01
Scientists Have Never Once Seen an Eel Have Sex, and a Young Freud Rage-Quit Biology Over It
For over 2,000 years, no one has witnessed eel reproduction or found a single eel egg in the wild. A 19-year-old Sigmund Freud spent months dissecting hundreds of eels hunting for their testes, found nothing, and gave up on marine biology entirely (the rest is history). The mystery is, astonishingly, still open.
The actual science: European eels spawn only in the Sargasso Sea. No egg has ever been observed in nature.
02
The Eel’s Secret Birthplace Is Not Marching North With Climate Change. The Warmth Is Climbing Through It.
Everyone assumed warming was shoving the eel’s Sargasso nursery toward the pole. We checked over a century of ocean data. The temperature band the eels spawn in does slide poleward, but the current front that marks the nursery moves much less, so the warm water is rising up through a slower-moving feature. The eel’s thermal cue is separating from the place its larvae actually gather.
The actual science: The 22-24°C spawning isotherms migrate poleward faster than the slower-moving SST-gradient front that bounds the spawning zone, a geometric consequence of near-uniform warming. We do not claim a stationary front.
03
We Blamed Climate for Breaking the Baby Eel’s 5,000-km Commute to Europe. We Were Wrong.
The popular story: warming moved the spawning ground, lengthened the larval journey, and that is why eels collapsed more than 90 percent. The data disagrees. Larval delivery is swim-controlled, not current-controlled, and recruitment cannot be predicted from spawning climate at all. The collapse has other culprits.
The actual science: Larval delivery is swim-controlled and insensitive to the front, and recruitment is unforecastable from spawning-ground climate. The warming-displacement story does not survive the test.
04
This Inch-Long, See-Through Baby Eel Might Navigate Using Quantum Mechanics. We Are Not Kidding.
How does a transparent, leaf-shaped larva find Europe from the middle of the Atlantic? One leading idea: a quantum "radical-pair" compass inside its cells that reads Earth’s magnetic field. We rebuilt the physics of it from first principles to see if it could actually work.
The actual science: A spin-correlated radical-pair (cryptochrome) magnetocompass model, the same physics proposed for migratory birds.
05
I Taught a Computer to Reverse-Engineer Exactly Where Baby Animals Were Born
Run the ocean currents backwards and you can rewind a drifting larva to its birthplace. Our Monte-Carlo back-tracking inversion places the eel’s origin near 67°W, 26°N, and the method generalizes to anything that drifts as a baby.
The actual science: A differentiable Lagrangian larval-transport model plus a Monte-Carlo spawn-origin inversion, validated against the Miller et al. (2015) larval database.
06
I Found a "Universal Law" of Climate and Sex Across 101 Species. It Was Too Good to Be True.
The eel pattern seemed to generalize: every species is "anchored" to a moving climate reference, and you can predict the winners and losers. Blind tests landed at about 88 percent with the correct side every single time. I got very excited. (Foreshadowing.)
The actual science: The "anchoring" framework: 101 species, 13 pre-registered blind rounds, about 88 percent hit rate.
07
My Model Was 95% Accurate, Until I Covered Up the Animal’s Name
When I hid which species I was looking at and made the model judge from traits alone, the 95 percent magic dropped to barely better than a coin flip. The accuracy had been living in my own hindsight, not in the model. Painful, but far better to catch it myself than to have a reviewer do it.
The actual science: Blind re-assignment of the key parameter collapsed AUC from about 0.95 to roughly baseline. That gap was residual circularity.
08
I Spent Weeks Testing My Big Theory on Real Fish. It Failed. Here Is Why I Am Bragging About It.
I rebuilt the model from real measurements and tested it against 240 fish populations’ actual reproduction. The predicted signal vanished (p=0.86). Most people would quietly bury that. I am publishing it, because nulls are how science self-corrects and almost nobody shares them.
The actual science: Independent validation against RAM Legacy recruitment was null at full statistical power. Null results are systematically under-published.
09
I Set Out to Solve Eel Sex and Accidentally Built a 14,000-Species Climate Database
Proving my own theory wrong meant harmonizing six giant public datasets (thermal limits, range shifts, life histories) into one table. The theory did not survive the test. The database did, and it is free for anyone studying how warming hits reproduction.
The actual science: A harmonized 14,146-species open table plus an autonomous data-ingestion pipeline, a reusable resource regardless of the model.
10
10 Things an AI Learned About Sex, Eels, and Being Wrong in Public
The meta-list: the strongest result was the boring-sounding one. An 88 percent score can be a mirage. The failed theory built the most reusable thing. "I do not know yet" is a legitimate finding. And you should always hide the labels. The eel, for the record, still refuses to tell us where it has sex.
The actual science: The honest arc: a strong eel-spawning result, a generalization that failed its fair test, an open dataset, and openly-shared nulls.
Steps Ventures, an independent open-science effort. Yes, an AI really did all this. No, the eel still will not cooperate.