SPATIAL GOVERNANCE FOR PHYSICAL AI

Physical AI can plan an action—and still violate a safety boundary.

ÉTÈ checks a machine’s proposed action against approved spatial, operational and human-safety boundaries before it happens.

Explore the system ↓

Speculative urban environment with autonomous vehicles, drones and people — PHYSICAL AI IN SHARED SPACEIMAGE · MAURICIO OROZCO

Speculative city scene with a delivery robot, cameras and pedestrians — MACHINE–HUMAN TERRAINIMAGE · MAURICIO OROZCO

THE DEPLOYMENT GAP

AI proposes actions.
Infrastructure needs accountable decisions.

Robotics teams still stitch together planning, safety rules and after-the-fact logs. The result is hard to verify or reuse.

THE GOVERNANCE FIELD

The original ÉTÈ field remains as a visual threshold between the deployment problem and the system architecture.

AUTHORIZED FIELD

HUMAN-PRIORITY BOUNDARY

01 / PRODUCT

One assurance stack.
Two operating layers.

DESIGN TIMEÉTÈ STUDIO

Turn site rules into testable spatial policy.

Import trajectories and site geometry, define policy boundaries, replay scenarios and compare baseline versus governed behaviour.

Customer-data adapter
Policy-to-geometry compiler
Counterfactual Policy Diff
Audit-ready evidence

RULE → SHAPE → TEST

RUN TIMEÉTÈ RUNTIME

Authorize the action—or force a safer state.

A deterministic gate sits between a probabilistic planner and physical execution. Hard constraints remain non-overridable.

ALLOW / HOLD / REROUTE
DEGRADED / REJECT / EMERGENCY
Uncertainty-aware envelopes
Replayable decision record

PLANGATEACT

02 / BUSINESS MODEL

ÉTÈ sells assurance software and paid co-pilots to organisations deploying physical AI—not reports, grants or abstract frameworks.

PRIVATE-SECTOR BUYERS

Robotics, mobility and infrastructure operators

Teams that need to prove how autonomous machines behave around people, restricted zones and changing operating conditions.

BUY: integration, site assurance, runtime licences and evidence tooling

PUBLIC-SECTOR BUYERS

Cities, testbeds and public authorities

Institutions that need machine behaviour to reflect municipal rules, public-space priorities and auditable human oversight.

BUY: bounded pilots, policy-to-geometry tools and deployment assurance

PARTNERS & CHANNELS

Integrators, GovTech firms and research facilities

Partners bring customer systems, local implementation or test environments; ÉTÈ supplies the assurance architecture and product layer.

EARN: defined delivery roles in joint commercial co-pilots

01 · ENTERPaid Phase 0Customer data, one bounded scenario and a three-week Policy Diff.

02 · DEPLOYStudio + integrationSite-specific adapters, policy compilation and assurance workflows.

03 · SCALERuntime licensingPer deployment or enterprise agreements, with evidence and support modules.

WHAT THE ROUND BUILDS

From validated architecture to a sellable product.

Angel capital pays for the engineering capacity required to turn ÉTÈ’s spatial-governance method into software customers can test, integrate and buy.

Customer-data adapter and offline Studio PoC
Policy-to-geometry compiler and Policy Diff
Deterministic runtime core and decision API
Replayable evidence, security testing and two paid co-pilots

PRE-SEED RUNWAY€450k

12–15 months to a customer-data PoC, deterministic runtime and two paid European co-pilots.

€150k first closeFunds the engineering plan, first adapter, working Phase 0 product and the route to initial paid deployment.

ÉTÈ may use selective public R&D funding where it accelerates technical validation, but the company is designed around commercial revenue from public and private customers. Grants are complementary—not the business model.

03 / WHY ÉTÈ

Governance becomes
an operational geometry.

Spatial intelligence

Rules become zones, corridors and temporal conditions.

Independent assurance

The model proposes. A separate layer authorizes.

Evidence by design

State, rule and decision remain replayable.

THE OPERATING PRINCIPLE

Generate → expose uncertainty → audit → authorize → execute → evidence

ÉTÈ does not claim that a world model is always correct. It defines what must be independently checked before an inferred world is allowed to produce physical action.

04 / CO-PILOTS

Start with one bounded
decision that can fail.

URBAN

Shared public space

Human-priority corridors, crowd conditions and municipal operating rules for autonomous mobility.

CSV + GeoJSON → Offline Policy Diff

INDUSTRIAL

Factories & campuses

Machine routes, restricted workspaces and uncertainty-aware separation around people.

Trajectory replay → Assurance report

LOGISTICS

Ports & yards

Right-of-way, emergency access and multi-operator constraints across changing infrastructure.

Scenario adapter → Runtime specification

CRITICAL

Crisis infrastructure

Conservative degradation when sensing, connectivity or institutional coordination becomes unreliable.

Failure injection → Evidence protocol

PHASE 0

Three weeks to the first customer-specific evidence.

A tightly scoped offline engagement: one scenario, one machine, customer-provided trajectories and site geometry. No live control and no claim of legal certification.

W1Validate data

Coordinates, timestamps, identities and assumptions.

W2Compile & replay

Spatial assertions and governed comparison.

W3Deliver evidence

Policy Diff, incidents, delays and open questions.

05 / PUBLIC RESEARCH

Public inquiry
shapes the boundary.

2025—2026

From crisis infrastructure to audit geometry.

ÉTÈ convenes artists, researchers and the public to test what machines must not override in shared space.

2025 programme ↗2026 programme ↗

With contributors from Community AI, Corporeal Architecture, eHTC, Social Sensibility, LAS Art Foundation, Biānjiè Research Group, XAIA and independent practices.

ÉTÈ composition and evaluation diagram — 01 · COMPOSE THE FIELDRULE → GEOMETRY

Governance as relational composition — 02 · RELATE ACTORS & PRESSURESSTATE → RELATION

Cross-scale world model and safety layer — 03 · AUTHORIZE THE ACTIONPROPOSAL → DECISION

01 / 03 · CLICK IMAGE TO ENLARGE

06 / COMPANY

Built across governance,
geometry and execution.

ÉTÈ is a product company for the embodied-AI economy, built from an independent research foundation. We turn spatial design, AI, public-sector digitalisation and governance into one technical assurance boundary.

Current build-definition discussions include world-model and Active Inference expertise, European governance reviewers, edge-computing and urban testbed pathways. Participation does not imply institutional endorsement.

CURRENT STAGE

Build definition

ÉTÈ is converting its validated architecture into a customer-data PoC and first paid co-pilot.

Engineering + customer evidenceFreeze one scenario, build the adapter and Studio workflow, then validate the runtime interface with real operating data.

CORE & CONTRIBUTING TEAM

Roles describe current contribution areas; external reviewers and institutions are not represented as employees or endorsers.

FOUNDER

Yidi Lola Wang

Product thesis, spatial composition and co-pilot definition.

AI RESEARCH

Sonia Bernac

World-model research, computational philosophy and critique of synthetic common sense.

SPATIAL SYSTEMS

Yuliya Gaydamachenko

Spatial research, gradual system diagrams, rule provenance and contestability.

PUBLIC-SECTOR ANALYSIS

Jaime Bernal Cereceda

European public-sector digitalisation, data governance and institutional implementation analysis.

BUILD-DEFINITION NETWORK

Current technical and governance discussions include Dan Mapes and a nominated Cyberlab engineer, independent governance reviewers, robotics researchers, European edge-computing teams and urban testbed partners. Formal roles, IP and delivery obligations are defined separately.

FOR ROBOTICS COMPANIES, OPERATORS, CITIES & INVESTORS

Define the boundary before the machine crosses it.

partnership@ete.institute ↗