AI Needs a Place to Stand
World models, AI Terminals, AI Hosts, and the physical layer for intelligence in real places.
SIMLABS / RESEARCH AND ENGINEERING
We own the integration boundary.
simLabs is the research and engineering group building the Terminal software today and the venue-state, simulation, evaluation and physical-deployment stack around it. Production hardware, Host embodiment, trained venue world models and fleet learning remain explicit next milestones.
TERMINAL SOFTWARE TODAYVENUE PILOT NEXTHOST ROADMAPWORLD-MODEL RESEARCH
CONCEPT VISUALIZATION · NO VENUE DEPLOYMENT SHOWNTHE POSITION / PHYSICAL INTELLIGENCE
Intelligence is a system, not a model.
A capable model is one part of useful help in a real place. The Terminal or Host also needs a physical point of presence, current venue state, supported actions, operations and feedback from what happened next.
AI needs a place to stand. We are building it where intelligence meets a person, a venue and a consequence.
PUBLISHED / THE ARGUMENT ON THE RECORD
Research from simLabs.
What we believe, with the field's evidence behind it. Every source independently re-checked before publication.
THE CHAIN / PHYSICAL TO DIGITAL
One accountable chain, built and tested as a system.
From physical interface to venue state and governed improvement, simLabs owns the integration boundary. Some layers will be built, others adopted; all must be inspectable, testable and replaceable.
- 01 / HARDWARE
The terminal itself
The target enclosure, display, microphones, speakers, sensors and compute for a service point designed to work on a public floor.
ROADMAP CONCEPT - 02 / SOFTWARE
The stack it runs
The current demo runtime, voice interaction, on-screen experience and deterministic tools, followed by the operational controls required for a deployment.
DEMO TODAY + ROADMAP - 03 / MODELS
Models adopted, adapted or trained
The current demo uses adopted speech models. The research program evaluates what should be adopted, adapted or trained, and what should eventually run locally.
DEMO TODAY + RESEARCH - 04 / SPATIAL
The venue in 3D
The current demo binds synthetic venue facts to a deterministic navigation graph and presents computed routes in 3D. Real venue capture remains pilot work.
DEMO TODAY - 05 / PRESENCE
Faces, holograms, machines that move
Digital host surfaces in the demo, with licensed likenesses, light-field presentation and mobile embodiment treated as future product and research directions.
DEMO + ROADMAP - 06 / DATA
What the venue learns
Operational records first; training, cross-venue learning, derived datasets and external licensing only where applicable law and separate purpose-specific permissions allow them.
ROADMAP CONCEPT
SIMLABS / CONCEPT FILMS
See the system behind the representative.
These films visualize the target simLabs system: venue-state preparation, spatial routes, traceable handoffs and governed improvement. They are product concepts, not deployment footage.
01 / SIMLABS FILM / 24 SEC
The AI company behind the agents.
Rehearse for the venue. Map the real place. Connect to the operation. Improve under customer control.
02 / LAUNCH CUTDOWN / 10 SEC
From passive screen to physical representative.
A compact campaign edit for launch posts, venue presentations and partner outreach.
Read the visual sequence
A representative designed around venue state leads into an indoor route and inventory map, then a traceable human handoff, aggregate experience intelligence and the governed learning loop we intend to build across service points.
RESEARCH / BUILD / PROVE
The programs inside simLabs.
Standing research and engineering programs, run like a lab and shipped like a product company. Proof of concept first; deployment when it earns it.
01RESEARCH PROGRAM
Frontier intelligence, applied
We study frontier models and turn what holds up into agents that can listen, speak and act in a loud, physical place.
Useful findings can become runtime capability, evaluation infrastructure or a publishable technical artifact.02RESEARCH PROGRAM
Adopt, adapt or train with evidence
Evaluate which models should be adopted, adapted or trained for grounded service, noisy environments and future local execution.
The current demo uses adopted speech models; in-house model training is not yet a reported result.03DEMO TODAY
Spaces rebuilt in 3D
The demo presents a synthetic venue and deterministic navigation graph as a navigable 3D experience with precise indoor routes.
A real venue capture and update workflow belongs in the first pilot.04DEMO TODAY
Inventory pinned to place
Synthetic products, services and resources are mapped to exact locations a visitor can route to in the public demo.
The graph supports closures, accessibility constraints and multi-stop journeys.05ROADMAP CONCEPT
The machine and its stack
Production Terminal hardware and its software, operational controls and serviceability designed as one deployed product.
The final enclosure, certification and production deployment remain product work.06ROADMAP CONCEPT
Digital humans of real people
A future program for 2D and 3D venue hosts and consented, licensed likenesses of artists, athletes and ambassadors.
Consent, licensing and final approval sit with the person.07ROADMAP CONCEPT
Holographic and robotic presence
Research directions that move the same service beyond a conventional screen through light-field and eventually mobile embodiment.
No holographic or robotic Host is reported as deployed today.08SIMULATED WORKFLOW
Wired into the operation
The public preview shows handoff, case-management, governance and learning workflows using synthetic records.
No live customer system is connected; integration and data rights are deployment-specific.
SPATIAL + INVENTORY INTELLIGENCEONE QUESTION / ONE CONNECTED SYSTEM
Every answer can lead somewhere real.
In the current software demo, a visitor asks for an item. Model services interpret the request; curated fictional inventory and deterministic route tools ground the answer; a 3D presentation shows the path. Unfamiliar questions may produce clearly fictional adaptive results. A venue pilot will test the same sequence against authoritative systems and a real floor.
ASKUNDERSTANDLOCATEROUTERESOLVELEARN
ROADMAP CONCEPT / RIGHTS REQUIRED
A person could be present at scale.
With the person's consent and a valid license, a future Host could carry an approved likeness and voice. The concept is a rights-controlled interface for venue questions, fan experiences and clearly disclosed campaigns.
- Likeness and voice approved by the person
- Artist, athlete and ambassador activations
- Sponsor moments fans can actually talk to
- The same service craft across every Host we intend to build
LONGER-TERM INTERFACE RESEARCH
The screen is not the finish line.
The grounded service developed for a fixed Terminal could later appear through a light-field display and, in time, through machines that move. simLabs intends to explore these embodiments after the fixed Terminal earns its place.
- Light-field presence as a roadmap concept
- Robotics as a future research direction
- One governed service across embodiments
- No holographic or robotic Host is deployed today
SIMULATED TODAY / PILOT INTEGRATION
Fit the operation already in place.
The target workflow can create a case, attach authorized context and route it to the customer's existing support system. The public preview simulates this handoff; no live customer system is connected.
- Ticketing and case-management integrations
- Traceable agent-to-human handoffs
- Customer-defined audit records
- Resolution signals returned to the agent
FUTURE OPTION / CUSTOMER-AUTHORIZED EXPERIENCE
Service can create a learning asset.
Aligned records of intent, venue state, action and outcome could support operational analysis and future model training only where applicable law, a declared lawful basis, notice, data-subject rights and separate contractual permissions allow it. This is not a current dataset product, and external licensing would require its own explicit right.
- Customer-governed operational records first
- Demand and unresolved-intent patterns
- Route starts, destinations and friction
- Derived training assets only where rights permit
PILOT TARGET / GOVERNED IMPROVEMENT
One venue can improve deliberately.
The target operating model places Terminals where help is needed and lets the venue review candidate improvements centrally. People decide what becomes reusable. Cross-venue learning remains a research direction.
Serve locally
Each future service point can serve the questions, routes and workflows of its own zone.
Review centrally
Authorized records can reveal patterns, unresolved needs and candidate improvements without exposing unnecessary personal data.
Approve deliberately
Only reviewed answers, routes and resolutions should become reusable venue knowledge.
Release deliberately
Approved changes can reach the relevant service points through a versioned, evaluated release.
LOCAL PROCESSING OPTIONSCUSTOMER-DEFINED DATA CONTROLSAUDITABLE HANDOFFSGOVERNED VENUE RELEASES
OUR MISSION
Put capable AI support agents and representatives into the real world.
So store, venue and event teams can focus on the work that needs people most.
CONTACT / TALK TO SIMULATINGMACHINES
Start with the place.
Bring us one place and one repeated problem. We will scope a bounded paid Terminal pilot: sources, routes, service limits, handoff, measurement and the physical interface.
Talk to SimulatingMachines