Live Now: Figure 03 Hits Blistering 2.6-Second Throughput in 8-Hour Unedited Shift

Key Takeaways

Hide

• Figure has initiated a live, 8-hour broadcast of its Figure 03 robots operating completely autonomously, fulfilling CEO Brett Adcock's promise to showcase "human performance levels."
• In the first 10 minutes of the stream, the system successfully oriented 230 packages, translating to roughly 2.6 seconds per item and backing up prior claims of sub-3-second "human parity".
• The unedited format reveals natural imperfections; the robot missed a label orientation on a cardboard box early in the stream, emphasizing real-time recovery over scripted perfection.
• Despite the impressive speed, some industry watchers have expressed disappointment that the stream features the exact same logistics task shown in previous demonstrations, limiting the display of generalized autonomy.

The high-stakes gamble of unedited, real-time robotics is currently playing out on screens across the internet. Following a blunt exchange on social media regarding humanoid endurance, Figure CEO Brett Adcock has delivered on his promise, launching an 8-hour livestream of the company's autonomous operations.

In posts on X preceding the launch, Adcock confirmed the parameters of the test: a full shift operating at human speeds, governed entirely by the company's Helix-02 neural network, with zero human intervention. For a company currently valued at $39 billion, the stakes for this public broadcast are immense.

Blistering Speed, Minor Hiccups

The livestream immediately plunges viewers onto the logistics floor. The visual is a familiar one for anyone following the company: a charcoal-grey Figure 03 unit stationed at a conveyor chute, tasked with identifying incoming packages and flipping them so their shipping labels face downwards.

What is immediately striking is the cadence. Adcock recently claimed the newest iteration of the humanoid had reached "human parity" in average shift throughput, operating at speeds under three seconds per package. The live broadcast provides the receipts. A ticker on the feed showed 230 packages processed within the first 10 minutes of the stream—an average throughput of roughly 2.6 seconds per item.

The packages are highly varied, consisting mostly of deformable, brightly colored poly bags interspersed with rigid cardboard boxes of differing dimensions. This variety highlights the strength of Figure's "Software 2.0" approach, which bypasses hand-coded C++ motion planning in favor of an end-to-end neural network that computes torque directly to control its motors.

However, the "we'll do it live" format also exposes the system's current limitations. It is not flawless. Early in the broadcast, the robot failed to flip a cardboard box correctly, allowing it to pass with the label facing the wrong direction. Yet, for automation veterans, this occasional failure rate is expected; the true test of the Helix-02 architecture lies in its ability to adapt and continue operating without system-halting errors.

The "Same Old Task" Critique

While the endurance and speed are technically remarkable, the early reaction from some online observers has been mixed. The core critique stems from the narrow scope of the demonstration.

The livestream is a direct response to a challenge issued by Dr. Scott Walter, Diligence Director for RoboStrategy, who argued that humanoids would have "limited utility" until they could prove they can handle an 8-hour shift of autonomous labor at human speeds with no intervention. The Figure 03 is undoubtedly succeeding at that specific challenge. However, the task itself is the exact same package-orientation workflow the company showcased nearly a year ago in its 60-minute endurance video.

For skeptics who want to see evidence of truly generalized, multi-purpose autonomy, flipping boxes for eight straight hours—even at a blistering 2.6 seconds a piece—feels like a limited flex.

Regardless of the online murmurs, Figure is currently doing exactly what it set out to do: proving that its hardware can physically survive the grueling repetition of a warehouse shift without human babysitters. As the clock continues to count up toward hour eight, the industry is watching closely to see if the machine's motors, and its AI, can hold the line.