A robot can learn how to move. Deployment requires knowing what task matters, what sequence is correct, what failure looks like, when to escalate, and how success is measured. Eru is that layer.
Real deployment requires a layer above locomotion — the layer that understands what the work is, why it matters, and whether it succeeded.
Locomotion, manipulation, joint control. Strong and improving fast — but indifferent to task context.
Knowing the task, its sequence, its constraints, its risk surface, its success conditions, and the cost of failure.
Nine capabilities, exposed as APIs, datasets, and evaluation suites. Each integrates above existing control stacks.
Human demonstrations of real workflows across 1,420 trade-grade task types — captured by Aegis, structured by PIN OS, outcome-linked.
erulabs/tasks · v0.4How skilled humans handle tools, objects, contact states, and corrections. Embodiment-agnostic — usable as priors for any platform.
erulabs/priors · v0.4Where actions fit inside real operations: sequence, dependencies, handoffs, and missing-step detection at the task graph level.
erulabs/workflow · v0.4What goes wrong, why, and what follows — pulled from real rework, change orders, claims, and safety events.
erulabs/failure · v0.4Calibrated benchmarks for whether robotic work meets real-world success conditions — inspection grade, not lab grade.
erulabs/eval · v0.4Long-tail scenarios mined from PIN — edge cases, adversarial conditions, and rare failures for training and stress testing.
erulabs/sim · v0.4When robots should act, assist, defer, or escalate. Mixed-fleet workflows with PIN OS dispatch and supervision.
erulabs/handoff · v0.4Persistent memory of sites, companies, workers, machines, and workflows — queryable by your stack on deployment.
erulabs/memory · v0.4Verifiable records that a robotic system has performed a task to spec, signed against PIN Certified — insurer-grade.
erulabs/certify · v0.4It sits above them. Your locomotion stack, joint controller, and manipulator policy are still yours.
Robotics companies are building bodies — and getting good at it. Eru models the work those bodies must understand: tasks, constraints, outcomes, and action knowledge across environments.
We're embodiment-agnostic by design. Pick the model that fits where you are.
Co-authored memos, shared datasets, and long-horizon roadmap alignment.
Access to task libraries, action priors, and failure data under commercial license.
Production API access to perception, workflow, action, and outcome heads.
Field deployments with PIN OS sites — your platform, our network, shared outcomes.
Build a robot tailored to a vertical we already operate in.
Deep strategic alignment with shared upside in mixed human-robot deployment.
Independent Eru-graded benchmarks against real-world success conditions.
Joint go-to-market into PIN's existing customer footprint.