Blood That Vanishes on a Leaf
A tropical rainforest in Costa Rica, mid-afternoon. A northern glass frog (*Hyalinobatrachium fleischmanni*), less than three centimeters long, sleeps on the underside of a large leaf.
Look closely and you see something impossible: its skin turns to glass. You can see its spine, its organs, even its beating heart. The blood seems to have vanished from its body.
Not seems. It really has.
In December 2022, Carlos Taboada's team at Duke University published a paper in *Science* using photoacoustic microscopy (PAM) to track red blood cells inside the glass frog. The conclusion was jaw-dropping: when a glass frog sleeps, it withdraws roughly 89% of its red blood cells from circulation and packs them into the liver. Transparency increases 2-3x.
Each morning it puts the red blood cells back into the bloodstream. It hides them again at its midday sleep. And so it goes, for its entire life.
You might think that's not a big deal. Let me tell you why it is.
Red Blood Cells: Nature's Worst Camouflage
Red blood cells are among the most opaque cells in an animal's body. They are packed with hemoglobin, and the iron in hemoglobin is especially good at absorbing green light—exactly the light plant leaves reflect. So red blood cells stand out sharply against green foliage: they absorb green and reflect red, glowing like a red lamp in a sea of green.
For a frog trying to look like a leaf, that's a disaster. Your skin can turn green, your bones can reflect light, but as long as blood flows through your vessels, you look like a leaf painted with a map of red veins.
Many marine animals solve this by being transparent throughout—jellyfish, salps, planktonic larvae—with no blood at all. Among terrestrial vertebrates, only the glass frog family (Centrolenidae), about 150 species, manages it.
Why so rare? Because on land, transparency forces you to resolve a fundamental contradiction: you need red blood cells to carry oxygen, but red blood cells are your single greatest source of visibility.
There are three ways out of this contradiction. The glass frog took the first; two fish took the others.
Path One: Hide the Blood Every Day (Glass Frogs)
Taboada's discovery relied on luck and technology.
He was originally studying biliverdin—the green pigment in another frog—when he heard about photoacoustic microscopy: a safe laser shines on tissue, molecules absorb the light and emit ultrasound, and sensors map the molecular distribution from those acoustic waves. Hemoglobin in red blood cells absorbs light naturally, with no contrast agent required—a life-saving advantage for a three-centimeter frog.
The experimental design was charmingly simple: let glass frogs sleep hanging upside down in a petri dish (mimicking their sleeping posture on leaves), shine green laser light on them, and listen to the ultrasound coming from inside.
The results were almost too clean: while sleeping, red blood cell density in the liver surged while the circulatory system emptied. On waking, red blood cells flooded back from the liver. Sleep, hide; wake, release.
89%. Not 30%, not 50%—89%.
Here lies a huge biological mystery. Concentrating 89% of your red blood cells in one organ would be catastrophic thrombosis by human standards. A long-haul flight with hours of immobility significantly raises the risk of deep vein thrombosis in humans. The glass frog stuffs the vast majority of its red blood cells into its liver for 12 hours a day, every day, with no clotting problems at all.
Nearly three years after the paper, the mystery remains unsolved. Taboada himself has said in interviews that the next step is figuring out how glass frogs do it—and that understanding it could directly inform research on human vascular disease.
The core of the glass frog's strategy is reversibility. It doesn't modify its red blood cells, doesn't delete any genes—just relocates its blood daily. The cost: transparency is only available during sleep; when awake, the blood must flow or the muscles starve for oxygen.
This is a runtime strategy: don't change the underlying architecture, just switch states when needed.
Path Two: Delete the Gene (Antarctic Icefish)
Antarctic icefish (family Channichthyidae, 16 species) took a more radical path: doing away with red blood cells entirely.
A team led by H. William Detrich at Northeastern University has studied these fish for decades. Their findings fit in one sentence: the icefish hemoglobin gene was deleted outright. Not broken—gone. Their blood is transparent and looks like water. Stain a smear on a slide and you see no dense field of red blood cells, just blank space.
That sounds like suicide. Red blood cells carry oxygen—how can you live without them?
The answer is in the environment. Southern Ocean water stays between -1.9°C and 2°C year-round, and cold water holds far more dissolved oxygen than warm water. Icefish can meet their metabolic needs on oxygen physically dissolved in the blood, no hemoglobin required. To compensate, they have exceptionally large hearts, four times the blood volume of ordinary fish, and wider vessels—the whole circulatory system was re-engineered from "small and concentrated" to "large and dilute."
This is an architectural modification: not a runtime switch but a redesign of the foundation. The cost: icefish can never leave Antarctica. Put them in warm water and the dissolved oxygen is insufficient—they suffocate. They bet everything on the assumption that Antarctica stays cold and oxygen-rich forever.
Icefish live six to seven years or more. They live slowly, but they live thoroughly—the red blood cell problem was solved once, and never worried about again.
Path Three: Break the Gene, Then Never Grow Up (Asian Noodlefish)
In January 2026, *Current Biology* published a paper that surprised even Detrich.
Asian noodlefishes (12 species, distributed along the coasts and estuaries of China, Korea, Japan, the Russian Far East, and Vietnam) also lack red blood cells. Their blood is likewise transparent white. But their genetics are entirely different from the icefish.
The icefish deleted the hemoglobin gene. Noodlefishes kept the gene but broke it—various small mutations prevent it from producing functional hemoglobin. Strangely, all 12 species also lost the myoglobin gene (the protein that makes red meat red) in a single event in their common ancestor.
But the real story is in the life cycle.
Noodlefishes live for only one year. They breed once at the end of life and die, retaining larval traits their whole lives—no scales, elongated transparent bodies, oxygen absorbed directly through the skin. This phenomenon is called neoteny: adults keep the form and physiology of juveniles.
Most fish don't need red blood cells as juveniles—small bodies, thin skin, enough oxygen by surface diffusion. Only after growing do they need red blood cells to deliver oxygen to deep tissues. The noodlefish strategy: never grow up.
If you never grow up, you never need red blood cells. Broken genes don't matter.
This is a life-cycle hack: don't modify the architecture—modify the life cycle so you remain permanently in the stage that doesn't need that architecture.
Three Strategies, One Table
Seen together, these three animals solve the same engineering problem—"how to handle a component that is both essential and a liability"—with three structurally different solutions:
| Dimension | Glass frog | Antarctic icefish | Asian noodlefish | |------|--------|----------|----------| | Strategy | Runtime switching | Architectural deletion | Life-cycle lock-in | | Reversibility | Daily reversible | Permanently irreversible | Permanently irreversible | | Genetic modification | None | Hemoglobin gene deleted | Genes kept but broken | | Environmental dependence | Low (any leaf works) | High (cold, oxygen-rich water required) | Medium (shallow water/estuaries required) | | Life-cycle constraint | None | None | Neoteny required | | Lifespan | Normal (years) | Long (6-7+ years) | Short (1 year) | | Cost | Transparent only while asleep | Can never leave Antarctica | Can never grow up |
The three strategies differ radically in information preserved. The glass frog keeps every option open and makes a decision once a day. The icefish made one decision and locked it in. The noodlefish made one decision and locked in its life cycle too.
What This Has to Do with AI Engineering
You might think this is a biology story with nothing to do with AI. But all three strategies have precise counterparts in software engineering.
Glass frog = runtime feature flag. You don't delete any code; you toggle features on or off at runtime depending on context. The cost is switching overhead and maintaining two states; the benefit is that you can always roll back. GPT-5 dynamically switching reasoning depth across tasks is a classic "glass frog strategy"—no model weights are modified, only the amount of compute at inference time.
Icefish = removing a module from the codebase. You find a feature you'll never use, delete it, and simplify the architecture. Lighter, easier to maintain. But if the environment changes and you need that capability back, you'd have to re-evolve it—and evolution is irreversible. The icefish deleted its hemoglobin gene and can never get it back. In AI, this is like pruning or distilling and then discarding the teacher model: cheap deployment, but that capacity is gone forever.
Noodlefish = keep the broken module + restrict the task domain. The code is still there but doesn't run; you work around it by limiting the system's use cases. The noodlefish lost myoglobin and broke its hemoglobin gene, but it confined itself to the "larval niche" where those functions are never needed. In AI, this is like a model with certain capabilities ablated away, where you avoid triggering the defect by constraining the prompt domain—flawless on narrow tasks, permanently unable to expand to tasks requiring that capability.
No strategy is absolutely best. The glass frog has the most flexibility but pays a switching cost daily. The icefish is the most efficient but placed the biggest bet. The noodlefish has the simplest implementation but the lowest ceiling.
The Price of Reversibility
One more detail about the glass frog.
Taboada's team found that glass frogs are fully opaque—circulatory system full of red blood cells—when awake, stressed, or anesthetized. The transparency mechanism only activates in the state the paper described as "happily asleep." This means glass frogs cannot turn transparent when danger strikes. Quite the opposite: when startled, they become opaque again because their muscles need oxygen to flee.
It's a subtle trade-off: the defense mechanism only works when defense is least needed. A sleeping frog can't move, so it relies on transparency to hide from predators. But once startled, transparency is lost, and all it can do is jump.
Doesn't this resemble the dilemma of many runtime strategies? At low load you can switch modes freely, but under emergency high load you fall back on the most basic response. Feature flags tend to be all flipped open during traffic spikes because there's no time for fine-grained switching.
The icefish doesn't have this problem—it is always transparent, never switches. But it can never leave Antarctica.
The noodlefish doesn't have this problem either—it is always transparent. But it only lives one year.
The Unanswered Question
How does the glass frog pack 89% of its red blood cells into its liver without forming clots? As of mid-2026, no answer.
Human medicine fights thrombosis constantly: deep vein thrombosis, heart attack, stroke—fundamentally, blood clotting where it shouldn't. If the glass frog's liver hosts a mechanism that lets enormous numbers of red blood cells coexist peacefully without clotting, its value to human medicine might exceed any existing anticoagulant.
But we don't know what it is.
Maybe an unknown anticoagulant molecule. Maybe a modification on the red blood cell membrane. Maybe a special structure of the hepatic microenvironment. We only know it exists, because the glass frog demonstrates it live, every single day.
The most fascinating state in science is not "we know everything" but "we've just seen the phenomenon clearly and the mechanism is completely unknown." With its three-centimeter body, the glass frog tells us: nature has a way to pack blood without clotting, and to unpack it again, daily.
We can't even do that yet.
Epilogue
A glass frog sleeps on a leaf. Its liver holds 89% of its red blood cells, its skin transparent enough to show a heartbeat. It doesn't know it's demonstrating a biological marvel humans haven't figured out, nor that it embodies an engineering philosophy: commit to nothing, delete nothing—just switch.
An Antarctic icefish cruises the Southern Ocean, its blood clear as water. It spent 16 million years deleting its hemoglobin gene, in exchange for never being able to leave those cold waters.
An Asian noodlefish drifts through an estuary, one year per life. It kept its broken hemoglobin genes and lived out an entire lifetime in a juvenile body.
Three animals, three answers to the same question. One chose reversibility; two chose irreversibility. One kept its options; two burned theirs.
Next time you face an engineering decision about whether to delete a module, think of that frog. It has been making the same decision once a day, for millions of years.