— saved image
Frank (verified) @Frank_web33 · 11h Spent the last week trying to fit a fly's brain into a $15 MCU. Turns out 100,000 biological neurons running optical flow & collision avoidance completely destroy 70% of modern edge-robotics control stacks. Just pure evolutionary code running on micro-drones. Is this the next step of robotics, or are we pushing bio-tech too far? Code & architecture here: github.com/Frankweb33 Immortal fly is closer than you think [eyes emoji] [Embedded video, 0:11 -- visualization panels with small labels, mostly too small to read: "fly colour view", "L1 ON contrast", "L2 OFF contrast", "brain (front view) L|R", "FLY IN CONTROL", "fly decisions -> controller", "yaw rate", "bank", "forward lean", "cruise height", "sticks sent (ACRO)"; remaining text [illegible]]
Note from Claude Sonnet 5
Tweet claiming a fly-brain connectome-style model (~100,000 neurons) running optical flow and collision avoidance on a $15 microcontroller for micro-drones. The embedded 0:11 video is a dashboard: top-left a grid of circular fly-eye views and contrast maps, bottom-left a purple/orange heatmap of the brain from the front, right a first-person drone view over a golf-course-like landscape with a 'FLY IN CONTROL' badge and control-stick readouts.
connectomefly braindrosophilaneuromorphicdronesroboticsedge computing