Engineering transformation · Software-driven practices · AI

Engineering must evolve.
So must how we engineer.

Markets expect shorter development cycles, more frequent improvements and higher quality. Software-inspired practices are spreading across electronics, physical products and systems engineering; AI introduces new ways to explore, implement and evaluate changes. This lab examines how we can make engineering faster and more dependable without overlooking whole-product evidence.

We build the machine that builds the machine: improve the engineering environment alongside the product.

Develop the system.
Develop how we engineer it.

Conceptual approach

Product system

Software · electronics
Physical design · AI

Understand use

Model
and experiment

Integrate
and validate

New product ambitions
change the engineering work

Decide what to test
and what to build next.

New engineering capabilities
expand what we can try

Engineering
environment

People · models
Tools · AI agents

Develop methods and models

Work with
AI agents

Automate
and evaluate

Executable engineering

Executable models make behavior testable. Automated checks give AI agents actionable feedback: propose a change, execute it, inspect the result and improve it within agreed boundaries. Engineers determine whether the model and its checks are suitable for the product question.

We build the machine that builds the machine.

The recurring engineering challenge starts in Chapter 2.

Conceptual workflow

A fictional city considers improving a four-way intersection. First we establish stakeholders, system boundaries and uncertainty; later we use a bounded fixed-phase model to test one specific question. Read the introduction to the example.

Define + model

Frame the whole-system problem.

Clarify the needs of drivers, pedestrians, operators and maintainers before selecting a preferred solution or model.

Execute + evaluate

Investigate with bounded models.

Use controlled scenarios and tests to explore alternatives. AI agents may help refine bounded changes where independent automated checks provide meaningful feedback.

Interpret + decide

Integrate, verify and release responsibly.

Connect software evidence with electronics, physical interfaces, operational needs and release authority. An executable model is not a qualified controller.

The example is fictional. Fast automated feedback supports learning and agentic iteration; whole-product evidence determines which results are actually relevant to integration and release.

01 / Modelling and exploration

Make ideas
executable.

Use software models, simulations and physical prototypes to examine specific engineering questions. Deterministic calculations have predictable results for defined inputs; experiments reveal what changes when conditions or designs change.

02 / Teams and workflows

Engineer the
feedback loop.

Design automated behavioral and integration checks that tell agents what failed and why. Let them refine bounded changes, while preserving independent criteria and making demonstrated behavior—not generated output—the measure of progress.

03 / Release and feedback

Check against
reality.

Automated tests support fast inner loops. Product validity, physical interfaces, operational conditions and release authority form a broader loop with different evidence and cadences. A green test suite cannot answer every product question.

The course · Fifteen chapters

Begin with the changing engineering paradigm, then follow one fictional traffic intersection from uncertain intent to an integrated product and the engineering capabilities needed to sustain it.

All fifteen chapters are available below following initial author review. The examples are fictional teaching cases; interactive V3 graph and model tools remain in development. Earlier lessons remain available separately.

I — Frame the Work

II — Build and Test Knowledge

III — Develop an Integrated Product

IV — Establish Trust and Sustain Change

Earlier published lessons