Longer articles and deeper write-ups.
La sécurité des agents IA ne se décide pas seulement par marque d'outil. Elle dépend surtout du niveau d'autorité accordé : lire, proposer, préparer ou exécuter.
A message from Stan and Antoine: after Quivr, YC, open source, and our customer support pivot, we are stopping Quivr as our main company direction to build The Vibe Company, an AI-native company focused on real workflows.
A prompt is a disposable instruction. A skill is a versioned workflow package with triggers, scripts, references, review gates, artifacts, and memory.
MCP is standardizing tool access. The harder problem is giving agents a maintained company memory they can actually operate on.
Le mode opératoire pour transformer Linear, GitHub, Notion, Drive et vos exports en dossier CIR/CII vérifiable : repo atelier, matrice de preuves, agents spécialisés, LaTeX et validation humaine.
Comment on a transformé Linear, GitHub, Notion, Google Drive et nos exports internes en dossier CIR/CII de 100 pages. Pas avec un prompt magique, mais avec une vraie chaîne de preuve.
I built Quivr. The system I actually use every day to run my company does the opposite. 188 markdown files, an MCP server that teaches agents a methodology, no semantic search, no embeddings. Here's why this is where personal software is going — and how I wrote this article with it.
Tool access is the shallow promise of MCP. The real value is packaging company judgment - context, order, permissions, review, and taste - into a surface agents can actually use.
I’ve wanted to build my game for a while. But I needed a solid visual identity to finally start coding. Generic AIs were too inconsistent for a real project. The fix was Scenario.com. By training a custom model on 25 specific images (refined with Claude to hit a \"Zelda meets Clash Royale\" vibe), I finally got a unified look. AI didn't do the creative work for me; it just removed the technical wall. Bestiary is now officially in development.