Figure’s Humanoid Robots Sort Parcels in a Livestreamed Demonstration

Matthew Berman watches Figure’s package-handling test and estimates its eight-hour throughput. The demonstration used multiple robots and measured a narrow task.

MBMatthew BermanIntervenant · Forward Future
Chaîne
Matthew Berman
Publié
14 mai 2026
Vidéo
0:48
Lecture
1 min
Diapositives
2
Original
anglais
Ouvrir sur YouTube
En faisant défiler, la vidéo se réduit en bas à droite et continue. Les sous-titres suivent votre langue de lecture.
Langue de lectureDisponible en 2 langues
Cet article n’est pas encore disponible en français ; il s’affiche dans l’original en anglais.
RÉSUMÉ

En 30 secondes

In May 2026, Matthew Berman comments on Figure AI’s livestream of F.03 humanoid robots orienting packages on a conveyor. He finds the pace impressive, jokes about a person inside the robot and estimates roughly 12,000 packages over eight hours. The event was a company demonstration of a specific repetitive task, with multiple robots involved; the video does not establish an independently verified sorting record or general human-equivalent performance.

  1. 1A livestreamed sorting taskBerman watches a Figure F.03 handle parcels and orient their labels down. The demonstration concerns a specific conveyor task.
  2. 2A running counterHe notes 3,175 packages and jokes that the robot has a person inside, immediately clarifying that he is joking.
  3. 3An eight-hour estimateBerman estimates about 12,000 packages in eight hours; the short does not provide an independently verified count or accuracy rate.

À retenir

  1. Figure’s demonstration involved orienting packages so their labels faced down on a conveyor.
  2. Berman estimated about 12,000 packages in eight hours; his short does not independently verify the total or accuracy.
  3. Multiple robots participated in the longer livestream, so its duration is not evidence of one robot working continuously for the entire event.
  4. Human-speed comparisons depend on the specific task, setup, accuracy and measurement conditions.
TEXTE INTÉGRAL

Lire le texte remanié, chapitre par chapitre

Transcription seule
01

A humanoid handles packages

Berman watches a live Figure demonstration, remarks on its pace and jokes with viewers. He estimates a total over an eight-hour shift.

F.03 handles a parcel
DIAPOSITIVE 1 / 2F.03 handles a parcelThe demonstration shows the humanoid handling packages at a conveyor.

This Figure F.03 humanoid is handling parcels on a livestream initially planned as an eight-hour demonstration. It turns each package so its label faces down on a conveyor, and I find its pace impressive. Figure described the demonstration as autonomous and said its Helix-02 system controlled the robots; the claim of human speed is task- and setup-specific.

The counter shows 3,175 packages. I joke that there must be someone in the suit who needs a break, then clarify that I do not mean a person is inside the robot. It is entertaining to watch the count rise, even when some motions look awkward.

The package counter during the stream
DIAPOSITIVE 2 / 2The package counter during the streamThe on-screen count records progress in the demo, not independently verified accuracy.

I estimate that the operation could handle about 12,000 packages over eight hours, more than I think I would manage without a break. That is my estimate, not an independently audited count or error rate. [MIT CSAIL’s later account](https://cap.csail.mit.edu/stunt-or-substance-what-figures-robot-marathon-means-mit-csail-graduate-student-wil-norton) says several robots participated as the event continued beyond the original plan; the stream duration does not establish uninterrupted work by one unit. A later observer also noted errors, so the demonstration alone does not establish a general sorting record.

Sur la même chaîne

12 diapositives34:21EN
Matthew Berman· 27 oct. 2026Matthew Berman Examines Anthropic’s IPO Filing, AI Growth and GovernanceMatthew Berman, Sam Altman · Forward FutureThe presenter reviews the S-1’s financial metrics, compute strategy and voting structure while questioning the tension between rapid AI development and safety.AnthropicAI SafetyIA en entreprise+2
0:26EN
Matthew Berman· 25 oct. 2026Sam Altman on Human Values and Problem-Solving in an AI WorldSam Altman · OpenAIIn a short interview excerpt, Altman considers the enjoyment of solving hard problems and the human role in interpreting results.Human-AI CollaborationFuture of WorkAI Strategy
2 diapositives0:36EN
Matthew Berman· 25 oct. 2026Kimi K3.1 and a Rubik’s Cube Solver DemonstrationMatthew Berman · Forward FutureMatthew Berman shows a scrambled virtual cube being solved and gives the Kimi K3.1 test a favorable assessment.ÉvaluationsAgents de codeOpen-weight models
2 diapositives0:21EN
Matthew Berman· 24 oct. 2026An AI-Generated Kimi Benchmark Chart Used Invented ScoresMatthew Berman · Forward FutureMatthew Berman asks for a graph of Kimi K3.1 Pro benchmarks, then discovers that the displayed figures are illustrative placeholders.BenchmarksÉvaluationsGenerative media
6 diapositives25:50EN
Matthew Berman· 24 oct. 2026Kimi K3.1 Pro Tested: Strong 3D Demos, Uneven Results and Competitive PricingMatthew Berman · Forward FutureMatthew Berman tests the open-weight model on coding, 3D rendering and visual explanations, then compares benchmark results and prices.Open-weight modelsBenchmarksAI cost & performance+1
2 diapositives32:20EN
Matthew Berman· 23 oct. 2026Who Is Local AI For? The Costs, Trade-offs and a Hybrid FutureMatthew Berman · Forward FutureMatthew Berman weighs data control and independence against hardware cost, setup complexity and inference speed.Local inferenceOpen-weight modelsAI cost & performance+1