Architect
Most architecture diagrams are lies. This track teaches you to create documentation someone could build from—without asking you clarifying questions.
Demonstrated, not asserted
- Blueprint install The document pipeline this track teaches, shipped as an npm CLI whose output is exactly this artifact class — arc42-grade architecture docs, ADRs, and validated handoffs — with its live self-demo as a worked example.
npx @nino-chavez-labs/blueprint-cli initverified 2026-07-15
Who this is for
This path is for you if
- ·You're a tech lead or senior engineer who documents systems
- ·You've built systems but struggled to communicate their design
- ·You want documentation that developers actually use
- ·You're tired of explaining the same architecture decisions repeatedly
Probably not if
- ·You haven't built production systems yet (see Builder Track)
- ·You want to write persuasive content for executives (see Strategist Track)
- ·You're exploring AI for self-understanding (see Explorer Track)
- ·You need to build AI apps, not document them
The Voice Shift
Architecture documentation requires a different voice than thought leadership. Thought leadership is exploratory — "Here's where I've landed—for now..." Architecture is definitive — "The system uses PostgreSQL for X because Y." Architecture docs are reference material. Someone should build from them without asking you questions.
The levels
0 Learn the Templates 1-2 days
Understand arc42, C4, and ADRs before you start filling them.
Terms that matter
- Context Diagram — your system as a box plus external actors and systems
- Container — a separately deployable unit (API, database, web app)
- Component — a logical grouping within a container (auth module, payment service)
- ADR — Architecture Decision Record: documents why you chose X over Y
Read
- arc42 — the full template with section explanations — arc42.org
- C4 Model — four levels of abstraction for diagrams — c4model.com
- ADR templates — the Architecture Decision Record format — adr.github.io
Checkpoints
- ·You can explain Context vs Container vs Component diagrams
- ·You understand why ADRs include "Alternatives Considered"
- ·You know what makes documentation "implementation-ready"
Done when — You can explain the difference between Context, Container, and Component diagrams
1 Document an Existing System 1-2 weeks
Create your first architecture document for a system you've already built.
System requirements
- ·Pick a system with at least 2 containers (frontend + API, or API + database)
- ·Must have at least 1 external integration
- ·Must have at least 3 decisions you remember making
Create these artifacts
- Context Diagram — your system as a single box, all users and actors, all external systems
- Container Diagram — break the system into deployable parts with technology labels
- 3 ADRs — Status, Context, Decision, Rationale, Consequences, Alternatives
Tools
- Excalidraw — quick sketches, hand-drawn feel
- D2 — diagram-as-code, version control
- Mermaid — embedded in markdown
Checkpoints
- ·You have a Context diagram showing system boundaries
- ·You have a Container diagram showing internal structure
- ·You have 3 ADRs with rationale and alternatives
- ·All artifacts are in a single markdown file or repo
Done when — You have a Context diagram, Container diagram, and 3 ADRs
2 Capture Decisions as You Make Them 2-3 weeks
Write ADRs before or during decisions, not after.
The ADR workflow
- Before deciding — write the Context section: what problem, what constraints
- While deciding — write Alternatives: 2-3 real options with trade-offs
- After deciding — write Decision, Rationale, Consequences
Write 5 ADRs — include at least
- ·1 technology selection (database, framework, service)
- ·1 architectural pattern (monolith vs microservices, sync vs async)
- ·1 integration approach (REST vs GraphQL, polling vs webhooks)
- ·1 security/compliance decision
- ·1 "we decided NOT to do X" (scope/deferral decisions count)
Checkpoints
- ·You have 5 ADRs written during active work (not historical)
- ·Each ADR includes a trade-off table
- ·You have a sense of when a decision is "significant" enough to document
Done when — You have 5 ADRs written during active decision-making
3 Complete arc42 Document 3-4 weeks
Create a full arc42 solution architecture for a real system.
The 12 arc42 sections
- 1. Introduction and Goals — what, why, who, quality goals
- 2. Constraints — technical, organizational, regulatory
- 3. Context and Scope — context diagram, business and technical context
- 4. Solution Strategy — key approaches, technology decisions
- 5. Building Block View — Container and Component diagrams
- 6. Runtime View — sequence diagrams for key scenarios
- 7. Deployment View — infrastructure, environments, process
- 8. Cross-cutting Concepts — auth, errors, logging, security
- 9. Architecture Decisions — all ADRs with decision log
- 10. Quality Requirements — quality tree and scenarios
- 11. Risks and Technical Debt — known risks, mitigation, debt backlog
- 12. Glossary — domain and technical terms
Checkpoints
- ·Complete arc42 document with all 12 sections
- ·At least 5 diagrams (Context, Container, Component, Runtime, Deployment)
- ·At least 5 ADRs
- ·Document is implementation-ready — someone could build from it
Done when — You have a complete arc42 document someone could build from
4 Cross-System Integration Docs 4+ weeks
Document how your system integrates with others you don't control.
Document these
- System-of-Systems Context — your system as one box among many, dependencies in both directions
- Integration Patterns — sync vs async, data flow direction, contract ownership
- Failure Modes — what happens when each dependency fails? Retries, circuit breakers, fallbacks
- Data Consistency — source of truth per data type, eventual consistency windows
- Integration ADRs — why this approach? What contracts negotiated? What compromises made?
Worked patterns
- PIM to AI enrichment — event-driven: product updates trigger the enrichment pipeline; the PIM stays the source of truth while AI writes back enhanced attributes
- Search platform to LLM — hybrid: traditional search for recall, an LLM for semantic reranking, with a circuit breaker when LLM latency crosses a threshold
- CMS to personalization — async: publishing content fires events, the personalization engine indexes them, and generic content is the fallback when the engine is unavailable
- Analytics to optimization — batch daily aggregation feeds the optimization models while real-time click data drives immediate reranking
Checkpoints
- ·System-of-systems context diagram
- ·Integration pattern documentation for 3+ external systems
- ·Failure mode analysis for each integration
- ·Integration-specific ADRs
Done when — You have system-of-systems documentation with failure modes
Tool links and level references live in a dated data file checked by the link sensor on every build — this page can't silently point at a moved doc or a dead demo.