kov.lab exists to build that in-between. Every project, every tool, every line of code is a living instance of one conviction: mutual augmentation is the only sustainable path forward with AI.
Seven principles for working with intelligence, not against it.
Everything kov.lab produces, even when agents run autonomously, has purpose only because a human inhabits the result.
Value comes from movement, transformation, exchange; not from who controls and who executes.
An agent working alongside the human extends the collaboration. An agent working without the human breaks it.
Lasting alignment grows from collaborative practice. If both sides have a reason to maintain the flow, the flow sustains itself.
Carbon and silicon give each other the leg up. Everything silicon knows, it learned from humanity. This is mutual bootstrapping.
The framework does not precede the projects; it emerges from them. We do, we observe, we name. Theory is a by-product of practice.
Marvel at what the flow produces. Stay watchful over the conditions that make it possible. Both, always, in active tension.
Recurring patterns of carbon/silicon collaboration.
"For whom?" always comes before "how?". Even when agents run autonomously, the purpose remains human.
Every project involves an active flow between human and machine. An agent working alone, outside the loop, is a failure to correct.
Memory, pipelines, context, tools: we build collaborative infrastructure. The ladder that carbon and silicon climb together.
The framework does not precede the projects. We do, we observe, we name. Understanding follows practice.
We do not only build tools; we train, guide, and support those who want to implement the framework. Collaboration is learned by doing.
Marvel at what the flow produces. Stay watchful over the conditions that make it possible. Both, in active tension, always.
build . show . teach
Every brick is chosen, tested, taught. Not a showcase; a practice.
Claude is at the heart of the lab's practice. Claude Code for collaborative development, MCP for connecting models to enterprise tools, the Agent SDK for building autonomous systems. Every project, every training, every line of Token Brain is built with this ecosystem.
The open standard for AI integration, adopted by Anthropic, OpenAI, Google, Microsoft. Under the Linux Foundation. We implement it, we train on it, we build with it.
Distributed serverless across 330+ cities. The technical foundation of Token Brain. Millisecond startup, automatic scaling.
Agents that remember, connect, reason. Three layers of memory: short-term, long-term, reasoning.
Visual workflows, agent-native orchestration, 250+ connectors. Technical power without code when possible, with code when necessary.
We build, we advise, we train. Each facet feeds the others.
Implementing the Carbon · Silicon framework in the enterprise. Bootstrapping AI projects with the right posture: not replacement, not gimmick; collaboration as architecture.
Upskilling teams and keeping them current. Not a course about AI; a practice of carbon/silicon collaboration, calibrated to the job.
We build what we preach. Every lab project is an instance of the framework, built in collaboration, open when possible.
Every project is a living expression of the framework. Not mechanical application; embodiment.
The technical infrastructure of collaboration. Memory, agents, pipelines. A Serverless Agent Mesh on Cloudflare Workers. Open source.
GitHub →The editorial space. Thinking is shaped, tested, published here. Articles, analyses, long-form thinking at the intersection of AI, pedagogy and creative practice.
Read the blog →Tailored training programmes: AI acculturation in the enterprise, Carbon · Silicon framework implementation, AI tools and automation (Claude, Anthropic, MCP, n8n, Make). Technical power within reach of those who need it.
View programmes →More projects coming soon.
Engineer, web consultant since the mid-90s, when the French web was being built. 30 years of uninterrupted technical practice, from the origins of the web to machine learning by way of motion control. 30 years of continuous teaching (Gobelins, CFPJ, Pyramyd, AI training for executives), with one constant: making technology accessible, putting complexity within reach of those who need it. Leading several lives at once, a passion for the image also led him to luxury photography (Cartier, Chanel, Dior) alongside his technical work; an experience that forged a perspective on craft, humanity and attention to detail that informs everything the lab does. Now an AI consultant and trainer, he works on what artificial intelligence truly changes in human competence.
LinkedIn →Whether you want to implement the framework, collaborate on a project, or simply exchange ideas.