mode at grok.com. The (2) GhatGPT is the chatgpt.com free version. The next were run on the site openrouter.ai. The others are open source using ollama.ai to download and run on my local machine. I am very happy with both openrouter.ai and ollama.ai by the way.
Most of the AI think 5 to 100 Kg of frost in a cornice would be enough H2O to darken an RSL trail.
Most think that the time the trail fades away matches the time a hydrated soil would take to dehydrate.
Most think it might be possible for wind to collect that much frost over a winter.
Most think it is possible H2O Cornice Avalanches could still be the winner in the end.
