I’ve been thinking about how useful objects in our own Solar System are when we compare them to other stellar systems. Our situation has its idiosyncrasies and certainly does not represent a standard way for planetary systems to form. But we can learn a lot about what is happening at places like Beta Pictoris by studying what we can work out about the Sun’s protoplanetary disk and the factors that shaped it. Illumination can come about in both directions.
Think about that famous Voyager photograph of Earth, now the subject of an interesting new book by Jon Willis called The Pale Blue Data Point (Princeton, 2025). I’m working on this one and am not yet ready to review it, but when I do I’ll surely be discussing how the best we can do at studying a living terrestrial planet at a considerable distance is our own planet from 6 billion kilometers. We’ll use studies of the pale blue dot to inform our work with new instrumentation as we begin to resolve planets of the terrestrial kind.
But let’s look much further out, and a great deal further back in time. A 2003 detection at Beta Pictoris led eventually to confirmation of a planet in the early stages of formation there. Probing how exoplanets form is an ongoing task stuffed with questions and sparkling with new observations. As with every other aspect of exoplanet research, things are moving quickly in this area. Perhaps 25 million years old, this system offers information about the mechanisms involved in the early days of our own. Here on Earth, we also get the benefit of meteorites delivering ancient material for our inspection.
The role of Jupiter in shaping the protoplanetary disk is hard to miss. We’re beginning to learn that planetesimals, which are considered the building blocks of planets, did not form simultaneously around the Sun, and the mechanisms now coming into view affect any budding planetary system. In new work out of Rice University, senior author André Izidoro and graduate student Baibhav Srivastava have gone to work on dust evolution and planet formation using computer simulations that analyze the isotopic variation among meteorites as clues to a process that may be partially preserved in carbonaceous chondrites.
Image: Enhanced image of Jupiter by Kevin M. Gill (CC-BY) based on images provided courtesy of NASA/JPL-Caltech/SwRI/MSSS (Credit: NASA).
The authors posit that dense bands of planetesimals, created by the gravitational effects of the early-forming Jupiter, were but the second generation of such objects in the system’s history. The earlier generation, whose survivors are noncarbonaceous (NC) magmatic iron meteorites, seems to have formed within the first million years. Some two to three million years would pass before the chondrites formed, containing within themselves calcium-aluminum–rich inclusions from that earlier time. The rounded grains called ‘chrondules’ contain once molten silicates that help to preserve that era.
The key fact: Meteorites from objects that formed during the first generation of planetesimal formation melted and differentiated, making retrieval of their original composition problematic. Chondrites, which formed later, better preserve dust from the early Solar System and also contain distinctive ‘chondrules,’ which solidified after going through an early molten state. But the very presence of this isotopic variation demands explanation. From the paper:
…the late accretion of a planetesimal population does not appear readily compatible with a key feature of the Solar System: its isotopic dichotomy. This dichotomy—between NC and carbonaceous (CC) meteorites —is typically attributed to an early and persistent separation between inner and outer disk reservoirs, established by the formation of Jupiter or a pressure bump. In this framework, Jupiter (or a pressure bump) acts as a barrier that prevents the inward drift of pebbles from the outer disk and mixing, preserving isotopic distinctiveness.
But this ‘barrier’ would also seem to prevent small solids moving inwards to the inner disk, so the question becomes, how did enough material remain to allow the formation of early planetesimals at the later era of the chondrites? What is needed is a way to ‘re-stock’ this reservoir of material. Hence this paper. The authors hypothesize a ‘replenished reservoir’ of inner disk materials gravitationally gathered in the gaps in the disk opened up by Jupiter. The accretion of the chondrites and the locations where the terrestrial planets formed are interconnected as the early disk is shaped by the gas giant.
André Izidoro (Rice University) is senior author of the paper:
“Chondrites are like time capsules from the dawn of the solar system. They have fallen to Earth over billions of years, where scientists collect and study them to unlock clues about our cosmic origins. The mystery has always been: Why did some of these meteorites form so late, 2 to 3 million years after the first solids? Our results show that Jupiter itself created the conditions for their delayed birth.”
Image: This is Figure 1 from the paper. Caption: Schematic illustration of the proposed evolutionary scenario for the early inner Solar System over the first ~3 Myr. (A) At early times (t ~ 0.1 Myr), radial drift and turbulent mixing transport dust grains across the disk. (B) Around ≲ 0.t to 1 Myr, primordial planetesimal formation occurs in rings. (C) By ~1.5 Myr, growing planetary embryos start to migrate inward under the influence of the gaseous protoplanetary disk, whereas Jupiter’s core enters rapid gas accretion phase. (D) Around ~2 Myr, Jupiter’s gravitational perturbations excite spiral density waves, inducing pressure bumps in the inner disk. Giant impacts among migrating embryos generate additional debris. Pressure bumps act as dust traps, halting inward drift of small solids and leading to dust accumulation. (E) Between ~2 and 3 Myr, dust accumulation at pressure bumps leads to the formation of a second generation of planetesimals. Rapid gas depletion in the inner disk, combined with the presence of these traps, limits the inward migration of growing embryos. (F) By ~3 Myr, the inner gas disk is largely dissipated, resulting in a system composed of terrestrial embryos and a second generation of planetesimals—potentially the parent bodies of ordinary and enstatite chondrites—whereas the inner disk evolves into a gas-depleted cavity.
A separation between material from the outer Solar System and the inner regions preserved the distinctive isotopic signatures in the two populations. Opening up this gap, according to the authors, enabled regions where new planetesimals could grow into rocky worlds. Meanwhile, the presence of the gas giant also prevented the flow of gaseous materials toward the inner system, suppressing what might have been migration of young planets like ours toward the Sun. These are helpful simulations in that they sketch a way for planetesimals to form without being drawn into our star, but there are broad issues that remain unanswered here, as the paper acknowledges:
Our simulations demonstrate that Jupiter’s induced rapid inner gaseous disk depletion, gaps, and rings are broadly consistent with both the birthplaces of the terrestrial planets and the accretion ages of the parent bodies of NC chondrites. Our results suggest that Jupiter formed early, within ~1.5 to 2 Myr of the Solar System’s onset, and strongly influenced the inner disk evolution….
And here is reason for caution:
…we…neglect the effects of Jupiter’s gas-driven migration. This simplification is motivated by the fact that, once Jupiter opens a deep gap in a low-viscosity disk, its migration is expected to be fairly slow, particularly as the inner disk becomes depleted. Simulations show that, in low-viscosity disks, migration can be halted or reversed depending on the local disk structure. In reality, Jupiter probably formed beyond the initial position assumed in our simulations and first migrated via type I migration and eventually entered in the type II regime… but its exact migration history is difficult to constrain.
The authors thus guide the direction of future research into further consideration of Jupiter’s migration and its effects upon disk dynamics. Continuing study of young disks like that afforded by the Atacama Large Millimeter/submillimeter Array (ALMA) and other telescopes will help to clarify the ways in which disks can spawn first gas giants and then rocky worlds.
The paper is Srivastava & Izidoro, “The late formation of chondrites as a consequence of Jupiter-induced gaps and rings,” Science Advances Vol. 11, No. 43 (22 October 2025). Full text.
The assumed premise of this paper that Jupiter was needed to form the first planetesimals is not correct. There were already pressure bumps, radial density gradients, turbulence, instability zones in the accretion disk without Jupiter.
The snow line matters here for the density of the planets being more dense in the inner solar system and also most important the centrifugal force which is the angular momentum of the accretion disk prevents immediate fall into the Sun. There are other reasons, but these are the main ones. This idea came to me first so I confirmed it on Open AI Chat GPT.
@Geoffrey
Are you aware that ChatGPT is sycophantic and tries to support the POV and beliefs of the prompter? This is now so well known that OpenAI is promising to [try to] prevent this, as it can lead to emotional attachment, psychosis, and even self-harm.
I have 2 suggestions to try to test for this:
1. Prompt again with a different, but incompatible hypothesis. Does ChatGPT confirm this hypothesis, too, or explain why that is incorrect?
2. Add that the response must be supported with at least one citation that supports the hypothesis, so that you can check the accuracy of the response. This avoids the experience I have had in the past when reading Prager U’s econ articles that were not supported by the cited econ papers. I have checked the content of citations on other Gen AI on technical issues to occasionally find similar mistakes in the “interpretation” of the cited papers’ content. This should be an obvious problem because Gen AI uses context statistics to guess at the next word/phrase. There is no understanding of the prompt or the texts it can locate. The “intelligence” is a simulation, not real.
Even my probably wrong speculations are at least understood by me. That is not true of Gen AI, and until they have a true understanding of the world, I doubt that any claim of AGI in the near future based on LLM technology is realistic. The “thinking” mode of current AI isn’t really thinking at all, as we understand human minds work, whether our thinking produces a correct or incorrect response. IMO, Gen AI remains more like Kahneman’s “System 1, or thinking fast”. This is good for simulating human speech, but not for thinking, unless that thinking is to execute an algorithm, something that works for logic-based tools such as mathematics and coding a computer language.
Stating that ChatGPT supports a hypothesis is more like an appeal to authority. Worse, because that “authority” may not be more than a “hallucination” (i.e., b****hit), unless supported by a relevant citation[s].
The usefulness of CatGT or any other escapes me, seeing no advantage over a simple google search which brings me the primary sources.
“Stating that ChatGPT supports a hypothesis is more like an appeal to authority. Worse, because that “authority” may not be more than a “hallucination” (i.e., b****hit), unless supported by a relevant citation[s]..” I agree completely with this, however, if one knows a subject, field or domain well that is has a knowledge of it, then one should really be able to corroborate it using thought experiments oneself as I have already written. Here we must assume some ideas in astrophysics to be axioms such as how an interstellar gas cloud collapses to from an accretion disk and proto planets in stellar nurseries have already been observed and therefore their physics is well known and supported. One should know that physics and research it before jumping to conclusions which is what I think that has happened in this paper.
Chat GPT only draws from this knowledge on the internet. I clearly wrote I came up with the idea first which means I used my own knowledge of planetology and astrophysics and then I corroborated using Chat GPT. I really don’t need to check any references as I assume the idea that the cloud of gas flattens out to make an accretion disk to be correct. i.e., the thing collapses vertically and flattens out due to the centrifugal force keeping the dusk and gas from moving horizontally which applies also to the orbits or any iron planetesimals. Centrifugal force is really classical physics and general relativity describes better the force that is gravity which is what keeps the orbits stable and the iron planetesimals in the inner solar system.
Then you are saying the authors’ model is wrong, and you are right. fair enough, so why bother asking a GenAI for confirmation of your idea? I generally accept that
– George Box
applies to statistics, but perhaps not to ALL computational algorithms.
CD has posted articles on new models of planetary formation, and yet somehow the results can change. This suggests that planetary formation is either not a fairly cut-and-dried phenomenon that is deterministic and based on some fundamental physics, or that the physics is perhaps more subtle, and that additional variables or algorithmic assumptions can produce different, perhaps emergent, results. This may be attributed to tweaking the chaos that is inherent in systems that rely on precise measurements to be deterministic, rather than stochastic, i.e., knowing the position and velocity of every particle results in the perfect prediction of the universe into the far future, doesn’t work due to the “Uncertainty principle” at the atomic level.
However, classical physics does work, e.g., Boyle’s Law on gases, and increasing entropy determines the most probable state of 2 gases of different temperatures mixing.
As Michael Fidler notes below in his experiment with Grok, changing the prompt does result in different responses, with the AI even correcting itself in a “conversation”. This is almost inevitable with the stochastic LLM model, where responses cannot converge to a common solution as there is no known response in the training – exactly as one would expect if the prompt is for a novel idea.
So I ask you again, test out the prompt and response on ChatGPT (which model are you using and with what type of response?) to see if the responses are convergent and confirm your idea, or whether they change under different prompts, as Michael Fidler found with Grok. You could use different context documents to try (futilely, I would predict) to get confirmation of your model idea.
Regarding physics models. Isn’t the observation about galaxy rotation across its radius only solved with either dark matter, or changes in the equations governing the effect of gravity, or…? Observations should always take precedence over models, which are simplified representations of what one believes is reality.
I don’t find that Chat GPT’s comes to the wrong conclusions. In fact I have had great difficulty in stumping it with subjects and fields I am an expert or savvy. It does draw information in a way that conforms to what is the generally accepted view in astrophysics, and does not look for errors unless you ask it too. I refer to Chat GPT not for reason of authority because we know that the first principles of physics are what makes something right and not authority. One always has to support one’s hypothesis with first principles in order for it to be right. I am using those principles to come to my conclusion. Anyone who uses those first principles will come to the same conclusion and one has to know how to do that, to use one’s scientific intuition using thought experiments and ask questions like is there anything wrong with this idea. I use chat GPT as a reference assuming everyone else also knows how to use their scientific intuition which is not always the case. My prediction is that I don’t think that this paper will be accepted by astrophysicists who stick to first principles so it is not just my idea, but something already generally known like telescopic observations of actual accretion disks supporting these first principles.
The fact that Chat GPT can change its mind is one of its best features becuase it shows at least from my questions that it can see if an idea conforms to first principles and correct ideas which were generally accepted, but did not stick to first principles. As far as novel ideas go, they still have to stick to the foundations in physics and we should never assume just because an idea is popular and generally accepted in astrophysics that it is a foundation idea in physics. Sometimes Chat GPT does stick to what is generally accepted, but if one gives it one’s references and the first principles, I have found that it agrees with me, but only when I give it the answer that what is generally accepted was based observations not supported by modern first principles such as GR, SR, QFT, etc so it is an obsolete observation or the wrong principles were used which makes the hypothesis wrong since the premise is wrong. Recall, that the philosophy of physics is that the first principles are considered to be a priori, mind independent, objective and therefore part of physical reality itself,, in other words, physicsts don’t consider it to be a belief system since the four forces of nature don’t care whether or not we believe in them, for they continue to work anyway.
I also can find out information about subjects I don’t know and get information. I always as why and how something works. Now it takes the scientific intuition to corroborate that. One has to read some papers or books on a particular subject to get ideas which are the answers to questions. You want to know how something works. You got an idea looked it up and you were right, because somebody already solved that problem using their knowledge. Their knowledge becomes your knowledge. One has to have those foundations before one can discover something new or at least know if something is valid. Otherwise one will simply rediscover the wheel.
I went through that with Grok. It would come up with the usual well known material. Then if change the parameters it would give a different answer and correct itself. You can get good results but you have to dig deeper and keep asking the right questions.
As for Jupiter, I have aways thought the earth was in some ideal conditions in it’s location that made it’s formation beneficial for life. This should become apparent when we find similar systems.
Interesting article in Nature on why water may be very common in super earths.
Published: 29 October 2025
Building wet planets through high-pressure magma–hydrogen reactions.
https://www.nature.com/articles/s41586-025-09630-7
All current public AI are the equivalent of glorified search engines. They don’t know anything more than what exists on the Internet, and even then they do not actually “know” anything in the way a human does – or even other mammals, for that matter.
As for an AI ever gaining true sentience/consciousness and having real feeling, which are an essential part of being truly aware and smart, at least for us, this next piece adds that any emotions on an AI’s part such as HAL 9000 – or in present and future true AIs – will be programmed actions and responses, not a sign of real feelings, or at least not emotions such as we possess and recognize:
https://theconversation.com/ai-like-hal-9000-can-never-exist-because-real-emotions-arent-programmable-94141
As for Jupiter being in the vicinity of where Earth is now, that is a paradigm that took the finding of real such exoworlds to overcome. Before then most astronomers assumed alien solar systems would be variations of our Sol system, with small rocky planets near the star and the big gas giants in the outer regions.
Even in Carl Sagan’s 1980 book Cosmos, the diagrams of various predicted solar systems generated by a computer had all but one such system looking similar to ours. The one exception, if I remember correctly, had a huge gas giant nearest to its star and a bunch of little ones trailing off into more distant solar orbits. Or it may have been an actual star instead!
I thought Sagan had borrowed the diagram idea above from this book, but I could not find online the diagrams I am thinking of. This is an excellent classic work to own and read anyway, being very relevant to the topic of this post…
https://www.rand.org/content/dam/rand/pubs/commercial_books/2007/RAND_CB179-1.pdf
In the movie, the interview with the 2 astronauts with the BBC, when asked if HAL has real emotions, Bowman says that while it feels like that, it could just be programming. This makes sense, as back in the 1960s, AI was almost entirely symbolic. However, in the book (copyright 1968), Clarke writes that Minsky and [I. J. ?] Good had worked out how neural networks could be created automatically in the 1980s, which allowed these artificial neural networks [ANN] to be grown like a human brain. [This was a remarkably prescient observation, even if Minsky and Papert demolished the use of simple ANNs (Perceptrons) as not able to handle certain classifications. This pushed AI development back to symbolic AI until 3-layer networks were shown how to be made and trained with backpropagation for synapse weights, which could handle all separation cases.] But back to the book:
(clearly a dig at Dreyfus.) Later, Clarke writes that HAL could think because HAL could easily pass the Turing Test. This was accepted until recently, as few symbolic AI programs could reliably pass this test. With the computing power that allowed huge ANNs to be built with vast amounts of text to train on, AIs can now easily pass the Turing Test except in certain circumstances. However, the idea that passing this test was effectively synonymous with human thinking, as Turing thought back in 1951, is now hotly contested. The GenAI companies use “thinking” mode as a choice when responding to prompts. Is this marketing, a cynical way to train users to believe this is happening, or wishful thinking to support the idea of imminent AGI? AFAIK, GenAI is not considered to be truly emotional, nor sentient, although, like HAL9000, a subset of users think it is.
Whether emotions are needed for sentience is another matter. Sociopaths have to simulate empathy and emotions, but we don’t believe they are not sentient. Is consciousness needed? Philosophers have the concept of zombies, intelligence without consciousness. We may be on the cusp of being able to decide this…if we can decide how consciousness is invoked in brains. We have the “spot on the forehead” test for animals (and babies) looking at themselves in a mirror. If they recognize that the reflection is of themselves with this spot that needs to be removed, the animal is considered conscious. Mirror Testing: Which Animals Demonstrate Visual Self-Recognition?.
Re: habitable Planets for man – Dole.
I have this book, and the more popsci one written with Isaac Asimov. The only material concerning planetary arrangements is “Spacing of the Planets in the Solar System,” pp.49-52. No other arrangements are discussed. I vaguely recall a SciAm article about de novo simulation of planetary formation where one arrangement was for the gas giants to be close to the sun and the rocky planets deeper into space. (published in the 1980s?)
While writing and researching my previous post for this thread, I wanted to add this great article I found below. It gives a very nice and well-illustrated overall history of how humanity has envisioned our planetary neighborhood through the millennia – which is an important perspective to have while we figure out what our Sol system used to look like…
https://brewminate.com/modeling-the-cosmos-ancient-greece-to-carl-sagan/
And on a related note, I can never recommend this historical discussion on the subject highly enough…
https://tofspot.blogspot.com/2013/10/the-great-ptolemaic-smackdown-table-of.html
These models can be the mathematical equivalent of motivated reasoning. We come up with a model and fiddle with the parameters until it produced the expected results. We then declare the model to be “correct”. Or as Hamlet would say, “aye, there’s the rub!”
Many (infinite?) models can reach the same result by means of parameter adjustment. That is unconvincing, and most scientists understand that. It is far more interesting when a model can, without unique parameter adjustment, generate stellar systems for two, three and more that match observations.
I have my own experience beating models into submission. As compute power increased over time to support more granular and accurate physical modelling (variation of FEA) those old contrived models were shown to be entirely wrong. We then proceeded to beat the latest models into submission. And so it goes.
I suggest taking any current stellar system formation modelling system with a large grain of salt. Stellar system formation is an exceedingly difficult challenge for models and they all get it wrong for a variety of reasons. Whether some of the current crop are indeed useful is TBD.
Opinions of modelling experts (or worse, LLM) are really just that: opinions. With time and effort we’ll do better. Think of them more as hypothesis generators and testers rather than definitive answers.
This an important paper in helping to explain the distribution of planetary systems and it may give a pointer to the Fermi paradox.
First off, the percentage of planetary systems with cold Jupiters is very low—about 1% as best I can tell. A far more common type of system is one where similar-sized planets form and migrate in, to close pack around the primary. In some cases, such as Trappist 1, you get neat resonant chains. In a lot of systems though, as the planets migrate inwards to pack tightly around the star, there is planet-planet scattering. Another more common star system is one with a hot Jupiter. These are thought to disrupt planetary formation, scattering planets, as the gas-giant migrates inwards.
On most of the close packed systems, the planets are inside the habitable zone, and those that are in the habitable zone have migrated from the out system, which means they have too much in the way of volatiles, so they finish up being ocean worlds at best.
These planets tend to size in proportion the their primary ranging from Earth-sized for small red-dwarfs to Neptune-sized for G stars like our sun.
From this paper, it appears that get get an Earth-like planet, a relatively dry planet far from the star, you need a cold Jupiter to cut the inward gas flow off, and this is rare. Gold Jupiters are almost non-existent around Red Dwarves, and their frequency increases with system metallicy, which means that early in the universe our type of system would have been very rare. This, and other factors making complex life rare, could account for why we are the first technological civilization to emerge in our galaxy.