How a chain forms, in plain words

What are we doing here.

§Chains are not a separate feature

There is no button marked "start a line of reasoning," and no construct called "argument" or "chain" anywhere in the machine's vocabulary. A chain is simply what you get, for free, the moment one doer's published output happens to be the exact fact type a different listener was already waiting for. Nothing has to be told this is a chain — the machine has no idea, and does not need to; it only ever asks each listener, independently, "has what you were waiting for arrived," and each doer, independently, "has everything you demand fired." A line of reasoning is a pattern a reader sees in the wiring, not a thing the machine tracks as such.

§A seven-step chain, walked once, plainly

One of the system's own study designs makes an unusually clean seven-step chain visible: a person refines their own understanding through seven named levels — what am I seeking to understand (L1); where does it apply and where does it not (L2); how do the parts depend on each other (L3); what shape must the answer take (L4); into what pieces does it break down (L5); what changes and what stays stable (L6); and, finally, does the whole answer serve what was actually asked, checked against the person's own response (L7). Each level's listener waits for exactly two things: the person's own refinement at that level, and the previous level's published output, arriving as a fact of a type named knowledge.L<n-1> — nothing before both are present, no reordering possible. The output of L6 becomes, unmodified in kind, the very fact L7's own listener is watching for. Seven pairs, wired so each one's condition names the last one's result — and the whole thing reads, from a distance, exactly like a structured argument advancing level by level, because that is precisely what it mechanically is.

§Chains do not have to be a straight line

The four-question study behind most of this book's worked examples is not a single chain at all — it is four separate, independent lines running side by side, each opening its own delegated sub-investigation, each releasing its own finding, and a fifth listener waiting for all four before it does anything: the report. Nothing forces a line of reasoning to be sequential. A listener's condition can just as easily name several unrelated things at once, and the doer attached to it will not fire until every one of them has arrived — which is the machine's native way of expressing "do not synthesise until every branch of the investigation is actually in."

§Where a delegated sub-investigation fits into a chain

Part 7's Cell is not a break in a line of reasoning — it is a segment of one, fenced off and returned. A chain can perfectly well have one of its steps be "hand this off, wait for a person to sign off on what came back, and only then continue" — and the return, once it happens, re-enters the chain exactly the same way any other step's output would: as a fact of a declared type, landing in the listener that was waiting for it all along.

§Where knowledge actually crosses a boundary

Every one of the examples above shares one detail worth naming up front, because 2.3 and 2.4 build the rest of the part around it: a finding never simply appears inside a new listener's slot by magic. Something has to carry it there — the person's own intention, re-entering a delegated Cell as its starting material; a Cell's accepted export, carried back into the exact slot that asked for it; a released finding, cited again inside a later step without being retyped by hand. That one carrying mechanism, used identically every single time knowledge needs to cross from where it was established to where it is needed next, is what the rest of this part is actually about.