Lattice Robotics is the Bulletin's recurring example for why the Web4 thesis is not only a software thesis. The company builds hardware enclosures — industrial cabinets, edge boxes, sensor packages — with an agentic orchestration layer embedded at the firmware level. The argument is structural: that an agentic system without ownership of its compute layer is, eventually, rate-limited into irrelevance by the platforms it depends on.

The company was founded in Tokyo by Akihito Mori and Naomi Sutter. Mori's background is in industrial control systems; Sutter is a former platform engineer with an unusual depth in low-level orchestration. The combination is rare for the category — most agentic-AI companies are software-only, and most embedded-systems companies have shown little interest in agentic abstractions. Lattice is one of the few that has taken both halves of the problem seriously.

Their commercial product is, at this stage, deliberately narrow. Lattice ships a sealed compute unit, sized for industrial environments, that runs a small set of agentic workloads locally with a defined integration to cloud agents. The unit is sold into specific verticals — light manufacturing, water utilities, [TKTK: third vertical to be confirmed]. The Bulletin has tracked the company since its first published architecture note in early 2024 and has used Lattice as a recurring counterweight to software-only Web4 framings.

The company's editorial position, which is also the position the Bulletin has cited most often, is that agentic systems will increasingly need to argue with the underlying platform layers they depend on. Cloud quotas, rate limits, model availability, identity primitives, integration permissions — all of those become structural constraints on what an agentic workforce can actually ship. Lattice's argument is that a meaningful slice of the agentic-systems category will end up running on edge hardware its operators own, and that the firmware layer is the right place to bake in orchestration.

It is, in the Bulletin's view, a position that may turn out to be partially right rather than fully right. Most of the agentic workloads the directory tracks do not have the kind of physical-operations footprint that demands edge hardware. But Lattice's argument is not that all agentic work belongs at the edge. It is that the parts that do belong there are structurally more important than they currently get credit for. That argument has aged well in the eighteen months since the company first made it publicly.

Lattice is also the directory's clearest example of a company that is intentionally not trying to be a platform. Their public positioning is that they are an embedded-systems vendor with an unusually opinionated firmware story, not a competitor to the larger orchestration platforms the directory tracks. The Bulletin has appreciated the restraint. The category does not need every hardware-adjacent player to claim it is an operating system.

We have not been able to fully evaluate Lattice's customer traction. The two deployments the company has discussed publicly are small, and the company has not disclosed [TKTK: revenue or pilot conversion figures to be confirmed]. The directory entry is here on the strength of the engineering position and the editorial fit, not on a commercial assessment. We expect to update this entry as the company's deployment base grows or stalls.