System Design & Architecture

Thinking like an architect#

The mindset shelf — less “how to draw boxes,” more “what the job actually is.”

• Fowler’s Software Architecture Guide — the front door to the tag’s most-saved source.
• Thinking like an architect — “The architect’s job isn’t to be right. It’s to reduce uncertainty. Treat architecture as a series of experiments.”
• Architecture is not a blueprint — it’s a set of decisions — pairs perfectly with Fowler on Architecture Decision Records, the log of that decision history.
• The Expert Generalist (Fowler & Unmesh Joshi) — the strongest argument for your layers-first, cross-specialty trajectory.
• Butler Lampson: Hints and Principles for Computer System Design — the talk you watched in one sitting (“watched this great talk last night”).
• Simon Brown: five things every developer should know about software architecture — the keynote slides.
• Rik Hepworth: Azure Architecture — Choosing Wisely — “architecture is a team sport.”
• Is High Quality Software Worth the Cost? and Technical Debt — the two Fowler essays you keep re-saving for stakeholder conversations.
• the Solution Architecture Document, with a lean template — the enterprise-architect artifact, minus the ceremony.

Architecting AI & agentic systems#

The tag’s new center of gravity, and the section that matters most for where you’re headed. The recurring tension in your saves: deterministic systems now have to coexist with non-deterministic ones.

• The oil-and-water moment — the Architect’s V Impact Canvas for mixing deterministic and non-deterministic components; the cleanest framing of the whole problem.
• Agentic AI Architecture Framework for Enterprises — InfoQ’s three-tier (Foundation → Workflow → Autonomous) enterprise evolution path.
• the definitive blueprint for enterprise agentic AI architecture — the end-to-end reference.
• Reference architectures from the clouds: Google Cloud’s multi-tenant agentic AI system · choosing agentic AI architecture components · the multi-agent AI system reference · architecting for agentic AI on AWS — and the Google advocate’s caution about single-agent-first: “Simmer down. One is probably fine.”
• Patterns: Agentic Design Patterns — the free Google ebook · awesome-agentic-patterns, the curated catalogue · Guillaume Laforge’s agentic design patterns slides · 8 multi-agent patterns in ADK · all-agentic-architectures — 17+ with runnable code.
• Event-driven agents (your Kafka-brain applied to agents): Sean Falconer: the future of AI agents is event-driven · Kai Waehner: Kafka + MCP + A2A · event-driven multi-agents with Kafka, Flink, and LangChain · Bindu: turn any agent into a living microservice.
• Production shape: the 7-layer production-grade agentic system · a Foundry agent harness: tool governance via Azure API gateway · architecting conversational observability for cloud apps — the last two sit exactly on top of your agentic incident-management work.
• Case studies: Grab’s multi-agent engineering support system · Google’s scaling principles for agentic architectures · how Perplexity built an AI Google · the architecture behind Lovable and Bolt.
• The still-true 2024 foundations: what is a cognitive architecture and own your cognitive architecture, outsource the agent infrastructure.
• The contrarian shelf: multi-agent AI is the new microservices — and that’s a warning · “agentic systems get the hype, but distributed systems are the future of AI” · local-first inference escalation with confidence gates.
• The enterprise-architecture profession itself: Gartner: heads of EA must rewire EA for AI.

The martinfowler.com subscription#

Your steadiest source — and since 2025 it reads like a serialized book on GenAI-era architecture. The through-line: classic engineering discipline applied to non-deterministic systems.

• the GenAI patterns series (Fowler & Bharani Subramaniam) — starts with evals, for good reason.
• agentic AI security — the primer on what agents do to your threat model.
• harness engineering — Birgitta Böckeler’s mental model and sensors for coding agents.
• patterns for reducing friction in AI-assisted development — “AI assistants are like junior developers with infinite energy but zero context.”
• Legacy modernization with AI, a quiet superpower for an enterprise architect: legacy modernization with GenAI and black box to blueprint — reverse-engineering lost source code.

Distributed systems — the fundamentals#

The layer that doesn’t expire. Everything in the agentic section eventually falls back on this shelf.

• MIT 6.824 — the distributed systems course; readings and lecture videos.
• Gossip Glomers — Kyle Kingsbury’s distributed systems challenges; learning by breaking things.
• Lindsey Kuper’s CSE138 lectures — “the ideal blend of approachable introduction and academic rigor.”
• an illustrated proof of the CAP theorem · the distributed systems reading list.
• Designing Distributed Systems — Brendan Burns’s hands-on labs — the 2019 save that started this thread.
• the map of system topologies — “I love this kind of stuff,” you wrote. Whatever you’re building probably fits one of these.
• message queues explained with analogies · Bernd Ruecker on long-running processes in modern architectures.
• debugging event-driven systems — five self-inflicted problems — dead-letter queues, eventual consistency, and other 2 a.m. friends.
• “I used to think distributed systems were hard; then I learned about billing systems” — the humbling counterpoint.

Microservices, monoliths & the pattern shelf#

The DevOps-era backbone of this tag — and notably, your saves aged into skepticism before the industry did.

• Death by a Thousand Microservices — the polemic.
• Adrian Cockcroft on the Prime Video microservices-to-monolith story — the sober read on 2023’s noisiest architecture debate.
• microservices lessons from Netflix · Airbnb’s journey to post-microservices.
• modular monolith with DDD and evolutionary architecture by example — the two repos you’d actually build from.
• DDD by examples — the library project — domain-driven design as working code, newly relevant as a shared language for coding agents.
• the RIG model for designing sagas — reversible, irreversible, guaranteed.
• Design Patterns That Deliver · design patterns for humans — the patterns refresher, two speeds.
• Adam Tornhill’s clear software design principles — recurring patterns from building both AI-native products and mature codebases.

Scalability case studies#

The InfoQ thread — real systems, real failure modes, saved steadily for a decade.

• InfoQ’s Architectures You’ve Always Wondered About — Netflix, AWS, Pinterest, Cloudflare in one minibook.
• Dropbox Chrono — scalable, consistent metadata caching.
• Uber’s chat rebuild on GraphQL subscriptions · Zalando’s transformer-based search and ranking platform.
• DoorDash Aperture — failure mitigation for microservices — retry storms, death spirals, metastable failures: the vocabulary of cascading failure.
• Monzo’s lessons from Cassandra at scale · London Stock Exchange Group’s chaos engineering on AWS.
• the InfoQ Architecture and Design Trends Report 2025 — where the industry thinks it’s going.

Diagrams & architecture-as-code#

A surprisingly deep sub-thread — you collect diagramming tools the way some people collect fonts, and lately the tools draw themselves.

• The classics: the C4 model, explained · Luca Mezzalira on documenting architecture — ADRs, RFCs, event storming, C4 · architecture-as-code with Structurizr.
• Diagrams from code: Diagrams — diagram-as-code in Python · LikeC4 — live diagrams from code · Terravision — architecture diagrams straight from Terraform.
• The AI wave: archify — polished diagrams from plain English · Eraser’s AI architecture diagram generator · codeflow — GitHub URLs into interactive architecture maps · comparing LLMs for creating architecture diagrams.
• Interactive design canvases: paperdraw — design distributed systems in the browser · layerd — layer-based 2D system design.
• the architecture diagramming tools list — your own “[bookmark for later],” now bookmarked here.

The interview shelf#

You’re past the interview-grind stage, but this shelf doubles as a fundamentals curriculum — and the newest addition is aimed exactly at your next role.

• The AI Engineer’s System Design Interview Guide — where the classic interview meets the new job description.
• the System Design Primer — the canonical repo, re-saved in 2026; still the starting point.
• best-system-design-resources · Complete System Design — 100+ case studies · awesome-software-architecture.
• 30 blogs to learn 30 system design concepts — scalability, availability, CAP, ACID, one blog each.
• Worked designs: the URL shortener · scaling an app to 10 million users on AWS.
• The reading meta-layer: 75 engineering blogs worth reading for system design · the software architecture books list · top 10 of 40+ Udemy architecture courses, road-tested.

All links on this topic#