March 2, 2026

When trout met mouse, chaos ensued

Launch HN: OctaPulse (YC W26) – Robotics and computer vision for fish farming

Robots check your trout; commenters roast the fish cursor and ask if it works in murky water

TLDR: OctaPulse launched robots to inspect farmed fish and is already live with North America’s biggest trout producer. Comments clashed over a gimmicky fish cursor, doubts about murky-water accuracy and data labeling, and praise for a 'for everyone' alternative to pricey fish tech—highlighting big stakes for sustainable seafood.

OctaPulse splashed onto Hacker News with a big claim: robots that inspect farmed fish so farmers don’t have to net, sedate, and hand-measure them. They’re already live with North America’s biggest trout producer and say their waterproof gear, smart cameras, and small on-site computers can count and size fish with less stress. But the comment section instantly hooked on something else: the website’s fish-shaped mouse pointer. One camp chuckled; another cried distraction, with one top quip calling it “cute, but extremely annoying.” The cursor became the unofficial mascot—and punching bag—of launch day.

Then came the real talk. A skeptical chorus asked if the system works when the water’s murky—“Can it see through soup?” vibes—while a home aquaponics grower cheered the potential to make farm-raised fish better quality and more respected. A class-war splash hit too: one zinger framed this as the accessible cousin to high-end fish tech, “Shinkei was for the rich; you guys make it for all.” Behind the scenes, a practical question cut through the hype: who’s doing the tedious data labeling to train this thing? Love the mission or not, the crowd agrees the stakes are huge: most of the world relies on fish protein, the U.S. imports 90% of its seafood, and smarter farms could matter. Dive in at tryoctapulse.com.

Key Points

  • OctaPulse is building a robotics and computer vision platform to automate fish inspection, deployed with North America’s largest trout producer.
  • The founders cite global seafood dependence and poor data visibility in a $350B aquaculture industry as motivation.
  • Manual hatchery sampling is labor-intensive (≈5 minutes per fish), stresses fish, and yields sparse data; OctaPulse aims to automate this.
  • Technical stack includes Luxonis OAK cameras (Myriad X) for on-camera inference and Nvidia Jetsons (Orin Nano/NX) for heavier tasks.
  • Models use CNNs, transformers, YOLO detection, custom segmentation, and keypoints, optimized via TensorRT, OpenVINO, ONNX Runtime with INT8 quantization and careful calibration.

Hottest takes

"The fish cursor is cute, but extremely annoying." — chadash
"Have you had any issues with turbidity so far?" — donalbrecht
"Shinkei was for the rich; you guys make it for all" — dogclaw
Made with <3 by @siedrix and @shesho from CDMX. Powered by Forge&Hive.