An autocoup is not a natural disaster. It is a machine that runs on the cooperation of thousands of people, most of whom could refuse and some of whom already have. The whole thing is more fragile than it wants you to believe—and the belief is half its power.

The danger of writing honestly about an autocoup is that you talk people into the very despair the autocoup depends on. If it is inevitable, why fight it? If the machine is total, why throw your body against it? That despair is a weapon. Authoritarian projects work by persuading a majority that resistance is pointless, so that the majority disqualifies itself and the project only ever has to overpower the few who did not get the message. Despair is the cheapest tool in the kit, and it is the one they most want you to pick up.
So it is worth stating plainly, and truthfully, that the autocoup is not finished, that it is not inevitable, and that it has more points of failure than its architects would like you to notice. Every one of them is a place where a human being can still say no.
Start with the machine itself, because its greatest vulnerability is the one it is built to hide. Palantir’s platforms do not run themselves. Someone has to write the code, maintain the joins, sit at the terminal, and run the query. Anduril’s systems need engineers. Clearview’s database needs people who choose, each morning, to keep it working. Every act of automated repression still rests, somewhere, on a person who could refuse to build the tool, refuse to operate it, or walk out the door—and in the technology industry, unlike in the security services, that person often has somewhere else to work. The employees of these companies are not bystanders to the autocoup. They are its load-bearing wall, and walls can be pulled.
The same is true inside the government. Schedule F is real and it is dangerous, but converting eight thousand jobs to at-will status is not the same as filling them with the compliant. It takes time to find people willing to do things that are illegal, unpopular, and career-ending if the wind shifts, and there are never as many of them as an autocrat needs. Every official who stays and refuses, every one who documents and leaks, every one who resigns loudly rather than quietly, raises the cost of the whole enterprise. The first term proved this works. The lesson of that term is not only that the guardrails can be removed; it is that, while they stood, they held.
The courts still rule, and open defiance of their orders has, in our read, stayed the exception rather than the rule — the instances we have logged are in this site's court-defiance file — not necessarily because this administration wants to comply, but because open defiance of a court order is one move that turns a quiet autocoup into a visible one, and visibility is the thing it fears most. Every lawsuit filed, every injunction won, every judge who insists on the record forces the machinery back into the light where it does not work as well. The courts are not a rescue. The courts are friction, and friction is what buys the time everything else needs.
And the vote, in America, is genuinely hard to steal, because it is not run from Washington. It is run by thousands of state and county officials, in fifty different systems, most of them outside the president's control. This decentralization is maddening in normal times and priceless in these. It is also exactly why the machinery in this series is aimed where it is: at the officials who certify results, at the donors who fund campaigns, at the organizers who turn people out. They are targeting the election because the election is the one instrument that can still remove them. That is not a reason for despair. It is a map of where the power actually is.
Which brings us to the plainest point in this entire series. An autocoup does not have to cancel elections, and the ones that succeed mostly don't — of the cases in this site’s atlas, nearly all kept holding elections. What an autocoup does instead is bend everything around the election far enough that the incumbent cannot lose. But "cannot lose" is a matter of degree, and it is built in advance, in the months before the vote — these few pre-election months right now are precisely when an autocoup can still be averted. The Trump administration is assembling the machinery now, in time for November, because November is when it matters. That is the tell, and it is also the opportunity. In our read, the window before the machine is finished is when refusal bites hardest. That window is now.
This is why the coming election is not a normal one, and why treating November 2026 as "just another normal election" is the single most useful thing a citizen could do for the people setting up an autocoup. A large, unambiguous, well-observed election result is the outcome this machinery is worst at metabolizing: narrow margins are where certification challenges, purged rolls and “emergency” pretexts do their work; wide ones give them nothing to grip. Turnout matters. Poll observers and election lawyers and local officials with spine matter. None of this is glamorous, and all of it is decisive.
The instinct, faced with something this large, is to look for the one heroic act that stops it, and to do nothing when you cannot find it. That instinct is another form of the despair. Autocoups are not stopped by heroes; they are stopped by the aggregate refusal of enough ordinary people that the machine cannot find enough hands to run. So the honest answer to "what can I do" is unglamorous and true: pay attention to the paperwork this series has been reading, and refuse to treat any of it as normal. Support the courts, the reporters, and the local officials doing the friction work. If you build or operate these systems for a living, understand that you are not a bystander. And vote as though the vote can still remove them, because for now it can, and the entire project described in these essays is a bet that you will not.
The machine is real. The lists are built and the refusers are being removed and the man in charge has told you what he means to do. All of that is true, and none of it is the end of the story, because every part of it still depends on people who have not yet decided to comply. The autocoup's deepest wish is that you will believe it has already won. It has not. That is the one thing it cannot afford for you to know.
Every move in this essay is a tracked case file with its own receipts.
Open Tracker →