99 · Closing

This is the courtroom after both benches have been used — no verdict read out, only the record handed across and the door left unlocked.

What was the trial for?

Count the collisions this book has entered into evidence. Church and Turing settled the Entscheidungsproblem six weeks apart in 1936 without either having seen the other's proof, by methods that had nothing in common until Turing wrote the appendix showing they agreed (ch.09). Stephen Cook located the boundary of feasible computation in 1971; Leonid Levin, cut off from Western journals, found the identical theorem two years later (ch.10). Haskell Curry, William Howard, and N. G. de Bruijn each arrived at the correspondence between proofs and programs from a different direction, none of them aware of the others (ch.12). George Boole rebuilt Leibniz's calculus of reasoning without knowing Leibniz had been there, and only learned of the anticipation after his book was printed (ch.01). Four independent rediscoveries of the same seam is not four coincidences. It is what a single vein looks like when people keep striking it from opposite sides of a wall.

That is the whole claim, and it is why this book was built as a cross-examination rather than a survey. Philosophy and computing are one project — what can be known, said, and done by rule — separated by an accident of institutional history rather than by a difference in subject. Part I put the machine on the stand and let philosophy ask what a program is, whether a machine's answer counts as knowledge, how code means, whether a delegated decision keeps its morality, and who governs when architecture outruns law. Part II reversed the benches, and computation's questions were the more embarrassing ones: your ideal knower is computationally impossible (ch.10); your certainty is an engineering product with an error budget (ch.12); your account of trust is a threshold with a number on it (ch.14).

Neither court won, and that is not a failure of the method. Every hearing here ends the way an honest examination ends: with the claim narrower than it was, and the ground under it visible. Chapter 15 said why the trial cannot close. This chapter says who runs it next.

The route

Thirty days, one hour a day. Read the book's chapter first, then the source; the encyclopedia entries are load-bearing, not decoration, and every link below was checked before this file shipped. Where a chapter gets two days, spend the second one arguing with the first.

Two habits matter more than the schedule. Read one primary source per chapter rather than three summaries of it — the summaries agree with each other and the primary does not, and the disagreement is where the thinking is. And keep the last column: the exercises are the examination, and a route with no cross-examination in it is a reading list, not a training.

The license

Nothing in this book required a degree in either discipline to produce, and the record says so. Boole was a schoolmaster in Lincoln when he algebraized logic. Shannon was a twenty-one-year-old master's student when he showed that Boole's algebra was also a wiring diagram. Turing wrote the founding paper of computing for a mathematics journal and the founding question of machine minds for a philosophy journal, fourteen years apart, without changing fields — because there was no field to change. The department that separated them was founded in 1962, twenty-six years after the paper that made it possible. An institutional boundary that young has no authority over a reader.

That is the license, and it is narrower and better than permission to have opinions. It is permission to ask a question in the other discipline's vocabulary and to insist on an answer there: to ask an engineer what their system means and not accept a benchmark score, to ask a philosopher what their ideal agent would cost to run and not accept that the question is vulgar. Each field's habits protect it from exactly one kind of question, and it is always the other field's question.

The rules were set out in chapter 00 and they are the only equipment needed. The examiner sets the terms. Both voices appear, or it is a lecture. Every claim carries its evidence — a date, a work, a source that can be opened — because the fastest way to lose a cross-examination is to assert what the record will not support.

The record is now yours. The witnesses are still under oath, the questions in every chapter's Open questions were left open on purpose, and the chair at the front of the room has been empty since chapter 15. Sit down and begin.