What is cyber-truss

cyber-truss makes a system self-converging: a change lands wherever it is natural to make it, strains the connections it crosses, and comes back carrying the criteria it must now meet.

What it does

Let’s use software development as an example.

Traditionally, there are two ways a change actually gets made.

Someone files a bug or asks for a feature, and the work starts at the top. A ticket, then a PRD, then a design doc, then a specification, and finally the code. You commit to the specification before you have touched the code, and what the implementation teaches you never travels back up.

Or you find the problem yourself in the code and fix it. What you leave behind is the PRD that still describes the old behaviour, the mockups nobody redrew, and the ticket that was never opened. The fix shipped and the system around it disagrees with itself.

Both paths end in the same place.

cyber-truss lets you make the change where you understand it and get the rest anyway. It can do that because those artifacts are parts of the system too, with the connections between them declared rather than remembered. That change strains the connections it crosses, because a connection states a relation between two artifact-sets that has to hold, and it no longer does.

Clearing the strain does not push your change outward from where it landed. The change is lifted out of line diffs into artifact-set vocabulary, and every workflow that spans the set you changed reads it: often several, and never the whole process by default. Each one works out the intent behind your change, separated from the particular expression you used, and asks the sets above yours what that intent now requires of them. They answer without seeing your change. The workflow replays from the highest set that has to move, back down to yours, and your change is reconciled against the criteria that arrive. A workflow that reads what you changed works from it directly.

So the criteria that come back to your change were not written from it. Sometimes they confirm what you wrote, and it stands as you wrote it. More often they add to it: a case the quick fix didn’t cover, or a rule the specification never stated until your change made it necessary. The round trip improves the change instead of only recording it.

The guarantee this buys is confluence. Whichever artifact you started from, the system settles into a state that meets the same criteria.

Where this goes

The loop above is set out in full, with the reasoning for the order of its steps, in Canonical execution. The three readings the design is worked out from are in The lattice.