There used to be a working assumption in most adult lives that your closest friends could disagree with you about taxes, immigration, abortion, or war, and the friendship would survive the disagreement. That assumption is no longer safe. Surveys in the United States, the United Kingdom, and much of Europe show a steep rise since roughly 2012 in the percentage of people who say they would be unhappy if a close friend or family member supported the opposing political party. The number has roughly doubled in a decade. Politics is no longer one topic among many. For a growing slice of the population, it is the master category, the trait through which other traits are read.

A cross-political friendship today is therefore a different object than it was twenty years ago. It is no longer a default that needs no defending. It is a small act of resistance against the sorting machine that wants every relationship to align with every other relationship. Lilliana Mason calls this "the alignment of identities": party, religion, race, geography, and lifestyle increasingly stack on top of each other, so that knowing one fact about a person lets you guess the rest with uncomfortable accuracy. A friendship that crosses this alignment is mechanically harder to maintain because every news cycle reactivates the dividing line. It is also, for exactly the same reason, more valuable.

The personal scale matters here because the macro problem is built out of micro decisions. Nobody decides to live in a country where half the population will not date the other half. People decide, one conversation at a time, to mute an old friend, to skip a wedding, to stop replying to a thread. The country is downstream of those decisions. So is your own intellectual life. A friend who votes differently than you is, among other things, a free tutor in how roughly half your neighbors actually think — not the cartoon version pumped through your feed, but the real version, with all the contradictions and reasonable parts intact. Lose that tutor and you lose your last reliable calibration against your own side's distortions.

The hard part is that the techniques that used to keep these friendships intact have weakened. "We just don't talk about politics" works less well when politics has expanded to cover schooling, medicine, climate, what to eat, where to live, whether to have children. The neutral zone keeps shrinking. So the modern cross-political friendship has to do something the older version did not: it has to develop an explicit etiquette. Not avoidance, which fails, and not constant debate, which exhausts, but a deliberate practice — an agreement about what gets discussed at depth, what gets acknowledged and moved past, and what gets ruled out of bounds entirely.

A second difficulty is that political identity has fused with moral identity. Haidt's research on moral foundations shows that people across the spectrum are not merely calculating self-interest differently; they are weighing different moral concerns with different intensities. A friend on the other side is not a worse version of you. They are someone whose moral weights are set differently, often for biographical reasons you have never bothered to learn. The friendship survives, when it survives, because at least one party stops treating disagreement as evidence of defective character and starts treating it as evidence of a different starting configuration.

The third difficulty is informational. Eli Pariser's filter bubble is now a decade and a half old as a concept and has only deepened. You and your cross-political friend are no longer arguing over the same set of facts. You are arguing over which facts even exist, which sources are credible, and which events happened. This is the part that defeats most modern cross-political friendships. The repair is not to agree on facts in a single conversation. The repair is to agree, repeatedly and humbly, that you are each operating on a partial and curated information diet, and that the friendship is one of the few places where those diets can be compared rather than weaponized.

What you get in return, if you do the work, is rare and underpriced. You get a check on your own certainty. You get a person who will tell you when your side is being ridiculous, in a tone you can still hear. You get protection against the slow narrowing of your social world into a single hue. And you get a small, stubborn piece of evidence that the country is not as broken as the feed insists — because here you are, eating dinner with someone who voted the other way, and the dinner is fine. The personal scale is where unity is actually rebuilt or actually lost. There is no other scale where it happens.