Correlations stronger than classical
Imagine I mail you one glove and keep the other. Before either of us looks, we each have a 50/50 chance of holding the left glove. The moment you open your box and see 'left,' you instantly know mine is 'right.' That is a correlation — but it is a boring, classical one. The answer was already fixed the moment I packed the boxes; opening yours just revealed a fact that was true all along.
Entanglement is a correlation between qubits that is stronger than the glove story, and stranger. With two entangled qubits, neither one has a definite value before you look — not 'unknown to you,' but genuinely undecided. Yet the instant you measure one, the outcomes of the two are locked together more tightly than any pre-packed boxes could ever reproduce. Experiments (Bell tests) have ruled out the 'it was decided in advance' explanation. The link is real, and it is not hidden glove-swapping.
Bell pairs
The simplest entangled state is a Bell pair. Take two qubits and prepare them so that, when you measure both, you only ever see '00' or '11' — never '01' or '10' — each with 50% probability. The two qubits always agree, even though each one alone looks like a fair coin.
|psi> = (|00> + |11>) / sqrt(2)
Here is the honest, important part. You cannot rewrite that expression as (something for qubit A) times (something for qubit B). Try it and the algebra refuses. A quantum state that *can* be split that way is called separable; a Bell pair cannot, and that un-splittability is exactly what we mean by entanglement. The information lives in how the qubits relate, not in either one by itself.
Measuring one shapes the other
Now measure just one qubit of a Bell pair. Before measuring, that qubit is in a 50/50 superposition — you genuinely cannot predict whether you will read 0 or 1. Suppose you get 0. The Born rule and collapse tell us the joint state snaps to |00>. From that moment, the *other* qubit — which you have not touched, and which could be on the far side of the galaxy — is guaranteed to read 0 too.
It is tempting to say your measurement 'reached over and set' the distant qubit. Resist that picture. What changed is the statistics: before, the partner was a fair coin; after, its outcome is pinned to match yours. You did not push a value across space — you updated which correlated outcome the pair was committed to. Measuring one qubit shapes what the other will do, but only in the bookkeeping of correlations, never as a signal you control.
No faster-than-light messages
This is where the hype usually goes off the rails, so let us be blunt: entanglement cannot send a message faster than light. None. Not a bit. The correlation is real and instantaneous, but it is uselessly random unless you also compare notes through an ordinary, slower-than-light channel.
Why the firm 'no'? Whatever you do to your qubit, your friend's qubit, looked at on its own, keeps producing the same fair-coin 50/50 randomness. There is no measurement they can make, and no pattern they can watch for, that would tell them whether you measured first, what you got, or even whether you did anything at all. The link only *shows up* once you phone them — at light speed or slower — and line up your two columns of results. No phone call, no visible correlation, no message.
Why it's a resource
If entanglement can't send messages, why care? Because inside a quantum computer it is a resource — something you spend to do things that separable qubits cannot. A processor whose qubits never entangle can be simulated efficiently by an ordinary computer, so any hope of a real quantum advantage runs through entanglement at some point.
Concretely, entanglement powers protocols like quantum teleportation (moving a qubit's state using a shared Bell pair plus ordinary classical messages — and yes, those classical messages are why teleportation, too, obeys the light-speed limit). It is also part of how quantum algorithms route possibilities so they can interfere, which is the real mechanism behind quantum speedups. Be careful here: the speedup is never 'try every answer at once.' It is arranging amplitudes — entanglement helping connect them — so that wrong answers cancel and the right one is likely when you take your single measurement.