An AI agency designs, builds and ships AI tools: audit, prototype, systems integration, evals. Real engagement flow, roles, deliverables, Monka lesson.
CI security agent: scan secrets, dependencies and configs on every PR, triage alerts, prepare fixes, and keep a human on irreversible changes.
Theodo, Artefact, Mistral, Galadrim, Ekimetrics, Koïno, Juwa or The Vibe Company: who to pick based on your company size and your AI project.
HDS hosting for AI in France: activities to cover, 2024 sovereignty rules, model/logs/index flows, cloud pitfalls and a vendor checklist.
Generative AI in media: 7 practical newsroom use cases, with sourced examples, guardrails, the AI Act and Paris Charter, and where to start.
AI agency, freelancer or IT services firm: a decision guide by project type, with sourced day rates, total cost, the risks and when each option wins.
Enterprise agent harness: the loop, sandbox, skills, evals and risk-based review to build around the model before your AI agents reach production.
How to get cited by ChatGPT and Gemini: stay crawlable, publish citable pages, prove expertise, and measure citations every month for your brand.
Enterprise RAG explained: how to make generative AI answer from your own documents, with real sources, the right access rights, and the LocService case.
How a fleet of AI security agents hardens a SaaS inside its CI: secrets, dependencies, IAM and config, security PR review, with humans approving.
PHP2Shell and the Railway incident (October 2026): what the Black October cluster changes for securing AI agents, SaaS API secrets, and HDS environments.
Working with an AI development agency: what to prepare, the five stages to production, who does what, and the warning signs that a project is drifting.
ChatGPT Business or Enterprise vs a custom agent: what changes, what doesn't, and when to build your own.
What an MCP server is, what it changes for AI agents in real tools, and checks before install: permissions, auth, logging.
How to deploy an AI agent in production in Europe: the 4 building blocks, GDPR, the AI Act and ANSSI guidance, and five steps from pilot to daily use.
Which process to automate first with AI, and who can set it up: a scoring grid, in-house vs agency vs no-code tools, and the LocService case.
CIR, CII, Diag Data IA, Osez l'IA: fund an AI project in France, with official rates and the MediaGen / AFP case.
Automate customer support with AI without losing quality: LocService (87% of tickets), plus Klarna and Air Canada lessons.
From AI agent POC to production: a concrete checklist, with Monka and LocService as proof.
What an AI-native agency is, how it differs from a classic agency and what it means for companies buying services. Based on Aaron Epstein (Y Combinator).
How to choose an AI agency: four options, fixed price vs time and materials, POC checklist, 10 questions before you sign.
Roll out Claude Code in a mid-size or large company: managed settings, permissions, data, pilot and measurement.
AI and health data in France: who needs certification, the 6 HDS activities, the 2024 framework, GDPR and the CNIL. What to check before production.
DeepSWE finally measures why Codex with GPT-5.5 felt stronger than Claude Opus on long multi-file tasks: closing the loop.
🇬🇧May 28
A founder who uses AI every day spent a session making music with an artist. The tools generated fast. The hard part was knowing what was worth keeping.
🇬🇧May 27
The new SkillOpt paper is easy to misread as prompt optimization. The useful lesson is more operational: treat skill files as trainable, validated artifacts.
🇬🇧May 27
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.
🇫🇷Apr 29
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.
🇬🇧Apr 24
A prompt is a disposable instruction. A skill is a versioned workflow package with triggers, scripts, references, review gates, artifacts, and memory.
🇬🇧Apr 24
MCP is standardizing tool access. The harder problem is giving agents a maintained company memory they can actually operate on.
🇬🇧Apr 23
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.
🇫🇷Apr 23
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.
🇫🇷Apr 22
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.
🇬🇧Apr 15
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.
🇬🇧Apr 8
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.
🇬🇧Feb 17
