Lucci Labs

Lucci Labs is an independent research studio working on tool-using agents, local inference, and the interfaces around them. We are interested in what makes intelligent systems dependable in practice: how they receive context, how they act on it, and how the people using them stay oriented to what the system knows and does.

The method is consistent. We formalize a question, build the smallest system that can test it, and publish what the work reveals — including where the idea holds and where it breaks. Theory and implementation move together; neither is allowed to drift far from the other.

The current direction centers on context as infrastructure: treating the information an agent receives as versioned, observable system components rather than a prompt assembled once at the beginning of a run. The research notes live under Writing; the instruments built alongside them under Building.

Everything is open source by default. The code lives on GitHub, and the lab reads its mail at contact@luccilabs.xyz.