In April 2025, in a UC Berkeley lab, five people took turns resting their chins on an apparatus and staring at a pinpoint of light the size of a fingernail at arm's length. Each of them said roughly the same thing afterward: "Wow."
What they saw is a color called olo. A blue-green, a peacock green — or in the words of those who experienced it, "a deeply saturated blue-green that makes the most saturated colors in nature look pale by comparison."
This is not a newly synthesized pigment, not a new display technology, not a new dye. Olo is a color that does not exist in nature at all — not merely "undiscovered," but physically impossible to elicit with any natural light source. It can only be seen by using lasers to target the photoreceptor cells of your retina one by one, with cell-level precision.
This is the first time humans have seen a color that evolution never permitted us to see.
1. Why Some Colors Are Forever Invisible to You
To understand how strange olo is, you first need an uncomfortable fact: the colors you see are not the world's true colors — they are your brain's guesses about light signals.
The human retina has three types of cone cells — S, M, and L — sensitive to short, medium, and long wavelengths:
- S cones: sensitive to the blue band (~420 nm)
- M cones: sensitive to the green band (~530 nm)
- L cones: sensitive to the red band (~560 nm)
The system seems sensible, but it has an embarrassing design flaw: the sensitivity curves of M and L cones overlap heavily — about 85% of light that activates M cones also activates L cones. This isn't an engineering oversight; it's evolution cutting corners. Primates are the only mammals with trichromatic vision, arising from a gene duplication on the X chromosome roughly 30 million years ago. The duplicated opsin genes diverged only slightly — one leaning green, one leaning red — leaving a huge overlap.
The consequence: no natural beam of light can activate only your M cones. Your brain has never received an "M-cones-only" signal, so it has never had to generate a color experience for that case. But "M cones only" is a perfectly legal combination in color space. It was there all along — locked away by evolution.
Ren Ng, a professor of electrical engineering and computer science at Berkeley, once asked: "What if we could activate only the M cones? What would we see? Would it be the greenest green ever?"
The answer is olo.
2. The Oz System: Writing Movies onto the Retina
Turning "activate only the M cones" from a thought experiment into reality was extraordinarily difficult.
Problem one: M and L cones look nearly identical and are interleaved across the retina. How do you tell them apart? The solution came from Professor Roorda's earlier work — an instrument that images a living retina cell by cell, identifying each photoreceptor's type from its response to different wavelengths. This produces a personal "cone map" for each subject.
Problem two: Even with the map, how do you activate only M cones without hitting neighboring L cones? The Oz system's core is a very fine green laser (the same color as a laser pointer) that rapidly scans the retina, firing microsecond-scale energy pulses only when aimed at a target cell. It scans thousands of times per second, calling out each cone individually.
The precision is remarkable: Oz can "draw" images on the retina — not by projecting an image, but by writing directly onto your photoreceptors. Subjects could see letters, moving dots, even pictures of babies and fish, all "drawn" with a single-color laser.
How big is the display area? "About the size of your fingernail at arm's length," Roorda says. They'd like to make it IMAX-sized, but they're limited by how much retina the device can scan.
When Oz activated only M cones, subjects saw olo.
3. The "Wow" Moment
Experiment designer Hannah Doyle devised an elegant control: she made the laser "jitter" — slightly offsetting its aim so pulses landed randomly on all cone types instead of only M cones. Olo instantly disappeared, replaced by ordinary green laser light.
"I wasn't a subject, but I've seen olo since," Doyle said. "It's stunning. When the laser jitters, the laser's normal green looks almost yellow by contrast — the difference is that strong."
That detail deserves a pause. Olo is not "a brighter green" — it is a color for which your color experience contains no reference point. When it vanishes, ordinary green looks yellowish by comparison, the way white looks bluish when you step from a dark room into sunlight. Your visual system recalibrates.
Roorda puts it plainly: "When I compare olo with other monochromatic lights side by side, I genuinely have that 'wow' experience."
The most saturated colors in nature are monochromatic light — lasers. But even monochromatic light spills over onto adjacent cones. Olo is far more saturated than the most saturated natural color because it bypasses that spillover entirely.
4. Mary's Room and the Philosophy of Olo
In 1982, philosopher Frank Jackson proposed a famous thought experiment:
> Mary is a color scientist who knows everything about color — physics, optics, neuroscience, psychology. But she has lived her whole life in a black-and-white room and has never seen any color. One day she steps out and sees red for the first time. > > The question: Does she learn something new?
Jackson argued yes — she gains knowledge of "what it's like to see red," which no physical description can convey. This "knowledge argument" is used against physicalism.
Olo flips Mary's room inside out. Olo is not a color someone "knows everything about but has never seen." It is a color that is physically describable, mathematically legal, and yet never experienced by any human in evolutionary history. Mary at least knew red existed in the world. Before the Oz system, nobody had seen olo — and nobody even knew they hadn't seen it, because it wasn't in anyone's color vocabulary.
This reveals something deeper than Mary's room: your color experience space contains "holes" — not unexplored regions, but regions architecturally unreachable by your visual system. Olo is a dark room in color space. The door was always there; the key just required lasers and cell-level precision.
More unsettling: there may be more than one such dark room. What if Oz could precisely control S cones? What if all three cone types could be activated in ratios impossible in nature? What if humans were given a fourth cone (simulated tetrachromacy)? Every technological advance might open another dark room.
5. Tetrachromats: Women Walking Ahead of Evolution
Speaking of a fourth cone: some real people have one. Most humans are trichromats, seeing roughly 1 to 10 million colors. But some women carry four functional cone types — the extra one typically sensitive to a narrow band between orange and red — and could theoretically distinguish about 100 million colors.
The genetic basis is the same story as trichromacy's "flaw": opsin genes on the X chromosome. Women have two X chromosomes; if the M/L opsin genes on each carry different mutations, she may end up with four distinct cones. Men have only one X, making tetrachromacy nearly impossible for them.
Confirmed tetrachromats are rare. The artist Concetta Antico is the most famous — her paintings are known for exceptionally vivid color, because she sees a far richer world.
But here's an interesting twist: the new colors a tetrachromat sees, like olo, are colors that are "theoretically present in nature but nearly impossible to encounter in daily experience." Natural light almost never produces the wavelength combinations that isolate her fourth cone. She has the hardware, but the world rarely gives her the right input.
The Oz authors are aware of this. They suggest Oz could one day let ordinary people experience tetrachromatic vision by precisely controlling their photoreceptors. Oz can show us not only colors evolution withheld, but worlds only a lucky few have ever seen.
6. When AI Is Also Locked in the Room
Here the story jumps somewhere seemingly unrelated: AI.
Large language models are trained on human data. Every text, image, and conversation they learn from comes from the world of human experience — so an AI's "color space" is isomorphic to ours. It knows red is #FF0000, knows what "a sky washed clean" means, can paint starry skies in Van Gogh's style.
But AI doesn't know olo.
Not because it isn't smart enough, but because olo isn't in any training data. It was created in April 2025. No text described it before then; no image contains it. Even today, no screen can display it — it can only be experienced by writing it directly onto a retina with lasers.
This leaves AI with its own dark room: it knows how cones work, knows about M/L overlap, knows "M-cones-only" is mathematically legal — but it cannot predict what olo looks like. Because "what does olo look like" is, like "what does red feel like," a question answerable only through first-person experience.
Mary's room, recreated in AI.
More generally, olo reveals a boundary of intelligence: any system trained on finite data has an experiential space it cannot reach — not "not yet learned," but "architecturally unreachable through learning." For humans, the boundary is set by cone overlap; for AI, by the distribution of training data.
But there's a key difference: humans can use technology (the Oz system) to bypass their biological boundary. What can AI use to bypass its data boundary?
One possible answer: direct interaction with the physical world. An AI that can run its own experiments, collect its own data, and discover new phenomena is essentially building its own "Oz system" — using new inputs to open dark rooms that don't exist in its training data.
This is why AI-for-Science is considered one of the most promising directions — not because it helps humans compute, but because doing experiments is the only way to bypass the training-data boundary. Every experiment is a direct visit to the unknown space of experience.
7. A New Door
Olo evokes an image: the space of human color experience is a big house we've lived in for millions of years, assuming every room had been explored. The Oz system found a door in the wall, and behind it is a room we've never entered — containing a color we've never seen.
More exciting still: there may be more doors.
Roorda says: "We found we can recreate normal visual experience just by manipulating cells — not projecting images, but directly stimulating photoreceptors. We also found we can extend that experience. It's an open question: if we extend the signals, or generate new sensory inputs, can the brain understand them, appreciate them? I like to believe it can. The human brain is a remarkable organ, extraordinarily good at making sense of inputs — whether familiar or entirely new."
Read that passage twice. It's not just about color — it's about the brain's relationship with the unknown. The brain isn't designed for specific inputs; it's a general-purpose meaning-making machine. Give it new signals, and it tries to understand. Give it olo, and it sees a new color. Give it tetrachromatic signals, and it may see a new world.
Perhaps that's what we should hope for from AI too: not imitating human experience, but accessing experiential spaces humans can't reach — and seeing what it brings back.
Olo is a color, but it is also a metaphor: in any system's space of experience, there are dark rooms locked by architecture. The real discovery is not exploring the edges of the known space, but finding the keys to the dark rooms.
The next time you see peacock green, pause and think: what you see is only a shadow of that color. Its true form, your eyes have never seen.
Perhaps they never will. But at least now you know it's there.