A concept, explained

Essential complexity.

On the model that outlives the tools — from business architecture to code.

Robert Blust · Software Engineer & Architect

rests onexperience Essential Complexity

My through-line

From business architecture to code — end-to-end traceability.

Four areas shaped my career: software development, methodology, IT architecture and business architecture. I know all four.

The goal is always end-to-end continuity — from the business domain to implementation. Only the path there was expensive and hard for a long time.

rests onskill Software engineering skill Software process engineering skill Enterprise architecture skill Business architecture

The standard

A model captures a problem in its essential complexity — no more.

Capture the essential, leave out the unimportant. That is the real craft.

My standard for every model, for about 15 years.

rests onskill Model-driven engineering experience Eclipse Modeling Platform Working Group

Once hard — now easier

Before

UML · SysML · your own DSL

Powerful — but hard, specialized, expensive. Few can read it.

Today

AI understands natural language.

With the right meta-model: the facts as Markdown — cheap to create and maintain, readable for everyone.

Same precision — far less friction.

rests onskill Software modeling (UML, SysML, C4)skill Domain-specific language design

My approach

The Mental Model.

A structured knowledge base — I call it the Mental Model.

Domains & termsBooking · Finance · Door Access · Task
How we workRoles · teams · processes with gates
Why this wayArchitecture decisions · rules
How we measureKPIs

Facts as Markdown · structured by a meta-model · machine-readable

rests onconcept Mental Model

An example — Finance in a hotel system

Real model from LIKE MAGIC — Markdown + Mermaid

Folio: Financial account of a reservation — bundles all line items of a stay.
OrderItem: A single line item on the folio.
Payment: Payment that settles order items fully or partly.
Debitor: Who owes the bill — person or company.
PdfInvoice: The finalized invoice at check-out.

Every term in one sentence — no more, no less.

rests onexperience Co-Founder & Head of Technology skill Domain-driven design

The payoff

One model that everything can be measured against.

Requirements use only terms from the glossary — one language, no synonyms.

Implementation validated against glossary + ADRs.

New terms must land in the glossary first — otherwise no merge.

Not documentation but an artifact to validate against — so that change gets cheaper and safer.

rests onskill Requirements engineering skill Domain-driven design

Why it matters now

The right context makes AI far more reliable.

A model like the Mental Model gives AI the right context: it does not eliminate hallucinations — but it shrinks the space where AI has to guess, moving it from guessing to grounded reasoning. The cleaner the structure, the better the context. In practice: the Mental Model as a context layer for AI assistants.

rests onskill Context engineering kpi Grounded Answer Rate

Where it is heading

The pendulum swings — back to spec-driven development.

From “no big design up front” to spec-first — because AI makes specifications fast and affordable.

For years upfront modeling was too expensive, so agile deliberately dropped it. That math has changed.

rests onskill Spec-driven development phase Spec

Takeaway

Tools come and go — the model stays.

The standard stays: capture essential complexity. Only now the path is open to everyone.

That is how I still build today: close to the code, but with the model in mind. Thank you.

rests onvision One model, true everywhere