Research Methodology: Rigor and Accessibility

Sources and Verification

Documentary research relied on:

- Academic sources: Works by historians of science, university publications

- Primary sources when accessible: Ancient texts, descriptions of inventions

- Reference works: Kate Crawford (Atlas of AI), Vladan Joler, historians of computing

Critical point: To avoid anachronisms and chronological confusion, a cross-verification agent was specifically deployed for the periods of antiquity, the Middle Ages, and the Renaissance. This agent verifies that facts, dates, and actors do not get mixed between periods.

Balancing Scientific Accuracy and Accessibility

Three principles guided the writing:

1. Never simplify to the point of distortion: Technical concepts are explained but not watered down

2. Avoid anglicisms in French content: A deliberate choice to preserve the richness of the French language

3. Use extended metaphors: Each article develops a metaphor to anchor abstract concepts

Treatment of Non-Western Contributions

Particular effort was made to:

- Document medieval Arab inventions (Al-Jazari and his hydraulic automata)

- Include African systems of thought (numeration, divination)

- Contextualize knowledge transfers (translations, conquests, trade routes)

Acknowledged limitation: This is not an exhaustive work but rather a narrative opening to show that the history of AI is polyphonic.

Technical Architecture

The Problem with Standard Interfaces

To produce a project of this scale, standard conversational interfaces (ChatGPT, Claude online, Mistral Le Chat) quickly reach their limits:

- Restricted context window: Impossible to simultaneously manage 40+ articles

- No persistence: Each conversation starts from scratch

- Indexing difficulties: No integrated document management system

The Solution: An Augmented Development Environment

The project was entirely produced in Visual Studio Code with Claude Code, a tool that transforms AI into a development assistant capable of:

- Working on entire folders

- Maintaining consistency across files

- Indexing and referencing research

- Integrating external tools via MCP (Model Context Protocol)

Skills and Agents: Automating Repetitive Tasks

Rather than repeating the same instructions, "skills" (specialized capabilities) were created:

- avalon-author: A writing skill following the editorial guidelines (style, length, three-movement structure)

- Chronological verification agent: Verifies that no historical fact is placed in the wrong period

- Cross-analysis agent: Compares information across sources to detect inconsistencies

These agents function as specialized assistants, each with a specific mandate.