The free book · 8 chapters · 43 pages

The model decides. The harness executes.

From climbing ropes to Claude: a beginner's guide to the infrastructure that turns clever AI models into reliable working agents — the loop, the eight building blocks, and a hands-on build.

8chapters
8building blocks
5moves to your first harness
0dollars
60-SECOND CHECK
Watch first, read second

The whole idea, in 50 seconds

Why every reliable AI agent is really two things — a brain you rent and a body you build — told with a climbing rope and one very honest invoice.

Inside the book

Eight blocks. Every production harness has them.

Each block exists because the raw model has one specific, predictable weakness. Miss a block and the failure names itself: lying agents, context rot, silent errors.

01
System prompt
standing instructions
02
Tools
hands on the real world
03
Sandbox
a safe place to act
04
Filesystem
durable workspace
05
Memory & compaction
what stays in context
06
Verification
check the work
07
Guardrails & HITL
limits and approvals
08
Observability
logs, traces, audit
Free download

Take the book. Leave an email.

Join our daily AI newsletter — one sharp, five-minute read on agents, harnesses, and what actually ships — and the full book is yours instantly.

  • The complete 43-page book — figures, code sketches, glossary
  • Daily AI briefing, written for builders, never longer than a coffee
  • Unsubscribe any time; the book stays yours
No spam. One email a day, one book right now.
You're in. Welcome aboard. 🎉

Your first daily briefing lands tomorrow morning. Your book is ready now:

⬇ Download “Introduction to AI Agent Harness”