---
title: "Chapter 15: The Last Six"
book: "Collapse or Collapse"
author: "Vinay Pasricha"
url: https://vinaypasricha.com/books/collapse-or-collapse/chapter-15-the-last-six
status: "On Kindle: https://www.amazon.com/dp/B0HNL53J5R. The full text is shared on vinaypasricha.com with the author's permission."
---

# Chapter 15: The Last Six

## 1. Why a company has edges at all

In 1937 a twenty-six-year-old economist asked a question nobody had thought needed asking: if the market is so good at coordinating work, why do firms exist? Why is anything done inside a company by instruction rather than outside it by contract? Ronald Coase’s answer was that using the market costs something. Finding the other party, agreeing the price, writing the contract, checking it was kept: each is a transaction cost, and where those costs exceed the cost of simply telling someone what to do, a firm forms around the work. The edge of the firm sits where the two costs meet.[^52] Oliver Williamson spent a career on the same line, and his finding was that the edge moves: whenever the cost of a handoff across a boundary rises, the boundary moves to swallow it, and whenever it falls, the boundary retreats.[^53]

Both were describing the edge between a company and the world. This chapter is about the edges inside a company, between its Outcomes, and it argues that they obey the same law and reach the same end. Every boundary between two Outcomes is a handoff. Every handoff costs. When the cost of a handoff is a leak, and a vector can hold both sides of it, the boundary moves, and it moves in one direction only: the downstream Outcome absorbs the upstream one. It keeps moving until there is one Outcome, and then the vectors inside it merge, as chapter 14 showed, until there is one vector. The company that began chapter 13 with ten departments and became five Outcomes does not stop at five. Everything tends to collapse into one, and this chapter follows the tendency to its end, while saying plainly where a real company may stop short of it.

## 2. The leak is the reason

Chapter 3 said where outcomes leak: between two desks, at the moment a result passes from the hands that made it to the hands that need it. A department was a workaround for the fact that no one person held every skill, and the handoff was the price of the workaround. Chapter 13 collapsed the departments into Outcomes, and the reader may have assumed that the leaks went with them. They did not. They moved. Where Pravaha once leaked between sales and the order desk, it now leaks between the winning Outcome and the fulfilling one: an order won on a promise the plant cannot keep, a delivery date set by a vector that does not see the queue in front of the machines, a credit term agreed by one chain and chased by another. The Keeper of chapter 8 settled exactly such a dispute, between Winning and Delivery, over a findability report. That dispute was a leak with two masters standing on either side of it.

A boundary between Outcomes is a cleaner leak than a boundary between departments, because the record shows it and the Keeper can rule on it in a day. But it is a leak, and the doctrine does not tolerate a leak it can close. It closes it the way it closes every leak: by putting both sides of the handoff inside one chain, where an agent carries the result across without a person on either bank.

## 3. The customer is the test

The reason tells you that boundaries will go. The test tells you which boundary goes next, and it is the test the whole book has used: what does the customer feel?

A customer of Pravaha does not feel an enquiry answered and then, separately, an order fulfilled. He feels one thing: he asked, he was promised, the pump arrived, he paid. Two Outcomes serving one feeling are a department by another name. The customer test says that wherever one customer’s experience crosses an Outcome boundary, the boundary is in the wrong place, and the Outcome nearest the customer’s last touch, the cash, should hold the whole of it. The test also says which boundaries hold for longest. The lender who funds the second assembly cell is a different customer from the plant owner who buys the pump, and feels a different result, so the capital Outcome has a reason to stand apart that the design Outcome never had. Boundaries between different customers’ results dissolve last. Boundaries inside one customer’s result dissolve first.

## 4. The vector is the mechanism

Reason and test say what should happen. The vector is how it happens, and it happens without anyone ordering it.

Chapter 14 showed a vector deepening: the same vector producing more each month as its agents compound, absorbing its neighbours because a vector that has learned one market learns the next faster than a fresh vector can. The same force works across an Outcome boundary. The deep winning vector’s pricing agent already reads the plant’s live capacity to set a delivery date; its quartermaster already knows which order is late and why. The day the fulfilling Outcome’s master looks at her floor and finds that the winning vector’s agents hold more of her chain’s state than her own vectors do is the day the boundary has already moved, and all that remains is for the Keeper to write it down. Nobody reorganises the company. The vectors grow through the boundary, and the boundary is found, afterwards, not to be there.

This is the difference between a merger and an absorption. A merger is a decision. An absorption is a reading. The Keeper does not decide that Winning and Delivery will become one Outcome; he reads, on a Monday, that one vector now produces the result both Outcomes were built for, and the Outcome that no longer produces anything of its own is, in the language of chapter 16, written on the wall.

## 5. The order at Pravaha

Pravaha ended chapter 13 with five Outcomes. Here is the order in which they became one, given by the test, with no dates, for a reason section 7 will give.

First, order-to-cash absorbed enquiry-to-order. The customer feels enquiry to cash as one result; the handoff between the two was the firm’s largest leak, a promise made by one chain and kept by another; and the deep winning vector of chapter 14 already held the fulfilment state. One Outcome, enquiry to cash, under one master. The other master became her backup, which is the standby rule of chapter 7 applied to a company that is running out of masters.

Second, the plant absorbed design. Design-to-first-unit had one customer, the plant, and produced one result for it, a first unit. When the plant’s own vectors, with their drawing and costing agents, produced the first unit of a new design faster than the design Outcome could hand one over, the design Outcome had no customer left. Written on the wall.

Third, enquiry-to-cash absorbed the plant. This is the one the reader will doubt, because the plant is where the hands are, and it is the one the customer test insists on most, because a pump on the customer’s floor is the promise kept, and lead time, which is the plant’s number, is the thing the customer feels most. By then the plant’s vectors held its machines, its cells and its robots as agents, and the chain from a quote to a unit in a crate was one chain with two masters standing in the middle of it for no reason the record could show.

Fourth, the Outcome that sold the plant’s spare capacity to other manufacturers merged into the one chain when the chain began to be lent, as chapter 14 describes, because a lent vector is the way spare capacity is sold.

Last, capital. Its customer, the lender, was a different customer until the record was complete enough that the lender read Pravaha’s floor directly and lent against what he saw. From that day a raise was not a result produced by an Outcome; it was a gate on the one chain, signed by the Keeper. The last boundary went where the first one had gone, into the record.

One Outcome. Inside it, over whatever time it took, the vectors merged as chapter 14 showed, each deep one absorbing its neighbour, until one vector carried enquiry to cash for every pump Pravaha made and every pump it made for the companies that borrowed its chain.

## 6. The six

What remains when one vector carries the whole company?

One aligner and one tuner, in the cockpit, holding the gates of the one chain. That much chapter 14 promised.

![Picture 13: The last six](https://vinaypasricha.com/assets/images/collapse-or-collapse/pictures/13-the-last-six-1600.webp)

*Picture 13. The last six: five Outcomes collapsing into one, then one Outcome into one vector; the six who remain, in two rooms.*

A master, still. The reader expects her to go, since a floor of one tile is not a floor. She stays, because the floor was never the point of her. The master owns the playbook, the written design every agent in the chain is built from; she is the one court for every dispute an inspector raises; she holds the Monday test for the whole Outcome and reads the gap between the design rate and the actual one. A one-vector Outcome has a playbook, a court and a Monday test exactly as a hundred-vector Outcome does. What changes is that her control room streamlines to one tile holding thousands of agents, and the rules of chapter 7 all still run.

A Keeper, still, because there are gating events. An Outcome is born only by his decision. The outside, the board, the law and the lender, has one face. The treasury is his. And every irreversible act the playbook assigns above the master’s limit carries his signature. None of that is produced by the chain, so none of it disappears when the chain becomes one.

And two backups, one for the master and one for the Keeper. Chapter 7 gave the master a standby in another master, and chapter 8 gave the Keeper a standby on the board. When there is one Outcome there is no other master, and a board member who steps into a one-vector company at a day’s notice has not read the floor. So each seat has a backup of its own, selected on intention like every seat, and the two seats that stood beside the Keeper in chapter 8 dissolve into one of them: the Head of Outcomes has nothing to queue when there is one Outcome, and the Keeper’s outside is his own again once there is nobody between him and the lender.

Six people. Not four: the master and the Keeper are not the same craft, and a company whose Keeper also holds the playbook has nobody above the designer of the gates to hear an appeal against them, and chapter 17 needs one. Not eight: there is nothing for an eighth to hold. Six is the smallest core the doctrine knows how to design, and rule 12 of this chapter asks for a written reason for a seventh. It is a possible end, not an inevitable one. A company whose customers feel several different results, or whose hands no machine can yet replace, or whose law requires a person where the doctrine would put an agent, may hold at two or three Outcomes for many years, and the doctrine counts that as a company still collapsing, not a company that has failed. Where the customers feel results that are truly separate, a buyer of pumps and a lender to the plant, say, the boundary between them is not a leak and the right end is more than one Outcome.

## 7. Two rooms, shared

The backups do not wait. A backup who shadows is a person outside the structure, and the tile would count her; a backup who runs the edge is away from the room when the room needs her. So the backups work alongside.

The master and her backup share the control room. Both read the one tile every morning; both see the inspector’s holds, the agent maker’s proposals, the design rate against the actual; either can take the appeal, and the playbook says which of them signs which day. The Keeper and his backup share the command centre the same way: four columns, one Outcome, the doctrine tile, and the queue of gating events, read by two people who could each hold the table alone by lunch. Nothing is handed over when one is away, because nothing was ever held by one of them alone. That is the whole of the company’s insurance, and it costs two salaries.

## 8. Hands, and why there is no timeline

The reader who runs a plant has been waiting for the catch, and here it is. A company of six that makes pumps has a plant with nobody in it, and a plant with nobody in it is a plant whose hands are machines.

The doctrine’s answer is the same answer it gives for every skill: a hand is an agent. A humanoid, a drone, a wheeled robot, a loader, an arm at a bench, any non-human automation, is an agent inside the vector that needs it, built and upgraded by the vector’s agent maker, checked by the vector’s inspector, gated where its act is irreversible, and shown on the tile as one more colour. A plant vector may hold thousands of them. There is no robotics department, no automation head beside the master, no separate rule for a machine that moves. An agent with hands is an agent.

But an agent with hands is bought, installed, commissioned and trusted at the pace at which such agents exist, and that pace is not the doctrine’s to set. The collapse of the office, the chain from enquiry to invoice, needs agents that already exist and a decision to use them, and chapter 13 gave it a year. The collapse of the hands needs machines that can do what the hands did, in that plant, for that product, at that tolerance, and whether that takes a year or a decade depends on what is being made and what the company can afford to buy. A software company collapses to six faster than a pump maker. A pump maker collapses to six faster than a shipyard. So this chapter gives an order and no date. Everything tends towards one, in the order the customer test gives, at the pace the hands allow. A company that announces the date has mistaken the doctrine for a plan.

## 9. Rules for the collapse into one

1. Every boundary between two Outcomes is a handoff, and every handoff is a leak the doctrine works to close.
2. The downstream Outcome absorbs the upstream one. Absorption runs towards the cash.
3. The customer’s felt result sets the order: boundaries inside one result dissolve first; boundaries between different customers’ results dissolve last.
4. Absorption is a reading, not a decision. The Keeper records that one vector now produces both results; he does not order it.
5. The master of the absorbed Outcome becomes the backup of the absorbing one, while there are masters to spare.
6. Inside the one Outcome, vectors merge as chapter 14 describes, until one remains.
7. The smallest possible core is six: aligner, tuner, master, Keeper, and a backup for each of the last two. A company may stop short of it; none goes below it.
8. The master stays for the playbook, the court and the Monday test. The Keeper stays for the gating events.
9. Backups work alongside, sharing the room, never shadowing and never away.
10. Any non-human automation is an agent inside a vector, under the vector’s rules. A plant vector may hold thousands.
11. No timeline. The order is fixed; the pace is set by the hands.
12. Six is the floor, not a target date. A seventh needs a written reason.

## 10. The last merger

The last Outcome to dissolve at Pravaha was capital, and the day it went was a Tuesday, because the lender’s reading of the floor had been agreed on the Monday and the Keeper would not write a death into the record on the same day as the test.

The capital Outcome had one vector, one master and one pair. Its master had held the enquiry-to-cash floor for two years before the raise became her Outcome, and she had been the one who argued, at the Monday test, that the lender now read the floor as well as she did. The Keeper wrote one line in the record: that the capital-to-balance-sheet result was now produced by the one chain and gated at his signature, and that the Outcome was closed. No meeting was called, because there was nobody to call to it.

Its master became the backup to the master of the one chain, and walked that morning into the control room she had once held. The pair of its vector were paid out as chapter 13 pays leavers, and the agents that had kept the lender’s covenants and the bank’s reporting went where the work went, into the chain’s accounts agent, which already knew the covenants because it had been producing the numbers that satisfied them. By noon the control room held two people and one tile, and the command centre held two people, four columns and a doctrine tile that read zero, zero, zero.

Six. A company that had been a thousand people in ten departments, and then ninety-eight people in five Outcomes, and then one Outcome with about forty-five vectors, and then one vector lent to two other pump makers, was now six people and a chain, with Meera Kulkarni still at its gates, and it made more pumps than it had ever made, for more customers than it had ever had, and nothing in it waited on anyone.

The reader may find the number frightening. It is meant to be read the other way. Six is not what is left of a company after the doctrine has finished with it. Six is what a company can be, once everything that was a workaround has gone; how close a given company comes to it is set by its customers, its hands and its law.

[^52]: R. H. Coase, “The Nature of the Firm,” *Economica* 4, no. 16 (November 1937): 386–405.

[^53]: Oliver E. Williamson, *Markets and Hierarchies: Analysis and Antitrust Implications* (Free Press, 1975), and *The Economic Institutions of Capitalism* (Free Press, 1985).