engineer@aine:~$ whoami --verbose
AlfonsoGraziano.
An engineer who ships this work rather than observing it. The book is written from that seat, which is also why it argues with parts of the current consensus.
- AI Lead at Nearform
- Agropoli, Italy
- 20+ talks
- Role
- AI Lead at Nearform
- Based
- Agropoli, Italy
- Writing
- Learning AI-Native Software Engineering (O'Reilly)
- Built
- node-code-sandbox-mcp — 100k+ Docker Hub pulls
- Speaking
- 13+ appearances on this material
- Driving
- AI-native adoption across ~300 engineers
Alfonso Graziano is a hands-on AI Lead at Nearform, where he builds AI agents and runs edge AI tooling in production. He is driving the adoption of AI-native engineering across the whole engineering department, helping teams move from ad-hoc AI use to structured, production-ready workflows.
He built an MCP server for Node.js with over 100,000 downloads on Docker Hub, giving developers a practical way to hand an agent a real runtime. He has delivered more than 20 conference talks across Europe and the US on AI, deep technical topics, and career development, and writes for thousands of followers on LinkedIn.
Based in Italy, he writes and speaks from the perspective of someone who ships AI-powered systems, designs agent workflows, and coaches teams — which is the gap the book was written to close: between the theory people publish and the practice that actually survives production.
why this book exists
01
The tools were not the problem
Team after team bought licences and got nothing. The same tools in the hands of a team with shared context and real gates produced completely different outcomes. That difference is a discipline, and nobody had written it down.
02
Most of the writing is about building AI
There is a lot on LLMs, RAG and model APIs. There is very little on how you build ordinary software — a payments service, a migration, a dashboard — when AI is part of the stack.
03
Practices outlive tools
Every tool named in the book will look dated in eighteen months. Specs, context, gates and the harness will not. The book is deliberately written at that level.
book him
Four formats, all built on this material.
Conference talk
30 to 45 minutesThe role shift, the four pillars, and one worked example. Works for a general engineering audience.
Leadership talk
30 to 45 minutesFor CTOs, VPs and engineering managers: what changes at team and org level, and why licences alone move nothing.
Hands-on workshop
half day or full dayTeams work a real feature through the SDD loop on their own codebase, then build the first version of their harness.
Canvas session
90 minutesOne team, eight areas, a facilitated pass through the canvas, ending with a decision list and a ticket per open disagreement.
7 talks in rotation, 13 appearances
Slides and recordings for most of them are on the talks page. To book one, LinkedIn is the fastest route.
Working notes, before they are chapters.
The newsletter is where the thinking happens in public: what worked this month, what did not, and which parts of the manuscript changed because of it.