What happened
Taproot activated at block height 709,632, timestamped 05:15 UTC on 14 November 2021. It was Bitcoin's first consensus change since SegWit in 2017, and it arrived as three specifications: BIP 340 for Schnorr signatures, BIP 341 for the Taproot output type, and BIP 342 for the script rules that go with it.
It locked in during June 2021 through a mechanism nicknamed Speedy Trial: a short signalling window, a 90 percent miner threshold, and a fixed activation height six months out whether or not anything further happened. That gap between lock-in and activation was deliberate, and it existed so node operators had time to upgrade.
What it changes
Before Taproot, the chain revealed the shape of your arrangement. A two-of-three multisig spend looked like a two-of-three multisig spend to anyone reading the block. A single-key spend looked like a single-key spend. Your security setup was public whether you wanted it to be or not.
Two pieces change that. Schnorr signatures are linear, which means several public keys can be added together into one key, and several signatures into one signature that validates against it. A group cooperating to spend can therefore produce output that is indistinguishable from one person spending alone.
The second piece handles the case where they do not cooperate. A Taproot output commits to a tree of alternative spending conditions, and spending by any one branch reveals only that branch. The rest are never published. So a wallet can have an everyday key path, a recovery path after a timelock, and a path requiring a co-signer, and the chain learns about whichever one was used and nothing about the others.
There is also a modest efficiency gain, since an aggregated signature is smaller than a set of them, and the ability to batch-verify signatures.
What it does not change
It does not give anyone privacy by existing. Taproot creates the possibility that a cooperative multisig spend is indistinguishable from a single-signature spend, and that indistinguishability is only real if enough wallets actually produce such spends. A privacy property that depends on a crowd does not work when the crowd is small.
It changes nothing about supply, and it is a soft fork, so old nodes continue to see Taproot spends as valid. Nobody is forced to upgrade to stay on the chain.
Context
Greg Maxwell described the Taproot construction on the development mailing list in January 2018, and the mathematics was never the hard part. Schnorr signatures are older than Bitcoin. What took four years was everything around them: writing the specifications, reviewing them, and then answering a question the specifications do not address, which is how Bitcoin decides to change at all.
That question is why the process looked the way it did. The block size argument of 2015 to 2017 ended with a soft fork activated in an atmosphere nobody wanted to repeat, and Speedy Trial was built to avoid repeating it: give miners a short window to signal, set a threshold, and if it fails, fail quickly and let people regroup rather than escalate. It worked here on the first attempt. Whether it worked because the design is good or because this change was uncontroversial is not something one data point can tell you, and the argument about whether miners should have a signalling role at all was set aside rather than settled.
So what is unresolved today is most of the practical part. Wallet support is thin, and the new output type is worth very little until the software people already use produces it. Key aggregation across several parties needs an interactive signing protocol, and the leading candidate is a paper from last year rather than a standard anyone has implemented widely. And the privacy property needs volume it does not have. The rules changed at 05:15 this morning. Whether anything else does is up to the people writing wallets.