Figure AI Founder Brett Adcock Releases Video Showing Humanoid Robot Autonomously Ascending and Descending 15 Feet
Brett Adcock Demonstrates Figure Robot Autonomously Ascending and Descending 15 Feet
Figure AI founder Brett Adcock released a video showing his humanoid robot autonomously completing a 15-foot ascent and descent.
Adcock emphasized that the focus is on achieving end-to-end system autonomy, with speed set to significantly increase later.
This demonstration continues the recent autonomous progress of the Figure robot in navigating complex terrains such as climbing ladders and stairs, having previously achieved stair walking solely based on onboard camera perception.
Figure employs end-to-end reinforcement learning and neural network control, gradually reducing handwritten code in pursuit of room-level and larger-scale autonomy.
Adcock has previously showcased milestones such as the robot autonomously walking from the manufacturing workshop to headquarters and performing long-duration warehouse operations.
From a market mechanism perspective, the founder directly releases progress videos as a technology showcase, driven by the commercialization competition of humanoid robots, attracting attention and capital towards Figure and similar companies, benefiting early technology leaders while traditional automation solutions face replacement pressure.
Source: Public Information
ABAB AI Insight
Brett Adcock, who previously founded Archer Aviation, has shifted to humanoid robots, establishing Figure AI and rapidly advancing iterations from F.01 to F.03. The team has deleted over 100,000 lines of handwritten C++ code, fully transitioning to neural network end-to-end control, with Adcock frequently releasing autonomous demonstrations to build technological credibility.
In terms of capital pathways, Figure attracts talent and funding through high-profile progress videos, concentrating R&D resources on integrated models of perception, movement, and manipulation, motivated by the desire to seize the commercialization window for general-purpose humanoid robots, with funds flowing from investors and potential customers towards hardware iteration and data collection.
Similar cases can be seen with Tesla Optimus and Boston Dynamics' progressive autonomy demonstrations, as well as early drone companies establishing market expectations through public videos; currently, Figure is transitioning from laboratory demonstrations to industrial and home scenarios.
Structurally, this represents a technological replacement: traditional robots reliant on preset rules are being replaced by end-to-end neural networks, with the mechanism being that once full autonomous closed loops are established, speed and reliability can be continuously amplified through data and computing power, thereby reshaping the cost structure of labor-intensive industries.
ABAB News · Cognitive Laws
First establish end-to-end, and speed will naturally follow
15 feet of autonomy is more valuable than 100 feet of remote control
The true moat for robots is writing one less line of code