Gonçalo Esteves / capable under pressure

The Vault / Methods

Method

Model-based systems engineering, applied to a syllabus

treat a body of knowledge as a system with declared dependencies rather than a document — so that changing one part tells you every other part it touches.

status · drafted added 2 Sept 2026 organizationtraining

What it is: treat a body of knowledge as a system with declared dependencies rather than a document — so that changing one part tells you every other part it touches.

When to reach for it

When a change to one piece of a curriculum, a process or a product quietly breaks three others and nobody finds out for six months. Reach for it when the thing is written down but not modelled.

The steps

  1. Name the elements. Techniques, drills, prerequisites, assessments — each an element with an id.
  2. Declare the dependencies. What must be true before this element can be taught?
  3. Declare the verification. How do you know a student has it? An assessment is a requirement, not an opinion.
  4. Change one thing and trace the graph. His own words: "the same way that you would introduce a change to a car — where are the parts that it touches? What needs to be done? Then we need to verify it."
  5. Make it queryable. "Why isn't my curriculum agentic in the sense that everyone can consult it, can talk with it, can view it?"

Worked example

  • Not yet applied. Register idea 284-01, tagged NAME ON ITthe ideas register.
  • ⚠️ This is a build, and the build slot is occupied by Do Not Die On The MatACTIVE. The cheap experiment that is NOT a build: take one belt level, model it as a system with dependencies, then ask one instructor a question the document should be able to answer. An evening. If it takes a month, it is a project and it needs the slot deliberately.

Where it fails

  • ⚠️ The taxonomy trap, and it is the central risk here. "Everybody wrote martial arts down; nobody wrote it down the same way." Same named kick, three schools — one thrusts, one pulls, one snaps, all three correct. A model that cannot represent legitimate variation is one lineage asserting itself as ground truth. Build variance into the schema or do not build it. See Embodied AI.
  • Modelling cost is real. MBSE pays off on systems that change often and break expensively. A syllabus that changes twice a year may not clear that bar.
  • A model nobody queries is a document with extra steps. Step 5 is the point; steps 1–4 without it are overhead.

Where I learned it

Aerospace. It is how you change one requirement on a vehicle without discovering in integration that you broke four others.

Reading rule. Anything marked unverified or on-air is a claim made on the podcast and has not been checked against a primary source yet. It stays marked until it has. Sourced numbers link to where they came from.