Extend and Improve: The Design You Already Have Is the Job
New code is cheap to draft. The lasting move is older: add to a design someone still understands, or tidy that design until the next change is easy.
Engineering Manager / Technical Lead
A topic hub collecting every article tagged System Design. Use it to explore related posts and follow this theme across the site.
12 articles
New code is cheap to draft. The lasting move is older: add to a design someone still understands, or tidy that design until the next change is easy.
AI-native apps are not ChatGPT wrappers. They are systems where model inference, retrieval, memory, orchestration, tools, and evaluation are core infrastructure. Here is a working definition, the reference architecture, and a test you can apply to your own product.
A deep systems engineering guide to the essential algorithms behind modern infrastructure — matching markets, rolling hashes, network flows, cache replacement, probabilistic sketches, and vector search.
A practical Agno-only blueprint using Workflow, native Agents, session state, bounded repair loops, persisted HITL review, and isolated coding workspaces.
A practical LangGraph-centered blueprint using typed graph state, dynamic Send fan-out, PostgreSQL checkpoints, deterministic candidate testing, interrupts, and isolated coding sandboxes.
A practical tour of Core NATS — how subjects and wildcards route messages, why delivery is at-most-once, and how request/reply and queue groups replace load balancers for many microservice patterns.
How a generic order domain flows from a relational write database through NATS KV and index subjects into OpenSearch and an analytics warehouse — without mistaking plumbing for source of record.
JetStream KV looks like a key-value database, but it is a stream abstraction with a latest-value view — watch for changes, design keys as hierarchies, and treat entity types as key prefixes.
JetStream adds durable streams and consumer delivery state so publishers and subscribers can be temporally decoupled — capture, replay, and redeliver until acknowledged.
An honest comparison of delivery semantics, persistence, ops burden, fan-out, replay, portability, and cost — with symptom-based selection tables instead of benchmark wars.
A practical ladder for growing a software system from one app and one database into observable, event-driven, permission-aware, AI-ready architecture.
AI is changing the center of gravity in software engineering: less manual typing, more specification, validation, system design, and human judgment.
Quick find
Search by topic, title, framework, or pattern.
Loading search…
Esta página también está disponible en español. Leer en español