← Back to Library

Experiments with vibe science

Rohit Krishnan turns a five-year-old's question about a dinosaur's sail into a rigorous test of how climate volatility shapes the very architecture of life on Earth. By treating the Paleobiology Database as a laboratory and large language models as fallible but tireless research assistants, he uncovers a hidden rule: when the planet shakes, ecosystems stop being unique and start converging on a narrow set of survival strategies. This is not just paleontology; it is a warning label for our own warming world, suggesting that the distinct regional flavors of marine life are already being sanded down into a global monoculture of the hardiest, most generic roles.

The Mechanics of Vibe Science

Krishnan's central premise is deceptively simple: he wanted to know if environmental pressure forces nature to repeat itself. He writes, "My hypothesis here was something like: 'if the landscape is less stable, we will see ecosystems seem more similar'." The logic follows that under extreme stress, only the most essential "job portfolios" for survival remain viable, causing disparate regions to look functionally identical even if they share no actual species. This approach reframes the concept of convergent evolution—a topic often explored in deep dives on the Paleobiology Database—moving it from a story about specific animals to a story about systemic constraints.

Experiments with vibe science

The data, drawn from the Paleobiology Database (PBDB) and climate reconstructions from the Community Earth System Model (CESM), delivered a result that was both surprising and nuanced. Krishnan notes, "Volatility doesn't make regions that already share species more functionally similar... But it does raise the minimum similarity between regions that share nothing taxonomically, it sets a floor on how different two ecosystems are allowed to be." This finding challenges the intuitive notion that shared species drive similarity; instead, the environment itself dictates the minimum complexity of life.

"When climates are volatile, ecosystems converge. And we can see it across 540 million years of prehistory."

However, the story isn't a straight line. The correlation Krishnan found was almost entirely driven by the Mesozoic era, specifically the aftermath of the Permian-Triassic extinction, where 96% of marine species vanished. This aligns with historical patterns seen in the Paleobiology Database, where the "reset" of the Permian-Triassic boundary created a unique window for convergence. As Krishnan observes, "The Mesozoic was in the sweet spot of transition and it had the extinction event in the middle, meaning there's enough range for convergence for volatility to have anything to correlate with." Critics might argue that relying on a single catastrophic event to drive a 540-million-year trend risks overfitting the data, but Krishnan acknowledges this limitation, noting that the signal drops off in the Cenozoic because modern ecosystems are too entrenched to be easily homogenized.

The Human Cost of Data and the AI Assistant

The most compelling part of Krishnan's piece is not the fossil data, but his candid autopsy of using AI to do the work. He describes a workflow where he acts as the principal investigator, constantly correcting the "lazy" and "mediocre" tendencies of the models. He writes, "The models just absolutely love mediocrity... They can't wait to sand the edges off any crazy ideas you have." This is a crucial insight for any professional considering automated research: the tools are indefatigable but lack the boldness to follow a hunch where it leads.

Krishnan details the friction of this new workflow, noting that "there was no substitute for actually looking myself, and LLMs ability to judge their own work remains remarkably bad." He had to manually clean the workspace, correct the models' presumptions, and even force them to delete the "enormous surplus of temp folders" they generated. The process was less like commanding a supercomputer and more like managing a brilliant but chaotic intern who needs constant supervision. "Constant vigilance is essential!" he warns, highlighting that the "final boss" of any analysis remains data quality, not the sophistication of the algorithm.

"It's brilliant, it's indefatigable, it's a little dumb, it's annoying, it believes weird things, but it'll do whatever you ask it to."

This "vibe analytics" approach allows a non-expert to test hypotheses that would traditionally require a PhD, but it introduces a new risk: the confidence of the user may outstrip the reliability of the tool. Krishnan admits that his initial theory about tectonic plates causing convergence was wrong; the data showed it was temperature change, not geography, that mattered. "The plates matter because they cause climate volatility, not because of the geography per se," he concludes. This correction underscores the value of the method: it forces the researcher to confront the data's reality rather than their own intuition.

A Warning for the Modern Era

The ultimate payoff of Krishnan's experiment is a prediction for the present. With current warming rates sitting in the top 10% of the Phanerozoic record, the same mechanism that homogenized ancient seas should be active today. Krishnan posits, "If this theory is right, marine ecosystems today should be losing their regional distinctiveness and converging on a narrower job menu." The data already shows that "suspension feeders" are expanding while "mobile predators" are shrinking during volatile periods.

This is not a neutral observation. It suggests that the rich tapestry of global marine life is being stripped away, replaced by a resilient but impoverished baseline of survival. Krishnan connects this back to his son's question about the Spinosaurus, noting that while he still doesn't have a perfect answer for the sail, he can now explain that the Cretaceous oceans were "converging on a limited menu of ecological jobs." The implication for the modern world is stark: we are not just losing species; we are losing the diversity of function that makes ecosystems robust.

"When in volatile climates the entire job portfolio homogenizes across regions regardless of which specific jobs expand or contract."

A counterargument worth considering is whether modern human intervention—such as fishing quotas or marine protected areas—could break this natural cycle. Krishnan's model assumes natural volatility, but the current crisis is anthropogenic and accelerating. If the "floor" of similarity is rising too fast, the entrenched incumbents of the Cenozoic may not have time to adapt, leading to a collapse rather than a convergence. The "liquid markets" analogy Krishnan uses for economics holds here: if the market is too choppy, nothing emerges at all.

Bottom Line

Krishnan's "vibe science" experiment succeeds not because it provides a final answer, but because it demonstrates a new way to ask questions of the deep past using modern tools. The strongest part of the argument is the identification of a "floor" for ecosystem similarity under stress, a non-obvious pattern that challenges standard evolutionary narratives. Its biggest vulnerability lies in the uneven quality of the fossil record and the current limitations of AI agents to handle truly exploratory research without human intervention. Readers should watch for the next phase of this work: testing whether the predicted convergence is already visible in modern marine data, a signal that would confirm we are entering a new, less diverse epoch of Earth's history.

Deep Dives

Explore these related deep dives:

  • Convergent Evolution: Limited Forms Most Beautiful Amazon · Better World Books by George McGhee

  • Paleobiology Database

    The author explicitly relies on this specific, non-commercial repository of fossil occurrences to test hypotheses about ecosystem stability, making it the foundational data source for the entire 'vibe science' experiment.

  • Community Earth System Model

    This specific climate simulation tool allowed the author to generate 10-million-year snapshots of the Phanerozoic, providing the necessary environmental context to correlate tectonic instability with biodiversity patterns.

  • Convergent evolution

    While the author mentions this as a favorite topic, a deep dive into the specific mechanisms of convergence explains the child's core question about why unrelated species like Spinosaurus and sailfish evolved similar sails under different pressures.

Sources

Experiments with vibe science

by Rohit Krishnan · Strange Loop Canon · Read full article

Some of you who know me know that I’m obsessed with prehistoric animals. It restarted because of my older son got obsessed with animals both alive and extinct when he was two years old, and in the more than half decade since then it’s become an all consuming passion in the Krishnan household.

At some point a few months ago, my younger son, the 5yo, asked me why his favourite dinosaur the Spinosaurus evolved that way and then went extinct. Convergent evolution being a favourite topic in our home, he asks why the sail had to look that way, and how it related to the sail of a sailfish. He knew the normal explanations from books and youtube, the sail helps spread away heat or be more streamlined swimming in water, but he asked anyway, as five year olds do, with intensity and expectation of a perfect answer.

Obviously only being an amateur paleontologist in my off hours I had no good answer. But I did have Codex. So I figured, let’s do this right. I should be able to go get some information about prehistoric animals, research it, and see if there was anything interesting in there I could proffer as an explanation.

Anyway, things got out of hand.

Since people have asked before about my research workflows I’ve been wondering about writing something, and so thought this was a great case study to write up. Especially since I’m not an expert in the field and therefore am liable to have made any number of silly mistakes, makes it much more fun!

Basically, turns out there’s this database called PBDB, the Paleobiology Database, which has details about nearly 2 million fossil occurrences - what was found, where, when, and more. I downloaded it and started playing with it. It was much (much) better for marine fossils because the record is better (even invertebrates have hard shells that preserve well and deposited in sedimentary, plus PBDB has better annotations for some reason) so that’s what I looked at. And for climate, I used reconstructions from a global Earth system model (CESM) that simulates what Earth’s climate looked like at 10-million-year snapshots across the entire Phanerozoic.

I had a firm belief that Earth is unique in having tectonic plates and that’s a major reason for our biodiversity, because it occasionally etch-a-sketches the lot and ecological niches emerge. I’ve had the same ...