A U-2 surveillance aircraft photographed a field near Srebrenica before July 11, 1995: empty ground. Photographs of the same field taken in the days that followed showed freshly turned earth — the outline of what would be identified as mass graves.^[3] The intelligence apparatus had already done its job. It had the correct answer before almost anyone outside a small circle in Washington did.

That answer did not travel far, and not because it could not have. The CIA's deputy director of intelligence, John Gannon, was permitted to give only an abridged account of the massacre in testimony to Congress on August 9, 1995 — the same day the fuller evidence was presented to the UN Security Council.^[3] Declassified records assembled since then confirm the intelligence existed in far more specific form than what reached the public that day.^[4] The machine had produced a correct, timely, specific finding. What happened to that finding afterward was a decision, made by people, not a further output of the surveillance system that generated it.

The decision that actually mattered had already been made on the ground days earlier. Bosnian Serb forces under General Ratko Mladić overran the UN-declared "safe area" of Srebrenica on July 11, 1995, and in the days that followed murdered at least 7,000 Bosnian Muslim men and boys — the worst act of genocide in Europe since the Second World War.^[1] Radovan Karadžić, the Bosnian Serb political leader, was later found to bear command responsibility for the same crime.^[2]

The institution built to answer for that decision already existed, but only barely: the UN Security Council had created the International Criminal Tribunal for the former Yugoslavia in May 1993, under Resolution 827, while the war that would produce Srebrenica two years later was still being fought.^[2] Even created early, the Tribunal could only ever prosecute after the fact. Mladić was not convicted of genocide at Srebrenica until November 2017 — twenty-two years after the killing. Karadžić's sentence for the same crime was not raised to life imprisonment until 2019.^[2] A tribunal, however quickly it is stood up, is a machine that processes what a decision already produced. It cannot reach backward into the decision itself.

The international community's answer to that limitation was to stop building ad hoc tribunals and build one permanent court instead. The Rome Statute, adopted in 1998 and in force by 2002, created the International Criminal Court specifically, in its own stated purpose, to remedy the deficiencies of ad hoc tribunals and to deter future war criminals rather than simply try past ones.^[5] It was, explicitly, an attempt to move the mechanism from after to before.

It is still after. In March 2009, the ICC issued the first arrest warrant in its history for a sitting head of state, Sudan's Omar al-Bashir, for crimes committed in Darfur; genocide counts were added the following year.^[6] Al-Bashir spent the next decade traveling through Africa and beyond and won two more presidential elections, never arrested, because the court has no police force of its own and depends entirely on the cooperation of the same states — including some legally obligated to hand him over — that simply declined to act.^[7] The permanent court, built specifically to fix the after-the-fact problem, produced the same shape again: a correct legal finding, and a decision by other people, elsewhere, about whether that finding would ever touch the person it named.

None of this is a design flaw waiting on the next redesign. This site has already argued that a machine can be reconstructed because it has a fixed structure, and a decision cannot, because it doesn't. A tribunal, ad hoc or permanent, is a machine built from evidence, testimony, and law — extraordinarily good at reconstructing what a decision already did. It has never once been the thing standing in the room at the moment the decision was made, and every version of it built since Nuremberg has had to accept that same limit. That is the actual cycle: not incompetence, not a fixable bug in institutional design, but the same gap between the decision and the machine, showing up again in different buildings under different names.