Blog

Notes on evidence architecture.

30 Sept 2026

Pivot x SurrealDB: why we made the assertion the unit of analysis

How Pivot built its assertion graph for financial disclosure on SurrealDB, and why holding the graph, the vectors and per-matter access in one store cut retrieval cost 20 to 40 times.

15 Sept 2026

The control plane is fractal

The agent economy will not have one agent above every app. It will have nested control planes, and Pivot is both a destination and an orchestrator.

12 Sept 2026

Powered by the Polyphonic Assertion Ontology™

Two Forms E arrive describing the same marriage, and they disagree. That conflict is not a defect in the material. It is the material. Here is how we model it.

11 Sept 2026

Own your weights. In law, own the structure.

Sequoia says owning your models is no longer a performance sacrifice. In regulated legal work, the durable asset sits one layer below the model.

17 Jul 2026

Why family law, and why now.

We started with the hardest practice area — messy evidence, high liability, where generic legal AI fails. That's the strategy.

02 Jul 2026

Form E in the age of WhatsApp.

Form E was built for paper. Disclosure now lives in message threads, banking apps, and screenshots — that gap is where the real work is.

12 Jun 2026

Citation, not confidence.

Pivot can't hallucinate. Every answer traces back to source — the only standard fit for legal evidence.