Stage 2 — Advanced Solution Engineering
Stage 2 takes you from building individual features to engineering a whole solution for real people. After Stage 1, you deepen a specialisation and take responsibility for how its technical, human and business parts work together.
AI still writes code. Your work includes choosing the problem, judging the design, checking difficult failures and explaining why the solution deserves to be used.
This page shows the curriculum direction, the learning structure and the method. It is an outline; Stage 2's lessons are not available yet.
The areas you can take deeper
| Area | What you will learn to judge |
|---|---|
| Software engineering | Architecture, security, reliability and how a system grows; technical leadership and decisions that affect the whole solution |
| Agent engineering | Agents that retrieve evidence, use tools and coordinate work; meaningful evaluation, safety and difficult failure cases |
| Using AI at an expert level | Larger coding-agent workflows, clear delegation, research and independent checks on work you commission |
| Web development | Complex interfaces, accessibility, performance and the design choices behind dependable web applications |
| DevOps and platform engineering | Operating larger systems and providing safe tools that other people use to build and run their services |
Communication and business judgment run through the work. You practise understanding a person's problem, explaining trade-offs, presenting evidence and making a useful offer. Communication skills, body language and marketing belong here alongside the engineering.
These areas describe the scope. They do not set a final list of course titles or chapters. You follow one course at a time, bringing forward the foundations it needs.
How the learning is organised
Stage → Course → Module → Chapter → Lesson → optional Parts. Each module combines its main subject with the supporting concepts and the places it is applied. Common foundations come before specialised uses; prerequisites still come first.
Earlier material is retrieved and deepened rather than taught again from the beginning. For example, a larger agent system still depends on the Python, data, permissions and testing you learned earlier. What changes is the scale of the decisions and the strength of evidence you need.
The depth scale continues towards 2.5–3.0 where the chosen specialisation needs it: designing, teaching the reasoning and handling difficult cases. The depth of each topic depends on the decisions you must make and the work AI can reliably do.
DevOps and platform engineering
The operations specialisation extends Stage 1 Course 4. Its current module outline is:
| Module | What the work must establish |
|---|---|
| Engineer a Kubernetes runtime | A suitable place to run workloads, with controlled networking, storage and access |
| Reconcile and progressively deliver releases | Keep the running system aligned with a reviewed configuration and introduce changes in controlled steps |
| Engineer cloud infrastructure and its lifecycle | Explain how resources are created, changed, secured and eventually removed |
| Engineer reliability from telemetry and incidents | Use measurements and failure records to set useful reliability targets and improve recovery |
| Offer a safe self-service platform | Let other people operate within clear limits, with safe defaults and bounded automation |
| Defend an operated solution end to end | Demonstrate delivery, failure handling, recovery and a handover supported by evidence |
This outline ranges from 2.5 to 3.0, depending on the module. It is a specialisation, not a tool list every learner must install. The names and groupings may develop as the lessons are written.
The method: Spec-Driven Engineering
Stage 1 uses specifications to guide development. Stage 2 uses Spec-Driven Engineering for the whole solution: the problem, the people affected, the design, the implementation and its continued operation.
You begin with a named person's real problem. You define a useful result and its limits, weigh alternative designs, then direct the implementation. You test both normal behaviour and the failures that matter. Finally, you explain the evidence and decide whether the result is ready for people to rely on.
AI can investigate, propose designs, build, test and help operate the software. You still decide whose problem matters, what risks are acceptable and whether the evidence is sufficient. You also own the cost, the communication and the promises you make.
Both stages aim at work suitable for real users. Stage 2 asks you to justify it under harder conditions: more traffic, partial failures, hostile inputs and changing requirements.
Where to begin today
The book asks for four hours a day, with one lesson followed by practice. A later specialisation assumes the relevant earlier foundations; it does not assume you arrived as an expert.
While this stage is being developed, begin with Stage 0 if you are new to software, or use the Stage 1 overview to find the available lessons.