Cognitum bets WiFi can replace cameras for sensing — RuView hits 52K stars

Cognitum, a Toronto company founded in 2026, is betting that the WiFi signals already flooding every room can become a privacy-safe human sensor. Its open-source project RuView (also branded WiFi DensePose) reads channel state information from commodity WiFi with ESP32 nodes and reconstructs 17-keypoint human pose, breathing rate, heart rate, and through-wall presence - no cameras, no wearables, no cloud - citing Carnegie Mellon's 2023 DensePose From WiFi research as the scientific foundation.

The bet generated exceptional developer attention: the repo hit GitHub Trending in early March 2026, Runa Capital's ROSS Index measured 44.7K stars with 41.3x growth by March 25, and by May 13, 2026 it stood near 52K stars and 6.9K forks, with Cognitum selling hardware bundles around $140 and releasing ESP32 firmware updates such as v0.6.4-esp32 on May 7.

The credibility gap grew with the star count. Innovatopia reported in May 2026 that engineers on Hacker News and in Cybernews coverage said the repository lacked a working pipeline for extracting CSI from real WiFi hardware - closer to a framework or prototype - and that GitHub issues #185, #231, and #408 documented failed real-hardware reproductions. The developer's own README conceded that camera-free pose accuracy with proxy labels was only about 2.5% PCK@20, that the much-cited 92.9% figure came from a narrow camera-supervised test, and that the supervised evaluation phase was unfinished.

Cognitum kept shipping through the criticism - more hardware targets, edge modules, and later READMEs that retracted an earlier 100% presence claim and added claim-check tooling - while arguing that the underlying CSI-sensing research is real and that camera-free designs avoid GDPR and HIPAA imaging rules in eldercare, retail, and hospitality settings. As of September 5, 2026 the project is still live and maintained, with no funding disclosed.