// Track

Discipline

The Rules That Scale

The operational rules that prevent catastrophic failure: stop-and-replan, E2E validation, compound learning, security, cost discipline, and ticket hygiene.

Recommended: Complete Automation first

12 lessons~114 min total

Lessons are shown in recommended order. Complete them in sequence for the best experience — or jump to any lesson.

Lesson 13·9 min read

Stop-and-Replan: The Rule That Prevents Catastrophic AI Failure

The pattern that kills AI projects isn't bad models or bad prompts. It's forward-pushing past the moment you should have stopped. Here's the rule that fixes it.

Lesson 14·9 min read

E2E Validation: Why Shipped Isn't Done

The most dangerous phrase in engineering is 'I think it works.' Shipped means the code exists. Done means the user experience is verified. They are not the same thing.

Lesson 15·10 min read

The Compound Learning Loop: Building an AI System That Gets Smarter

Most AI users plateau because every session resets to zero. The builders who compound are the ones who capture, escalate, and enforce lessons across every session.

Lesson 17·10 min read

Security-First AI: Locking Down Your AI Agent Platform

An AI agent that can execute tools, post to Discord, and call external APIs is not just a productivity platform. It is an attack surface. Here is how to audit and harden it before something else does it for you.

Lesson 18·9 min read

How Many Tools? Diagnosing Over-Tooling in Your AI Stack

More tools means more capability — until it doesn't. Past a certain threshold, every tool you add makes your agent more confused, not more powerful.

Lesson 24·8 min read

Ticket Discipline: Why Fast AI Builders Need a Kanban

AI makes you build so fast that you lose track of what you built and what comes next. Without a kanban, you have speed but no continuity. Every session starts with 'where was I?' instead of 'what's next in the backlog?'

Lesson 34·10 min read

Incident Post-Mortems as Code: The Format That Turns Failures Into Firmware

Your AI system will make the same mistake 50 times unless you encode the correction into a file it reads at session start. Post-mortems belong in code, not in Slack threads.

Lesson 35·11 min read

Agent Isolation & Blast Radius: Protecting Parallel Work From Parallel Agents

A dispatched agent did what it thought was right — created a clean branch. That single action destroyed 130 uncommitted quiz questions across 26 lessons. Isolation is not optional.

Lesson 36·10 min read

Execution Confirmation: Don't Trust the Log

Your system logged 'Trade placed.' The dashboard shows a position. But the order never filled. You are managing a phantom — a record of something that never happened.

Lesson 37·11 min read

Deduplication at System Boundaries: Why Every Event Will Fire Twice

Same signal, same handler, same window. Two trades placed. Two emails sent. Two deploys triggered. Deduplication is not a feature — it is the gate that prevents your system from doubling every action.

Lesson 175·9 min read

Atomic Writes as DefaultNew

Four out of five state files in a production audit used direct writes. One crash mid-write, and the file is corrupt. The atomic write pattern takes three lines and makes corruption impossible.

Lesson 180·8 min read

CI Pipeline Debugging: Trust the Logs, Not Your AssumptionsNew

Two CI runs, two PRs — except both runs belong to the same PR. Five minutes wasted on impossible debugging because you trusted adjacent run IDs instead of spending two seconds to verify. This lesson builds the systematic habit that eliminates that class of mistake permanently.