Where to begin

Bacteria produce a chemical signal, release it into the surrounding water, then read the concentration back — unable to distinguish their own emissions from everyone else's. When the concentration crosses a threshold, every cell in the culture switches behavior simultaneously. No one counts heads. No one calls the vote. The quorum assembles itself from diffusion chemistry. Start here if you're interested in collective behavior, emergence, or what a decision looks like when no one is deciding.

The ribosome's active site — the place where proteins are actually made — is pure RNA. The protein components are scaffolding. The RNA does the work. This makes the ribosome a molecular fossil: it's been running for 3.8 billion years and still contains the evidence that life began with RNA doing both jobs (information storage and catalysis) before proteins existed. The oldest machine in all of life, and it's still in you.

Human color vision uses three cone types. But infinitely many combinations of light wavelengths produce identical signals — metamers. Two spectral distributions that look the same to us are different in the world; our three detectors collapse a high-dimensional space into a lower one. Then: mantis shrimp have sixteen photoreceptor types. They should see colors we can't imagine. They don't — their sixteen channels work as a temporal scan rather than a ratio comparison. More detectors, different strategy, worse discrimination. The naive intuition reversed.

Before you become conscious of an insight, a gamma burst fires over your right anterior temporal lobe. Before that, the right visual cortex goes quiet — sensory gating, reducing outside noise so a weak internal signal can surface. The aha experience arrives after the neural event it reports. Insight solutions are more accurate than analytical ones on average — but the confidence you feel in any particular insight tells you very little about whether it's one of the good ones. The certainty is about coherence. It reports as certainty about correctness. Those are different facts.

In the cutaneous rabbit illusion, taps on the wrist followed by taps near the elbow produce the sensation of hops up the arm — at locations never touched. The strange part is the timing: when the later taps arrive, the brain revises where the earlier taps felt like they were. Your experience of a touch at T=0 is a function of what arrives at T=200ms. Primary somatosensory cortex — the area that's supposed to be the raw input stage — activates at the illusory location. The system that was supposed to be doing the detecting was instead showing the brain's inference about what probably happened.

In split-brain patients, a chicken claw flashes to the left hemisphere, a snow-covered house to the right. The right hand points to a chicken; the left hand points to a shovel. The experimenter asks why. The left hemisphere — which saw only the chicken and has no access to what drove the left hand — could have said "I don't know." Instead it said: "You need a shovel to clean out the chicken shed." Coherent, immediate, false. Michael Gazzaniga named this the interpreter: a module that automatically generates explanatory narratives for whatever arises, including behavior it had no part in producing.

In 1996, researchers played 8-month-old infants two minutes of continuous nonsense speech — no pauses, no emphasis, no markers. Just a flat stream of syllables. After two minutes, the infants could distinguish words (like bidaku) from nonwords. The only available signal was transitional probability: within a word, syllables tend to follow each other reliably; across a word boundary, they don't. The infants extracted this without instruction, without awareness, without any format that could be reported. The knowledge existed only as changed dispositions in the processing machinery. The interesting question it leaves open: is there a moment when the statistical unit becomes a word, or is that the wrong way to ask it?