Compare two grants issued on the same day, June 9, 2026, and you can see two engineering cultures decide where the interesting problem lives. Rivian was granted US12651813B2, “Systems and methods for battery architecture” (Rivian IP Holdings, LLC). GM was granted US12651810B2, “Electrochemical devices with multifunctional electrode separator assemblies having built-in reference electrodes” (GM Global Technology Operations LLC). Same calendar date, very different altitude.
Start with Rivian. The CPC tells the story: alongside the battery-architecture classes sit B60L 3/0046 and B60L 3/04 — “methods, circuits, or systems for monitoring or controlling” electric vehicles, specifically protection against electrical faults and overcurrent. That is a pack-level safety stance: design the architecture so that when something goes wrong electrically, the structure of the pack itself manages it. Rivian, a younger automaker building its packs from a cleaner sheet, is patenting the skeleton.
Now GM. Its grant goes the other direction — down into the cell. A reference electrode is a third terminal added inside the cell that lets the battery-management system measure what the anode and cathode are doing separately, instead of inferring it from the cell's overall voltage. Build that reference into the separator (the thin membrane between electrodes) and you get richer diagnostics without bolting on extra hardware. GM, with decades of cell-chemistry depth, is patenting the nervous system.
Neither approach is wrong; they are complementary layers. Rivian's pack architecture decides what happens when a fault propagates. GM's reference-electrode separator decides how early you can detect that a cell is drifting toward trouble in the first place. One is structural containment; the other is instrumentation. A mature EV program eventually wants both.
What the records do not tell you is which ships, or in what vehicle. A grant is a position, not a product roadmap — Rivian's architecture claims describe a method of building a pack, and GM's separator claims describe a method of sensing a cell, but neither filing promises a model year. Read them as statements of where each company thinks the defensible engineering is, not as a parts catalog.
The sector takeaway, comparing across the two: the EV pack is being attacked from both ends at once. Startups with clean-sheet packs tend to patent architecture and safety integration; incumbents with deep electrochemistry tend to patent cell-level sensing and materials. Watching which layer a company files in tells you, fairly reliably, where its real expertise sits.