████ ██████ ██ ██ ████ ████████ ██████ ██ ██ ████████ ██ ██ ██ ████ ██ ██ ██ ████ ██ ██ ██ ██ ████████ ██ ████████ ██ ████ ████████ ████ ██ ██ ██ ██████ ██ ██ ██ ██ ██ ██ ██ ████ ██ ██ ██ ██ ██ ██ ██████ ██ ██ ██ ██ ████ ██████ ██ ████████
████ ██████ ██ ██ ████████ ██ ██ ██ ████ ██ ██ ████████ ██ ██ ████ ██████ ██ ██ ██ ██ ██ ██ ██ ██ ██████ ██ ██ ████████
SOFTWARE ENGINEERING
engineer@aine:~$ man ai-native-engineering
Stop prompting.Start engineering.
The practical guide for engineers who already ship production systems and now want AI in every part of the lifecycle — not as a tool bolted on the side, but as a core engineering capability. Structured, production-oriented, and deliberately tool-agnostic.
Code ANSE2026 gives you 30 days of free access to the O'Reilly platform. Or press / and go anywhere on this site.
11
chapters
392
sections
135k
words
10h
to read
37
figures
0
vendor lock-in
01 / the argument
A year ago the company bought every engineer an AI subscription. The bill went up. The velocity didn't.
That is the first thing the CTO says out loud in the book's foreword, and it is the sentence that most engineering leaders recognise immediately. The tools are not the problem. What is missing is a discipline: a way of working that turns raw model capability into software a team can ship, safely, under pressure, into a codebase as messy as the one you actually have.
Vibe coding is not evil. It is great for throwaway prototypes. The trap is using it on code a customer will touch. See the full comparison →
02 / what it teaches
Six capabilities that outlive the tool you use this quarter.
- 01ch3 ↗
Context engineering
Assemble the right code, docs, rules and examples at the right time, so the agent produces consistent output instead of occasional lucky output.
- 02ch5 ↗
Spec-driven development
Write intent down before you prompt. The spec is the durable artifact that survives every tool change and model upgrade.
- 03ch10 ↗
Harness engineering
Build the environment around the agent: rules, skills, tests, feedback loops. Agent equals model plus harness, and the harness is yours.
- 04ch8 ↗
Verification and gates
Once code is cheap, trust is the bottleneck. Four layers of checks let AI-written code meet the same bar as anything else you ship.
- 05ch9 ↗
Agent orchestration
Know when one agent is enough and when to run a fleet, and stop hallucinations from propagating through a pipeline.
- 06ch10 ↗
Scaling in teams
Licences change nothing. Shared context, standard stacks and honest metrics are what move a team, not a purchase order.
03 / the shapes to learn
Three diagrams do most of the work in this book.
An agent is a while loop around a model call. Once you see that, the rest of the tooling stops being magic.
Six phases, durable artifacts, and human checkpoints at the seams — never buried in the middle of a phase.
Cheap checks first, on everything. Expensive human attention last, and only where it changes the outcome.
04 / free, and no email required
Three things you can run this afternoon.
The team canvas
Eight areas, ~50 prompts. Run it as a 90-minute workshop and walk out with a decision list, not a poster.
run it →
Score your team
Forty questions from the chapters. A number out of 100, a radar, your blind spots, and three moves to make first.
run it →
AI-native in a week
Seven days of theory and practice with curated videos and articles. The fastest honest way in.
run it →
- 01Adopt the AI-native mindset, from implementer to orchestrator
- 02Master context engineering for consistent, production-ready output
- 03Navigate the AI tool landscape with criteria instead of hype
- 04Apply spec-driven development as a practice that survives tool changes
- 05Orchestrate single and multi-agent systems without the usual anti-patterns
- 06Integrate verification and quality gates so AI code ships with confidence
- 07Collaborate across functions using specs, acceptance criteria and constraints
- 08Scale AI practices from IC to team to organisation
05 / who wrote it
Alfonso Graziano
AI Lead at Nearform, where he builds AI agents and runs them in production, and drives AI-native adoption across an engineering department of 300. He wrote an MCP server for Node.js with over 100,000 pulls on Docker Hub, and has given 20+ conference talks across Europe and the US.
The book is written from that seat: someone shipping this work, not observing it.
06 / stay in the loop
Chapters land as they are written.
The manuscript is due 25 November 2026. Subscribe and you get each Early Release chapter as it goes live, plus the working notes that never make it into the book.
Early Release chapters as they land, working notes, and the things that did not make the manuscript. No spam, and one click to leave.