Notes for engineering teams that need deterministic guardrails around AI-generated code, structural rules, and pull request workflows.
We built the same TypeScript API in Express 5 and Hono 4, benchmarked both on Node.js 22, and compared routing speed, TypeScript DX, middleware ecosystem, dependencies, and runtime portability. The results are not what the benchmark headlines suggest.
aws-lambda
You are starting a new TypeScript API. Express 5 shipped as stable in early 2025 and the project is actively maintained ...
We built the same TypeScript API in Express 5 and Hono 4, benchmarked both on Node.js 22, and compared routing speed, TypeScript DX, middleware ecosystem, dependencies, and runtime portability. The results are not what the benchmark headlines suggest.
Compare @supabase/server, getClaims(), Hono's jwk middleware, and jose for verifying Supabase JWTs in a Hono API. Covers CORS setup, RLS preservation, common failures, and a production checklist.
Step-by-step guide to migrating a React SPA or Vite app from Cloudflare Pages to Workers: wrangler configuration, SPA routing, Pages Functions migration, environment variables, and a domain switchover sequence that avoids production downtime.
Four ways to verify a Supabase JWT in a Hono API—@supabase/server, getClaims(), JWK middleware, and the legacy JWT secret—compared by signing algorithm, network calls, and when to use each.
Authentication succeeds but the query returns nothing—or an INSERT fails with a policy violation. This guide shows how to trace the failure across Auth, PostgREST, and Postgres using Supabase Unified Logs, and how to fix the five most common RLS causes.
Cloudflare Pages' Git integration watches a single repository. If your content or data lives in a different repo, you need a Deploy Hook and a GitHub Actions workflow to connect them.
CLAUDE.md rules don't always hold. Most failures fall into one of six diagnosable categories—four fixable by improving the file, two requiring a different approach entirely. Here's how to identify which problem you're dealing with.
AI coding agents increase pull request volume, but review capacity stays flat. Adding an AI reviewer addresses the symptom, not the problem. This article explains why, and how moving deterministic architecture checks before human review changes the constraint.
AI coding agents generate locally correct code that can silently cross architectural boundaries. Instruction files like CLAUDE.md don't reliably prevent this—they're probabilistic input, not deterministic constraints. This article explains why, with a concrete enforcement example.