Atlas Nails a Cartwheel-Backflip as Boston Dynamics Pushes Robot Agility
- Student Hub
- 1 day ago
- 2 min read

Boston Dynamics has spent years training audiences to expect spectacle from its robots. The latest footage of its new Atlas humanoid delivers again, but this time with a quieter message about persistence, failure and steady progress in advanced robotics.
Early attempts at a backflip did not go smoothly. On its first try, Atlas lost a hand. Now, the robot links a backflip directly out of a cartwheel, lands cleanly and keeps its balance. The achievement signals how quickly the new, fully electric Atlas is closing the gap with its hydraulic predecessor, Atlas HD, which long set the benchmark for dynamic humanoid movement.
In a recently released video produced with the RAI Institute, Atlas executes the combined manoeuvre without visible damage. The movement looks deliberate rather than theatrical, more like a trained athlete chaining skills than a machine following rigid instructions. That distinction matters for a field chasing robots that can operate reliably in unpredictable environments.
Boston Dynamics does not hide the path to that result. The video includes failed attempts, awkward landings and moments where Atlas simply topples over. Those falls rarely end the exercise. The robot often catches itself or gets back up, reinforcing a central lesson in robotics development: progress depends on absorbing mistakes, not avoiding them.
Running presents a tougher challenge. When engineers push Atlas to increase speed, the robot sometimes crashes face-first, sheds components or collides with pallets placed in its path. Anyone who has taken on a new role too quickly or rushed a promotion without mastering the basics will recognise the trade-off. Speed can expose weaknesses that careful pacing keeps hidden.
Walking, by contrast, has improved markedly. Atlas now moves with a more natural gait, shedding the stiff, tentative steps that once made it look uncomfortable at slow speeds. Until now, that awkwardness carried little consequence. Most commercial applications prioritise stability and task execution over elegance. Still, a natural walk hints at broader readiness, especially for environments designed around human movement.
That readiness is becoming more tangible. Boston Dynamics has launched Atlas in a modified form as an enterprise product. The commercial version lacks some of the sleek design of the research model, but it reflects a familiar shift in technology: refinement gives way to durability once real-world use comes into view. The original research version of Atlas appears to have reached the end of its development cycle.
The stunts shown in the video were described as "one final push to test the limits". That framing suggests closure rather than bravado. Engineers often reach a point where further gains demand disproportionate effort, prompting a pivot from exploration to application.
The broader question lingers. If Atlas can recover from falls, adapt its movement and learn from repeated failure, how soon before such machines operate alongside people in warehouses, construction sites or disaster zones? And if a robot can afford to fail dozens of times in pursuit of a single clean landing, what might that teach organisations still reluctant to tolerate short-term setbacks for long-term capability?
Atlas’s cartwheel-backflip is impressive. Its real significance lies in what it reveals about the slow, disciplined work required to turn spectacular demos into practical tools.
Author: George Nathan Dulnuan





Comments