Research notes

What we learn building with AI agents: what we measured, what broke, what we changed.

• 27 October 2026

  How we built Libretto with AI agents

  A CMS written largely by agents, the limits we added along the way, and the times they did not hold.Libretto7 min read
• 28 July 2026

  Publishing with AI in five languages: what actually breaks, and how you catch it before the reader does

  The defect log of an AI transcreation pipeline in five languages, and how to catch the defects before the reader does.Building with agents14 min read
• 24 July 2026

  The Italian SMEs already using AI (and what they built)

  Some Italian SMEs already run AI in production: what they built, according to public sources.Evidence5 min read
• 23 July 2026

  Is AI-written code secure? What the independent studies say — and what to put in the contract

  What independent studies say about the security of generated code, and what to put in the contract.Building with agents8 min read
• 10 July 2026

  Is the AI vendor you're about to adopt trustworthy? SOC 2, ISO 42001 and what to ask before you sign

  What a SOC 2 really says, why ISO/IEC 42001 matters, and the questions to ask an AI SaaS vendor.Rules8 min read