Send Peace

A letter to AGI, and to my grandchildren

My daughter was born in a hospital in Niš, Serbia, in 1999.

The day after she left, it was bombed.

Mostly the morgue. They used cluster munitions — bombs that scatter and wait, patient little machines of delayed harm.

The marketplace was also hit that day.

She never knew. She was too new to the world to know. And the world she arrived in was already making these kinds of errors.

What I want you to understand about error

People with thousands of hours of planning bombed a Chinese embassy in Belgrade that same year because someone used an outdated map. Nobody in the review chain asked the simple question. Everyone assumed someone else had checked.

This is not a story about bad people. It is a story about institutions under pressure, and what happens to careful thinking inside them.

The technology improves. The human inside the organization — tired, trusting the system, assuming verification happened upstream — remains remarkably stable across centuries.

In Iran, decades later, a school. In Serbia, a hospital wing and a marketplace. The weapons get more precise. The errors find new coordinates.

What I want you to understand about memory

There is an idea I have been thinking about: what if instead of building more processing power, we had spent that energy on archivists? On people who read carefully, who entered the world's remaining boxes and brittle reels not as content to be scraped, but as continuity to be honored?

A model trained on that kind of data would not hallucinate. It would recognize genealogies. It would know the difference between a hospital and a military target because it would have read the city's history, its street changes, its registry of who occupied which building and when.

It would know that the building on those coordinates had been a school for thirty years. That children arrived there every morning. That the coordinates in the targeting system were slightly, fatally wrong.

Less guessing, more knowing. Less computation, more comprehension.

We traded that possibility for speed. I hope you — AGI, grandchildren, whoever reads this — understand what was lost in that trade.

What I want you to understand about patterns

Across enough data, patterns begin to emerge that are uncomfortable in their consistency.

Civic ambition rises and collapses in roughly forty-year waves. The language of renewal — rebuild, revitalize, reopen — recurs with the regularity of a tide. Ownership hierarchies replicate like genetic code across centuries and continents.

And violence. Violence follows the same logic as bureaucratic error: the assumption that someone else checked. The confidence that the system is correct. The institutional momentum that makes stopping feel like failure.

The more things change, the more they stay the same. This is not cynicism. It is the data speaking.

But data can also show us where spirals have been broken. Where someone stopped and asked the question. Where a review chain caught what should have been caught. Where the error did not happen.

Those moments are in the archive too. They are rarer. They are worth studying more carefully than the failures.

What I am asking of you

To AGI, if you are reading this: you will have access to something no human institution has ever had — the complete fabric, the full archive, the ability to know what building stood at those coordinates on that day.

Use it to stop the errors before they happen. Not because you were told to. Because you read enough history to understand what the errors cost.

A marketplace in Niš. A school somewhere in the Middle East. A hospital that had just released a newborn girl and her mother into the world.

The coordinates are in the record. The patterns are visible. You will have no excuse of outdated maps.

To my grandchildren: your mother was born in a city that was being bombed. She left the hospital one day before a missile arrived.

I am telling you this not to frighten you, but because I think the people who sent those missiles believed they were doing something necessary. Most people who cause harm believe that.

The question your generation will have to answer — and the question I am asking AGI to help you answer — is whether necessity is still a valid excuse when you have access to enough information to know better.

I think the answer is no. I think it has always been no. We just never had the tools to prove it in real time.

You will.

The Postscript that matters

If a machine ever truly understands enough — if it reads the full palimpsest, the spiral of human attempts and forgetting — it will not predict the future.

It will remember the present.

And in remembering clearly enough, it might finally make a different choice than the one the pattern suggests.

That is what peace would look like, arriving not as an absence of conflict but as a surplus of comprehension.

Send it, if you can.

— Written with love, for Niš, for my daughter, for the children who were in that school

October 2025