Tesla's Optimus humanoid robots performing impressive feats in public demonstrations are not operating autonomously,they're being remote-controlled by human operators. This candid admission from Ashok Elluswamy, Tesla's VP of AI Software and current leader of the Optimus program, cuts through months of hype surrounding the project and reveals where the actual engineering challenge lies.

What Does "Remote-Controlled" Actually Mean for Optimus?

When you watch Optimus pick up an object, walk across a room, or demonstrate dexterity in a demo video, a human operator is directing those high-level actions. The robot's onboard systems handle the moment-to-moment physics,balance, joint coordination, and force management,but the robot isn't perceiving its environment, reasoning about what to do next, and acting independently. It's executing instructions while keeping itself upright.

This distinction matters because it separates "following remote commands while maintaining balance" from "acting autonomously in an unstructured environment." That gap is where the genuinely hard AI problems live. According to previous reporting, the Optimus team has adopted a camera-centric architecture that mirrors Tesla's Full Self-Driving (FSD) software approach, using end-to-end foundational neural networks to process camera and sensor inputs and make decisions.

Why Is Physical AI Safety Harder Than Language Model Safety?

Elluswamy made a striking comparison: solving AI safety for physical robots makes the current discourse around large language model (LLM) safety look like "child's play." The difference is fundamental. When an LLM makes a mistake, you get bad text. When a robot makes a mistake, you get force applied to a person or object in the physical world.

Optimus uses custom rotary and linear actuators with torque sensing, engineered to detect unexpected resistance,contact with an object or a person,and reduce output force accordingly. But the critical unknown is the latency between contact and force reduction. In a factory or home environment, that latency window is where injuries happen.

The technical challenges Elluswamy has previously flagged include the "curse of dimensionality" in building foundational robotics models. The system must process enormous inputs,reportedly over 2 billion tokens in 30 seconds,and compress them into precise, low-latency physical outputs like joint torques. For comparison, Tesla's end-to-end FSD system runs its control loop at 36 times per second. A humanoid robot operating in close proximity to people needs comparable or faster response times, with far higher stakes for each decision cycle.

How Is Tesla Preparing Optimus for Consumer Deployment?

While the robot remains in remote-controlled development stages, Tesla is quietly laying groundwork for eventual consumer integration. A decompilation of Tesla app version 4.61.0 reveals Optimus-specific charger identifiers and an Optimus entry in the app's energy-site model, suggesting preparation for managing robot-related hardware through the existing Tesla app that owners use for their cars and home energy equipment.

The analysis identified three charger-related identifiers: optimus_charger_id, createBaseChargerId_OptimusChargerId, and ChargerId_OptimusChargerId. It also identified SITE_TYPE_OPTIMUS within the app's energy model. These findings suggest the robot could be represented within the existing Tesla app rather than requiring an entirely separate application.

However, these software references do not confirm autonomous docking, a consumer interface, or a launch date. A dedicated charger identifier does not reveal the charging hardware's physical design, electrical specifications, or how the robot reaches it. The software findings follow reports of new Optimus supplier audits in China and construction progress at Tesla's dedicated Texas robot factory.

Steps to Understanding Optimus Development Timeline

• Current Stage: Optimus robots in public demonstrations are remote-controlled by human operators, with onboard systems handling only moment-to-moment physics and balance.
• Safety Challenge: The team must solve latency problems between physical contact detection and force reduction, a problem that doesn't exist in language models but is critical for human safety.
• Software Preparation: Tesla is integrating Optimus into its existing app ecosystem through charger identifiers and site-type entries, preparing infrastructure for eventual consumer ownership.
• Manufacturing Ramp: Production lines are being installed at Fremont, and dedicated factory construction is underway in Texas, with initial robots intended for training-data collection and further development.

Elluswamy took over leadership of the Optimus program in June 2025. In July 2026, he stated publicly that Optimus "will not disappoint," signaling internal confidence in the program's trajectory. The current remote-control architecture isn't a failure,it's a development stage.

Tesla's FSD followed a similar path: early versions relied heavily on human intervention, with autonomy expanding incrementally as the system accumulated data and the safety envelope was validated. The difference is that FSD's safety envelope could be tested at scale on public roads with a driver ready to intervene. Validating a humanoid robot's safety envelope in uncontrolled environments,homes, warehouses, public spaces,is a harder problem to instrument and a harder failure mode to recover from.

Interpretability remains an open problem Elluswamy has flagged in end-to-end systems. When a neural network makes a driving decision, you can audit the outcome after the fact. When a robot applies unexpected force to a person, the post-hoc audit is considerably more consequential. The field doesn't yet have reliable tools to understand why a physical AI system made a specific motor decision in a specific moment.

Elluswamy's public candor about the remote-control reality sets a more credible baseline than polished demo reels alone. It also signals that the team knows exactly what the hard problem is: whether the end-to-end neural architecture that solved highway driving can be adapted to solve physical autonomy in human environments. That's the central question Optimus will spend the next several years answering.