Brett Adcock is simultaneously running two of the most ambitious AI hardware ventures in the world, and recent developments suggest his dual strategy is paying off. Figure AI, his humanoid robotics company, just completed an unprecedented 81-hour continuous autonomous operation sorting over 101,000 packages without human intervention. Meanwhile, Hark, his newly funded AI device startup, has secured $700 million in Series A funding and is assembling a world-class engineering team to launch consumer AI hardware by summer 2026.

The parallel momentum raises a crucial question: Is Adcock building complementary technologies, or spreading himself too thin? The answer appears to be the former. Both companies are pursuing different markets with different timelines, yet they share a common philosophy about how AI hardware should work.

What Just Happened With Figure AI's 81-Hour Robot Demonstration?

On May 13, 2026, Figure AI started a livestream that was supposed to run for eight hours. It didn't stop until 81 hours later. Three Figure 03 humanoid robots sorted 101,391 packages with zero human intervention, zero teleoperation, and zero failures during the entire run.

The technical achievement centers on a system called Helix-02, Figure AI's second-generation whole-body robot intelligence model. Unlike traditional robotics systems that use separate modules for vision, balance, and arm control, Helix-02 is a unified neural network that handles all these functions simultaneously on the robot itself, with no cloud dependency.

"Our original goal was an 8-hour run. After zero failures yesterday, we decided to keep going. We're now over 24 hours of continuous autonomous operation without a failure. This is uncharted territory," said Brett Adcock.

Brett Adcock, Founder and CEO, Figure AI

The livestream attracted millions of viewers and generated legitimate skepticism. Some observers questioned whether the robots' frequent head gestures indicated hidden teleoperation. Adcock's response was technically specific: the head movements are an emergent behavior from the whole-body controller clearing arm pathways automatically, not a sign of human control. The same gesture appears consistently in identical circumstances, which is the signature of a learned behavior rather than variable human input.

The demonstration included real limitations. When boxes overlapped or arrived misaligned, robots occasionally paused before resuming. These moments matter because they reveal how the system handles uncertainty. Helix-02 includes an automatic recovery mechanism that triggers when the AI encounters situations outside its training data. The robot recognizes it doesn't know what to do, resets, and continues without human intervention.

How Does This 81-Hour Run Change the Robotics Industry?

The endurance benchmark matters more than the raw numbers suggest. For logistics operators evaluating humanoid robots, sustained performance across multiple work shifts is more relevant than peak performance in a controlled ten-minute demo. Figure AI's demonstration shifted the conversation from "can robots do this at all" to "how long can they do it, at what reliability level, and at what cost".

The livestream also served as a market development exercise. Over 10 million views across platforms meant the demonstration reached mainstream audiences including logistics operators, warehouse managers, and supply chain executives who may not follow robotics news closely. Adcock's decision to livestream rather than release curated highlight footage was a deliberate transparency choice. A livestream cannot be edited, so the pauses and resets are visible alongside the impressive sustained performance.

Every one of those 101,391 package interactions generated operational data. Every pause and reset generated edge case data. This operational data flywheel is what makes deployed robots better over time, the same mechanism that powers improvements in other real-world AI systems.

What Is Hark, and Why Did It Raise $700 Million?

Hark is Adcock's second major venture, focused on consumer AI devices rather than industrial robots. The company raised $700 million in Series A funding led by Parkway Venture Capital, with participation from Nvidia, Intel Capital, AMD Ventures, Qualcomm Ventures, and Salesforce Ventures. The funding values Hark at $6 billion.

Hark is developing custom AI models designed to provide "advanced personalized intelligence" on consumer hardware. The models feature a multimodal architecture (meaning they process text, images, voice, and other data types) and persistent memory that saves user preferences. This allows the AI to proactively generate task suggestions based on what it learns about each user.

Unlike many AI companies that run frontier models in the cloud, Hark plans to deploy its models on optimized local hardware. This approach avoids the prohibitive cloud inference costs that would emerge if millions of users accessed advanced models simultaneously. The company is likely to build less advanced, more cost-efficient models that run directly on user devices, similar to Google's approach with Gemma 3n, a multimodal model optimized for devices with limited processing capacity.

A demo video on Hark's website shows the AI models reserving restaurant tables, making e-commerce purchases, and performing research, all through voice commands. The company plans to roll out its first models in summer 2026, with consumer devices launching at an unspecified later date.

How Is Hark Building Its Team to Compete With OpenAI and Google?

Hark is aggressively recruiting engineering talent from major technology companies. The startup has already hired 30 engineers from Google, Meta Platforms, and Amazon, with plans to expand its total engineering staff to 100 people by mid-2026.

The most recent high-profile hire is Abidur Chowdhury, a former Apple designer who introduced the slim iPhone 17 Air during Apple's September 2025 product event. Chowdhury joined Hark as head of design, bringing consumer hardware expertise to the AI device startup.

This hiring strategy reflects Hark's competitive positioning. Adcock invested $100 million of his own capital into the company in December 2025, providing the financial runway to compete with larger technology firms for specialized talent. The company is using this capital to upgrade its AI training infrastructure, which includes an on-premises cluster of Blackwell B200 graphics processors.

Steps to Understanding Adcock's Dual-Company Strategy

• Figure AI Focus: Industrial humanoid robots for warehouse and logistics applications, with proven 81-hour autonomous operation and a $39 billion valuation based on demonstrated technical capability.
• Hark Focus: Consumer AI devices with personalized intelligence models that run locally on hardware, targeting the mainstream market with a $6 billion valuation and summer 2026 launch timeline.
• Complementary Markets: Figure AI addresses the B2B logistics sector where reliability and endurance matter most, while Hark targets consumer and business users who want AI that understands their personal preferences and habits.
• Shared Philosophy: Both companies prioritize on-device intelligence over cloud dependency, reducing latency and costs while improving privacy and reliability.
• Talent Acquisition: Hark's aggressive hiring from Apple, Google, Meta, and Amazon signals that Adcock is building engineering depth to compete with established tech giants in the consumer hardware space.

Adcock's competitive environment is intensifying. Figure AI competes against Tesla's Optimus program, Apptronik's Apollo (which is deploying with Mercedes-Benz), and Chinese rivals including Unitree. Hark faces competition from OpenAI, which is reportedly developing a screenless AI device that could debut in early 2027, along with smart speakers, smart lamps, and wearables.

The timing of these announcements suggests Adcock is signaling confidence to investors and customers. Figure AI's 81-hour demonstration proved that humanoid robots can sustain real-world operation at scale. Hark's $700 million funding round and high-profile design hire demonstrate that consumer AI hardware is attracting serious capital and talent. Together, they position Adcock as a builder of multiple AI hardware categories, not just a robotics entrepreneur.