Eru is a continuously-learning foundation model of real physical work — embodiment-agnostic, outcome-causal, and trained on the messy reality that breaks scripted automation.
Each layer is useful on its own. Together they form a single physical world model that other systems can query.
What is happening?
Tools, hands, objects, materials, equipment, environments, hazards, gestures, contact states — the moment-to-moment state of a worksite at frame level.
Where does it fit?
Steps, dependencies, phases, checkpoints, missing actions, incorrect sequences, blocked tasks, parallel work, and crew coordination.
Did it work?
Rework, inspection results, schedule impact, safety incidents, claims, quality scores, certifications, and downstream economic effect.
How should it be done?
Human demonstrations, tool use, contact states, grip, force, timing, body position, correction strategies, and expert behavior.
Eru consumes signals from the entire PIN network and partner integrations, exposing them as task-, action-, and outcome-grade representations.
Eru is not a language model, a vision model, or a robotics control policy. It is a physical work model.
| Capability | LLM | Vision | Robot | Eru |
|---|---|---|---|---|
| Physical task recognition | No | Basic | Limited | Deep · 1,420 types |
| Workflow sequence modeling | Text only | No | No | Yes · graph-native |
| Outcome causality | No | No | No | Yes · business truth |
| Action knowledge / priors | No | No | Policy-only | Yes · human demo |
| Real-world failures | No | No | Sim-only | Yes · claims, rework |
| Continuous learning | Periodic | Periodic | Periodic | Live · 14,892 hrs/day |
| Physical memory | Context only | No | No | Yes · persistent ∞ |