Diego Sens
Founder, Silo
Silo is built by Diego Sens, a Brazilian corporate and litigation attorney writing the system from inside a working firm. The product reflects what a real practice needs from AI — not what a generic chatbot can plausibly demo.
I am a Brazilian lawyer admitted in Parana, under OAB/PR 96.036, and my day job is not adjacent to legal work. It is legal work. My background spans corporate and litigation matters, which means Silo was shaped around the actual pressure points of practice: fragmented corpora, procedural context scattered across systems, and the need to defend every citation to another professional who will read it skeptically.
The moment I decided to build Silo was not when models first became impressive. It was when they became useful enough that their failure mode started to matter. I could get fluent prose quickly, but not a reasoning trail I would trust in a filing, an internal memo, or a strategy call. The missing piece was not better wording. It was structure, provenance, and memory.
That is why I think a practicing lawyer is the right person to build this category. The hard part is not choosing a model provider. The hard part is knowing which parts of legal reasoning must stay inspectable, where a workflow can safely compress, and which mistakes are merely annoying versus professionally dangerous.
Why a practitioner, not a generalist
A generalist founder often sees legal AI as a search problem with a nicer interface. A practitioner sees a chain of professional obligations: court hierarchy, procedural posture, adverse authority, citation treatment, argument quality, and the difference between a paragraph that sounds right and a claim that can survive challenge. Those are not polish details. They determine what the product has to model from the start.
The asymmetry shows up just as much in what not to build. If you have practiced, you know that many seductive demo features do not remove the real bottleneck. The real bottleneck is usually context recovery: reconstructing what this case is about, which authorities actually matter, and how the current argument relates to the existing record. That is why Silo is built around structure and traceability rather than surface fluency alone.
Current phase
Silo is in private build and validation. Foundations are focused on densification of the knowledge graph in the corporate and M&A niche.
The most recent milestone was opening the technical data room at silo.legal/inside with live system numbers, architecture, recorded traces, and reviewer-facing context. That work matters because it forced the product to become legible to outside scrutiny, not only operational inside the build loop.