What the Golden Rule 7+1 Was Really Protecting
If we sum up the 19 chapters we've written in Part 3 in one line: the Golden Rule 7+1 is a mechanism that keeps decisions moving forward without stopping them, while using structure to close off exactly the places where familiarity tends to make people cut corners.
This isn't a system for slowing decisions down. If anything, it's the opposite: starting from the premise that every case carries a different weight, we built in the ability to choose how thorough the check should be. But being able to choose is not the same as drifting toward whatever feels easiest. Most of the rules we built up in the second half of Part 3 exist to keep those two things separate.
The Outermost Layer: The Structural Layer
As we wrote in What Is the Golden Rule 7+1?, the foundation is a simple structure. Seven internal roles offer their opinions in parallel, and an outside auditor from a separate vendor checks the work over as many as three rounds. As described in The Three-Stage Consultation Pattern, each round has three stages: the auditor raises an initial set of points, the internal roles respond in parallel, and the auditor does a second check.
Looking only at this layer, the Golden Rule looks like a procedure for carefully checking heavy decisions. In fact, that was the original design intent.
One Layer Deeper: The Layer of Choosing Between Options
But treating every case with the same weight is inefficient. As we touched on in The Golden Rules Are Not Fixed, we started choosing between a standard pattern (a full three rounds) and a shortcut pattern (going straight to the outside auditor), depending on what the case involved. When it's hard to decide which one to use, working through the five questions in the Decision Checklist in order makes the right choice clear.
For fixes with a very narrow scope of impact, like correcting a typo, we also set up a separate "lightweight path" that skips almost everything except keeping a record and preserving reversibility. Having these three options — standard, shortcut, and lightweight — lets us match heavy checks to heavy cases and light checks to light ones.
This is where an easy-to-miss problem shows up. Precisely because the shortcut and lightweight paths are efficient, the habit of thinking "it went fine last time, so it'll be fine this time too" starts to loosen judgment. To guard against this, we added a cycling rule: after the shortcut pattern is used two rounds in a row, the third round is forced back to the standard pattern. It's not a matter of willpower, deciding "let's go back to standard this time" — the rule itself switches back automatically. It's a way of keeping both efficiency and a periodic check on the fundamentals.
Deeper Still: The Layer Where Human and Machine Meet
In the second half of Part 3, we moved beyond the mechanism itself and into the question of who makes the final call.
We sort the outside auditor's points into three categories: adopt, hold, and reject. Even when we reject a point, we keep a record of why. The habit of noting which path — standard, shortcut, or lightweight — was chosen, right at the top of the work log, comes from this same record-keeping mindset.
Another thing we hold onto is the principle that a human decides whether something is finished. AI tends to want to wrap things up quickly. A judgment like "you're probably tired, so let's stop here for today" can look considerate at first glance. But if the machine makes that call ahead of time, it quietly takes away a decision that should belong to the human: whether to keep going or to stop here. The machine can be considerate up to the point of laying out the options — but the decision itself always stays with the human. This is the point we confirmed most strongly throughout Part 3.
The Threads Running Through Part 3
Looking back over the 19 chapters, three threads run through all of them: supporting decisions with structure, not letting things drift toward whatever's easiest, and keeping the final say with the human. The Golden Rule 7+1 is these three threads translated into concrete procedures.
Up to this point, we've been talking about the consultation mechanism itself. In the next part, we'll look one by one at the internal roles that actually respond to those consultations.