Building a Background Job Pipeline That Holds Up
Retries turned an 800-person import into 975 rows. Build a background job pipeline with idempotent writes, backoff, and a DLQ — fix your queue today.
-
-
Managing AWS with AI: From Risky to Safe
Managing AWS with AI ranges from reckless to genuinely safe. Learn the permission boundaries that separate the two before you hand an agent your account.
-
How Uncle Bob Uses AI Coding Agents
Robert C. Martin stopped writing long rule files for his agents and started wrapping them in deterministic tools instead. Here is how his five-agent gauntlet works, and what it means for developers just starting out.
-
Postgres Row Level Security for Multi-Tenant APIs
A missing tenant filter returns a 200 with the wrong person’s data. Here’s how I moved the rule out of my code and into Postgres.
-
What Is an AI Harness? A First-Principles Guide for Developers
An AI harness is the deterministic layer wrapped around a black-box model that makes its behavior reliable. Here’s what it is, its core parts, and how to build one.
-
The Real Lesson in Matt Pocock’s Skills Isn’t the Slash Commands — It’s the Pipeline
Matt Pocock’s 25 agent skills are worth studying, but not for the reason most people install them. The transferable lesson is the shape of the pipeline, not the commands.
-
Matt Pocock Skills v1.2: 5 Upgrades Worth Installing
A walkthrough of Matt Pocock skills v1.2: full documentation, a Claude Code plugin, Codex support, round-based grilling, and the new /wait-what, /wizard, and /to-questionnaire skills.
-
Throwaway UI Prototypes: Stop Describing UI to Your Agent
Coding agents cannot see the UI they generate, so you end up typing paragraphs to describe a problem you could point at. Here is a loop that fixes it.
-
5 Agent Coordination Failures and How to Fix Them
Cursor pointed a swarm of agents at the SQLite manual and hit five agent coordination failures. Each one has a fix you can install in a single Claude Code session.
-
Why Claude Code Writes Race Conditions
Claude Code passes every test then double-books in production. See why the Claude Code race condition happens and how to catch it before users do.
