Published through the company · July 2026
AI Discipline
A Senior Engineer's Field Manual for Agentic Development
Written by our founder and principal architect, and published independently through R&T BrightLight Technologies.
Field Manual
AI Discipline
A Senior Engineer's Field Manual for Agentic Development
The premise
AI agents can generate a sprint's worth of code in an afternoon — and flood a codebase with plausible, confident, subtly wrong code faster than any human can review it.
AI Discipline argues that the economics of engineering have inverted: writing code is now nearly free, and discipline is the scarce resource. It is a field manual for the engineers and leaders accountable for shipping safely in that world.
What it argues
The bottleneck moved. Most teams haven't.
Engineering practice was built around an assumption that no longer holds — that writing code is the expensive part. When generation becomes cheap, every process downstream of it has to be rebuilt.
Review is now the constraint
When an agent produces more code in an hour than a team can read in a day, human review stops being a quality gate and starts being a rubber stamp. The book is direct about what has to replace it.
Plausible is not correct
The failure mode of generated code isn't obvious breakage — it's code that reads well, passes a glance, and is quietly wrong in a way that surfaces in production months later.
Verification has to be automated
If correctness depends on someone being careful, it will fail at volume. Discipline that isn't enforced by a machine is a preference, not a standard.
Accountability doesn't transfer
The agent doesn't carry the pager, face the customer, or answer to the board. Whoever ships remains responsible for what shipped — a point the book refuses to soften.
Written for people who ship
A field manual, not a think piece. It's aimed at senior engineers, leads, and technical leaders who are already using these tools and are accountable for the results.
Practice over prediction
No forecasting about what AI will do to the industry in five years. It's about what to do on Monday, with the tools and the team you currently have.
Who it's for
If you're accountable for what ships, it's for you.
- Senior and lead engineers Already working with agents daily, and increasingly aware that the old review habits don't scale to the new volume.
- Engineering managers and directors Being asked how much faster the team will be, and needing a defensible answer about quality as well as speed.
- Technical founders and CTOs Building with a small team and heavy AI leverage, where a single bad pattern propagates fast and quietly.
- Business leaders funding software Who want to understand what the productivity claims actually mean, and what they cost if the discipline isn't there.
Details
Title
AI Discipline: A Senior Engineer's Field Manual for Agentic Development
Published
July 2026
Publisher
Independently published through R & T BrightLight Technologies, LLC
Where to buy
Inquiries
A note on why this exists
We're a software company, not a publisher.
This is the only book we publish, and we have no plans to become a publishing house. It exists because the argument in it is the same one we make to every client: that discipline, not speed, is what makes AI-assisted engineering safe to build a business on.
If the book describes a problem you recognize in your own organization, the rest of this site describes how we help fix it.
Press & speaking
Interviews, reviews, and speaking requests.
For review copies, podcast or press inquiries, or speaking on engineering practice in the age of agentic AI, get in touch.