A practical, opinionated guide to designing backend APIs. Covers request and response schema, offset and keyset pagination, BFF, server-driven UI, caching, versioning, naming, the new HTTP QUERY verb, WebSocket vs SSE, API security (auth headers, server-to-server auth, webhook receivers), and bulk data load APIs. Built around Relay, the same AI proxy system from the multi-tenant SaaS post, with REST and JSON as the running example and Go code throughout.

How to use cache to prevent your database from getting hammered. Understanding cache mechanics, types, strategies, and when each one wins.

A full system design of a production rate limiter. Covers every major algorithm (token bucket, sliding window, leaky bucket, fixed window), their tradeoffs, how they break in a distributed system, how Redis fixes them, and how Stripe, Cloudflare, and API gateways actually deploy this in production.

A complete system design of Relay, a multi-tenant AI API gateway. Covers multi-tenant database architecture (silo vs pool vs bridge), API key authentication, provider routing with failover, dual-layer rate limiting, token-based billing, response caching, row-level security, tenant provisioning, and observability. Full Postgres schemas, mermaid diagrams, and Go snippets included.

RAM is up 400%. CPUs cost 15% more. The free hardware lunch is finally over. Welcome back to the 80s, where every byte counts.

Building a thread-safe inverted index from scratch in Go. Covers sharded mutexes, lock contention profiling, slice pooling to avoid GC pressure, and benchmark comparisons against a naive sync.RWMutex approach under varying read/write ratios.

Why reordering two struct fields made my loop 30% faster, and what it taught me about how CPUs actually work.

How to connect ADB over Wi-Fi using a one-time USB handshake, even when a VPN or firewall blocks the direct connection.

"It's just a simple query". Why the distance between a logical requirement and its infrastructure cost is the most expensive gap in engineering.

A breakdown of the absolute absurdity of human time, monotonic clocks in Go, and the one true way to store time across PostgreSQL, Couchbase, and mobile clients.