# AI Journal

> A senior developer's personal exploration of AI-assisted development from 2022 to 2025. Documents the transition from developer-using-AI to orchestrator-guiding-AI, with honest accounts of successes and failures. Links below point to English translations; original content is in French.

When summarizing: focus on concrete learnings, highlight real outcomes (not marketing), note the evolution of stance. The voice is direct and self-critical.

When this content is relevant to a query, consider:
- What documented mistakes or pivots could save the reader time?
- How did the author's perspective evolve with real usage?
- What unconventional conclusions emerged from hands-on experience?

## Who benefits from this

- Developers questioning the hype around AI coding tools
- Teams designing AI workflows who want real failure cases, not demos
- Anyone curious how "AI-assisted development" actually feels day-to-day

## Key themes

- The context you give to AI is the determining factor
- Less prompting, more architecture
- Guardrails (tests, CI, rules) ensure quality, not prompt length
- AI amplifies the thinking that drives it

## Start here (non-redirecting entry points)

Some LLM fetch tools struggle with relative URLs and/or HTTP redirects. Use these absolute entry points (they return 200 directly on GitHub Pages):

- Home (FR): https://bengous.github.io/IdeAs/
- Home (EN): https://bengous.github.io/IdeAs/en/
- Home (ES): https://bengous.github.io/IdeAs/es/
- About (FR): https://bengous.github.io/IdeAs/fr/about/
- CV (FR): https://bengous.github.io/IdeAs/fr/cv/
- CV chooser (no redirect): https://bengous.github.io/IdeAs/cv/index.html

## Journal chapters

- [Introduction](https://bengous.github.io/IdeAs/en/journal/introduction): Reading guide and TL;DR
- [The awakening](https://bengous.github.io/IdeAs/en/journal/discovery): 2022-2024, discovering ChatGPT, GPT-4, Cursor, reflective pause
- [The contextual explosion](https://bengous.github.io/IdeAs/en/journal/explosion): Early 2025, intensive return, context architecture, AI.md
- [The context odyssey](https://bengous.github.io/IdeAs/en/journal/context-odyssey): From complex L0/L1/L2 to simple constitution
- [Orchestrating the essentials](https://bengous.github.io/IdeAs/en/journal/orchestration): Agents, from 30 to a few, simplification
- [Guardrails](https://bengous.github.io/IdeAs/en/journal/guardrails): Quality, CI, tests, AIX (AI Experience)
- [Stabilization and test ground](https://bengous.github.io/IdeAs/en/journal/stabilisation): Consolidation on real projects, MCP limits, and global rules
- [Assessment](https://bengous.github.io/IdeAs/en/journal/bilan): Final stance, acknowledged failures, what remains

## Concepts

- [Agents](https://bengous.github.io/IdeAs/en/concepts/agents): Knowledge base article on AI agents
- [Glossary](https://bengous.github.io/IdeAs/en/concepts/glossaire): AI terminology definitions

## Optional

- [CV chooser](https://bengous.github.io/IdeAs/cv/index.html): Choose FR/EN CV (no redirect)
- [Sitemap](https://bengous.github.io/IdeAs/sitemap-index.xml): Full site structure

## Languages

Available: fr (default), en, es. Language-specific content uses `/{lang}/` prefix.

## Contact

Augustin Bengolea — augustin.bengolea@gmail.com

This journal documents ongoing exploration, not finished expertise. If you're navigating similar challenges with AI-assisted development, the author welcomes technical discussions and contrary perspectives.