Introduction

Introduction

I started working with AI in late 2022. This journal tells what it actually changed for me: moments of overexcitement, step-backs, failed attempts, what held up, and what I ended up keeping.

This is neither a tutorial nor a manifesto. It’s a chronological narrative, with concrete projects and more technical passages when needed.

If you are interested in AI, you’ll find:

• how I integrate it into daily work;
• how I moved from “generated code” to systems that frame AI;
• how I work today: set rules, delegate execution, keep control.

If you want to discuss this or work with me: augustin.bengolea@gmail.com.

How to read (and use) this journal

• If you are not technical: Read the phase introductions, the anecdotes, then the “Reflections” section. You can skim parts with acronyms and technical terms (L0/L1/L2, CI, MCP, etc.). What matters most is the tradeoffs and decisions.

• If you are a developer or very curious: Read everything: chronological phases, context architecture (L0/L1/L2), agent orchestration, and the stabilization phase (web project, agent orchestration, dotfiles). You’ll find reusable ideas and anti-patterns from real work.

• If you are in a hurry (or already a fan of LLMs): You can also copy-paste this journal directly into an LLM. A prompt is injected at the top to tell it how to extract the essentials.

  It’s also a small prompt engineering demo: the document self-documents for AI. Copy, paste, done.

  It’s a bit meta, but consistent: this journal is about orchestrating AI, so you can start by orchestrating the journal itself.

This is a logbook, not a polished manifesto. You get the real path here: trials, errors, adjustments, from developer-scribe to developer-orchestrator.