Seeing Machines

The dashboards are green. The customer is on the phone. The pager hasn't gone off in eleven days, and your last incident took forty-seven minutes to diagnose. You don't have a monitoring problem. You have an observability problem.

Modern software systems have outgrown the way we watch them. The "three pillars" model - logs, metrics, traces - was useful in 2018 and has been quietly failing us ever since. Microservices sprawl. AI agents make decisions in loops nobody designed. Cost surprises arrive in monthly invoices. The classic tooling sees less of what matters every year, even as we spend more on it.

Seeing Machines is a manifesto for what comes next.

JJ Shen builds a new model from the ground up - seven strata of observability, from substrate plumbing to business outcomes - and shows how most teams stall at stratum four, where the technical signals are rich but the business reality is invisible. The "missing middle" is where the money lives, and the book spends most of its time there.

Through the recurring story of M82, a fictional logistics company climbing from year zero to year seven, the book walks the discipline as it actually develops in the wild: the dashboard graveyard, the deletion ceremony, the agent that ran away on a Wednesday and cost $14,000 by Friday before anyone noticed. Theory dies on contact with on-call. This book is what survives.

What you'll take away

• A practical seven-strata framework you can apply to any system in production


• The five-rung maturity model - and an honest test for where your team actually sits


• How to instrument AI agents so the next runaway loop costs forty dollars, not forty thousand


• Why most SLOs are theater, and how to build the rare kind with real consequences


• The four-slide deck that defends your platform budget across three CFO transitions


• Rituals that compound - observability hours, deletion ceremonies, the four-question alert test

Who this is for

Staff and senior engineers. Platform leads. SREs who have outgrown the SRE book. Engineering managers who need to explain to a CFO why the platform team matters. Anyone shipping AI agents into production and quietly worrying about what they'll find when the cloud bill arrives.

Who this isn't for: Beginners looking for a tooling tutorial. Buyers comparing observability vendors. Readers who want to be told their current stack is fine.

What the book promises

That the gap between what your systems emit and what your team understands is the most important thing you're not measuring - and that closing it is a five-year discipline, not a quarterly project. The book is the playbook for that climb.

The systems are talking. This book will help your team listen.