Everything I learned building agentic workflows in production, written down. CLAUDE.md, context budgets, parallel sessions, the lot.
Proof, not promises
One real session, trimmed for length: a plan before any code, an edge case caught, a course-correction, a result. No magic words.
New screen: a BLE pairing flow. Don't write code yet. Read useDevicePairing and the two screens that consume it, then give me a five-line plan and the one edge case I'll forget.
Read three files. Plan:
edge case The user backgrounds the app mid-scan. iOS suspends the BLE callback, so on resume the hook still reads scanning, forever.
Good catch — that's the whole feature. Drop the optimistic rollback, it's premature. On resume, re-arm the scan and time it out at ten seconds. Show me the hook diff only.
useDevicePairing.ts — added an AppState listener: on active, if status is scanning, restart the scan with a 10s timeout that transitions to error. Fourteen lines. No screen changes needed.
Ship it. Now write the test that backgrounds the app mid-scan and asserts we don't hang.
Real session. Sanitized, lightly trimmed.
One-time payment, lifetime access
No subscription. The first chapter is free, and there's a 14-day full refund — no questions asked.
No. Chapter 1 covers how Claude Code works, eight practical quick wins you can use immediately, and the one constraint that governs everything.
This course teaches Claude Code. The methodology (context management, spec writing, feedback loops) translates to other agentic tools, but the techniques, configs, and examples are Claude Code specific.
Email me within 14 days for a full refund, no questions asked.
No. It's text-based with code examples. You can read at your own pace and copy-paste directly.
Yes, the course is fully responsive. But you'll want a laptop nearby for the exercises.
Testing strategies with AI agents, team workflows (shared CLAUDE.md across 5+ devs), advanced debugging of multi-file logic bugs, and security hardening beyond basics. These are real topics. They're out of scope because each one deserves its own deep treatment, not a surface-level appendix.
The tools move fast. Autonomy defaults, the 1M context window, Skills going GA, and a mid-year billing change all landed while this course was being written, and the course was refreshed against every one. Updates are included. The deeper answer is the final lesson: the method it teaches is the durable part. When a setting changes, you relearn the setting, not the approach.