LinkedIn Larpmaxxing

This is a blistering take on the performative, low-effort engineering theater that dominates LinkedIn feeds. The writer documents a steady stream of identical computer-vision demos - people waving at cameras, pothole detectors, and other trivial YOLO projects - that are visually similar, practically useless, and rarely connected to any real system. The core argument is that the platform rewards looking interesting over getting better: flashy posts attract applause while criticism is suppressed because comments live on your professional profile. The result is repetitive “LARPing” of productivity where appearance substitutes for skill and measurable improvement.

To prove how trivial these demonstrations are, the writer built a detector to flag those very posts. They scraped 200 screenshots with simple tools, annotated them via Roboflow, exported YOLOv8-formatted labels, and trained a YOLOv8n model (img size 320, 50 epochs) using ultralytics on a modest PC; training took roughly 30 minutes and the whole experiment about 90 minutes. In short, creating a convincing LinkedIn demo is fast and easy, which explains why the feed is clogged with near-identical, low-impact projects dressed up as expertise.