A Book · Attacca
Specs In,
Software Out.
The bottleneck in software is no longer how fast you can write code. It’s how clearly you can describe what you want.
By Jhonn Moreno · 15 chapters · 4 appendices · ~79,000 words
What this book is
Models can generate thousands of lines of code in minutes. But ambiguous specs produce ambiguous software. The difference between a demo and a production system is the quality of what goes into the machine.
This is a practitioner’s book. Not theory, not hype. Real projects, real failures, and the methodology that came out of both. Built from shipped work — VZYN Labs, SonIA, Edifica, Regasificadora — not from benchmarks or demos.
If you’re building AI agent systems that need to be right, not just fast, this book is the map.
Contents
Read the book, chapter by chapter.
The Shift
- 01
The Bottleneck Has Moved
Thirteen agents. Two months of engineering. Hundreds of hours across a team of developers. Hexagonal architecture — a pattern I'd never heard of before the lead engineer explained it to me as "the rig
- 02
The March of Nines
Here is a number that should change how you think about AI agents: **65%.**
- 03
DORA Is Cracking
A randomized controlled trial published in 2025 found that experienced open-source developers using AI tools completed tasks **19% slower** than developers without AI tools.[^1]
- 04
Trust As Architecture
The first question I ask on every project is the same: **"What's the worst realistic outcome if this system gets it wrong?"**
The Framework
- 05
Intake: The Question That Governs Everything
Oscar Isaza described his business in fifteen minutes. A $42 million LNG regasification operation. International tankers. Cryogenic containers on barges. A five-year contract with Ecopetrol. Thermal p
- 06
Discovery: Understanding Before Changing
According to a 2024 Cisco study of 2,600 privacy and security professionals across the globe, more than one in four organizations — 27 percent — have banned generative AI tools entirely. Another 61 pe
- 07
The Spec: Writing for Machines, Not Humans
I built a Frankenstein once.
- 08
Build: Execution on Deterministic Rails
Hernan described his week as a "montaña rusa" — a roller coaster. He'd been working on Edifica, implementing features with Cursor and Claude Code. Some sessions produced exactly what he needed. Others
- 09
Test: Four Layers of Confidence
In February 2026, researchers at the Icahn School of Medicine at Mount Sinai published the first independent safety evaluation of ChatGPT Health — a tool that had attracted roughly forty million daily
- 10
Certify, Deploy, Maintain: The Last Mile
Most agent failures happen after deployment, not before.
Enforcement
- 11
Harness Engineering: The Model Is Not the Product
The same model. The same prompt. Two different harnesses. One scored 78% accuracy. The other scored 42%.[^1]
- 12
Intent Engineering: Aligning Agents With Organizations
Klarna deployed an AI customer service agent that resolved tickets faster than any human.[^1] By every prompt-level metric, it was succeeding. It was efficient. It was responsive. It was destroying cu
- 13
Simplicity: What Rob Pike Teaches Us About Agents
Rob Pike wrote five rules of programming decades before AI agents existed.[^1] They apply with double force now.
Getting To Work
Practitioner’s reference
Take it with you
Download the full book.
Free. No email required. Updated as the methodology evolves.