What Is SSOT (Single Source of Truth)?
The previous chapter covered the document-first principle (the idea of keeping written records as the basis for every decision). The core message was simple: leaving a record matters.
But when records keep accumulating, a different problem eventually appears. The records scatter across multiple places, and it becomes unclear which one is correct.
SSOT is the answer to that problem.
Q. What does SSOT stand for?
SSOT stands for Single Source of Truth. In practical terms, it means deciding that one specific location holds the authoritative version of a given piece of information — and that location alone.
"Only one place is correct." That is all it takes.
Imagine a rule is written in three separate files. At the start, all three say the same thing. Over time, one gets updated while another is left as it was. A third gets rewritten somewhere as a quick note.
Once things reach that state, nobody knows which file to look at. Different people reference different files, and contradictions arise. When someone points out the discrepancy, the response is "I was reading this one," and the situation becomes unmanageable.
SSOT is the practice designed to prevent that kind of contradiction.
Q. How does it work in practice?
There are two things to do.
First, choose one location and declare it the authoritative source.
Second, when referencing that information from anywhere else, do not place a copy there. Copies inevitably drift from the original. Instead, record only where the authoritative source lives and build a structure that goes there to look.
To use a concrete example: the series you are reading now has its design rules spread across several files. One of those files is the authoritative chapter list — the single place that manages chapter titles and their order.
Whether a chapter is being written or the publication schedule is being adjusted, "which chapter comes next" is determined by consulting that one file. Nothing is copied out to a separate spreadsheet or transcribed into a side note. By routing all references to a single location, there is no need to wonder which version is current.
Q. Does consolidating everything in one place make things inconvenient?
There are situations where consolidating to one place can feel like more work to maintain.
However, keeping copies in multiple places means every copy must be updated each time something changes. Missing even one creates an immediate discrepancy. Comparing "the effort of keeping everything in sync" against "the effort of updating a single location" shows that the latter is lighter over the long run.
There is another benefit. When everyone knows exactly where to look, the system becomes easier to understand for anyone who joins later. Being able to say "just read this file" makes explanations shorter and reduces the load on whoever is doing the explaining.
Q. How does this connect to AI organization?
Running multiple AI agents means information is passed around constantly. One agent generates something; another agent references it. If that reference points to multiple locations, each agent ends up operating from different information.
When a human can notice and correct the discrepancy, things may still be recoverable. But the more automated the system becomes, the more contradictions accumulate without anyone catching them.
That is why the information agents reference needs a designated authoritative location. This is especially important for rules, design principles, and constraints — anything that serves as the basis for a judgment call. Keeping those in one place is essential.
In the system described throughout this series, each agent has a defined rule file to read at startup. Every agent reads from the same location. That is SSOT in implementation.
Q. What is the difference between SSOT and the document-first principle?
The document-first principle is about behavior: recording the reasoning behind a decision as text. It is a habit — the practice of writing down why a judgment was made.
SSOT is about structure: deciding where to look. Even if records exist in various places, SSOT designates one as authoritative and routes all references there. It is a design decision.
The two address different things, but combining them makes both stronger.
Keeping records alone does not tell you which one is authoritative. Designating an authoritative source without putting records in it leaves that source empty. "Building up records through the document-first principle, then routing all references to those records through SSOT" — when both are in place, it becomes possible to trace who made what decision and why, while keeping contradictions from taking hold.
The first half of Part 2 has worked through the principles and concepts that form the foundation of AI organization: the separation of powers (here meaning: dividing execution / audit / approval between separate agents) as a structural starting point, the two-person rule (requiring two independent parties to confirm irreversible actions), how to handle irreversible actions, the document-first principle, and SSOT. Each of these works on its own, but combined they create a system where it is clear who did what and why each decision was made.
Before moving into system design, having a map of these terms and ideas makes the implementation chapters that follow considerably easier to follow.