[DSRP Evidence](https://dsrpevidence.org/)

# Context is the System – What AI Prompting and Systems Thinking Have in Common

## Details

**Authors** Prill, Patrick

**Year** 2026

**Publisher** Test Pappy

**Kind of work** press

**Discipline** Computer Science & AI

**Secondary disciplines** Systems Science

**Kind of engagement** public

**Applied** false

[Read it](https://testpappy.wordpress.com/2026/03/04/context-is-the-system-what-ai-prompting-and-systems-thinking-have-in-common/)

## In authors' words

### What they set out to do (purpose)

To argue that what makes an AI prompt useful is the same thing that makes an engineering decision sound — a description of the surrounding system — and to use the Cabreras' VMCL model to say what shared context human colleagues hold that an AI does not.

### How they did it (methods)

A blog post built on a worked contrast. Prill opens from an Elisabeth Hendrickson remark on context as the cumulative effect of past decisions that cannot now be changed, inverts it to ask what context is needed for good future decisions, runs a 'Context Scan,' and demonstrates with a pair of prompts: a vague one asking how to structure an authentication module, and one describing platform, team and constraints.

### What they found (results)

Prill concludes that missing context makes AI output generic and locally optimised, that front-loading a description of the system saves iteration time, that experienced engineers already gather context before deciding so AI use should follow the same practice, and that agents should be set up to request more context before starting. He maps repeated unhelpful correction cycles with an AI onto the systems-thinking 'fixes that fail' pattern. The support is reasoning plus one illustrative example, not data.

## Commentary

### In short

A software tester applies VMCL to the problem of AI prompting, using it to characterise the shared organizational context a human colleague holds and a model does not, and builds his prompt checklist from boundaries, elements, relations and perspectives.

### In more detail

Prill uses 'the Vision, Mission, Capacity, Learning model from Drs. Cabrera' to name what a colleague inside an organization already knows and an AI does not: the vision it is working toward, the mission it repeats, the capacity it has, and what it has learned. That asymmetry is his explanation for why a prompt that would be enough for a teammate is not enough for a model. He then gives the dimensions of context worth supplying in a technical prompt as boundaries, elements, relations and perspectives — the four patterns in substance, though he does not name DSRP in this post. The demonstration is a before-and-after pair: 'How should I structure my Auth module?' against a prompt that states the platform, the team and the constraints, with the second yielding a far more relevant answer. He argues that a component cannot be understood apart from the system around it, and reads the loop of repeated AI corrections as the 'fixes that fail' archetype, where each local repair leaves the underlying absence of context in place.

### Response

The use of VMCL is accurate and does real work here: the model is applied as an account of what an organization holds in common, which is what it is for, rather than as a planning template. Prill's prompt checklist is the four patterns arriving under other names — boundaries for distinctions, elements for part-whole, relations for relationships, and perspectives as given — which is a reasonable independent route to the same four and worth noting as such. Because he does not name DSRP in this post, the connection between his checklist and the framework he names in his other writing is left for the reader to make.

**Patterns it shows** D, S, R, P

**Added** 2026-10-19

**How to cite this** Prill, Patrick (2026). Context is the System – What AI Prompting and Systems Thinking Have in Common. Test Pappy.