May 30, 2026
Pedal to the petty
It Takes Two Neurons to Ride a Bicycle
Scientists say 2 brain cells can ride a bike, and the comments instantly lost it
TLDR: A quirky paper says a tiny two-neuron system can steer and balance a virtual bicycle, skipping the massive practice older computer approaches needed. Commenters were split between calling it elegant genius and accusing it of being bike logic dressed up as "neurons," with jokes about one neuron, two wheels, and self-driving bikes.
A delightfully weird old-school research paper claimed something almost insultingly simple: a tiny two-neuron setup could keep a virtual bicycle upright and steer it toward a target, without the huge training marathons older computer methods needed. The paper even has that irresistible mad-scientist energy, right down to the cheeky footnote admitting the title might be overselling it because, technically, one neuron might still be enough. Naturally, the community grabbed that and sprinted straight into chaos.
The biggest split in the comments was between "this is charmingly brilliant" and "come on, that’s basically hand-written bike logic wearing a neuron costume." One commenter praised the fun of the paper while joking that using two neurons is practically wasteful if one supercharged neuron could do the job. Another side-eyed the whole thing, saying it felt less like a true learning system and more like someone built a custom bike-riding function and slapped the word "neurons" on top. Still, even skeptics admitted the idea is encouraging: if something this small can handle a messy balancing act, maybe simple systems deserve more respect.
And then came the internet doing what it does best: jokes. One person immediately asked the obvious chaos-gremlin question: so… self-driving bicycles when? Another noticed the bike’s wobble traces looked like a blooming flower, which is honestly the most poetic way possible to describe near-crashing. And yes, someone made the perfect low-hanging-fruit gag: two neurons, two wheels — one for each? Between the skepticism, the nerd flexing, and the meme energy, the comments turned a quirky paper into a miniature drama about whether this is genius minimalism or just clever labeling.
Key Points
- •The article proposes a two-neuron network for controlling a simulated bicycle toward a desired location or path.
- •It contrasts this method with prior work that either required about 1,700 reinforcement-learning practice rides or exact equation-based modeling of a specific bicycle.
- •The paper says the controller naturally prioritizes stability in the short term while remaining accurate for long-range goals.
- •The author describes virtual bicycle riding as counterintuitive and explains that turning right initially requires pushing the handlebars left to induce a right lean.
- •The project uses a general robot physics simulator that models rigid bodies, inertia, applied forces, and hinge constraints before instantiating a bicycle model.