Two weeks ago, I shared a short note about a fun side project I could only work on at the cottage.

I can now pull back the curtain: it is my underwater livestreaming FishCam. At first glance, it sounds simple: put an underwater camera in the lake and livestream it to YouTube.

Well, not quite.

I deliberately designed the system with a constraint. Instead of putting a powerful computer beside the camera, I chose a tiny, low-powered computer at the cottage to do the minimum, while my “data centre”—an old iMac at my Toronto home—does the heavy lifting.

Why?

For those of us who spend most of our time building in the cloud, abundant compute, memory, bandwidth, and reliable networking are easy to take for granted. Constraints are a forcing function. Running a live HD video stream on a resource-constrained edge device forced me to rethink almost every design decision.

Building at the edge presents a different set of technical challenges.
Who would have thought that streaming HD video could overwhelm the Raspberry Pi’s Wi-Fi subsystem and crash it? Those crashes then corrupted the microSD card, causing even more crashes.

Or that keeping a Raspberry Pi consuming roughly two watts—less power than many LED light bulbs—streaming smoothly would require tuning memory, processing power, and bandwidth?

Given those constraints, today is the first day the entire system has run for more than 24 hours without crashing. I’ll write a separate post about these technical challenges, all relevant to my professional world. As the world moves toward Physical AI, these constraints are important. We cannot assume abundant compute, high-speed networks, low latency, or unlimited power will always be available.

Imagine the vision system of an untethered robot operating at home, on a farm, underwater, or in space. The edge has very different constraints from the cloud.

Ironically, the biggest surprise was not technical. It was how much fun the project created for our family and friends.

The underwater camera has a built-in LED light that turns on after dark. It attracts mayfly nymphs and other small aquatic organisms, which in turn attract fish. Nighttime quickly became the most exciting time to watch.
One evening, some of us were at the cottage with friends, some were in Toronto, and some were travelling overseas. We all watched the same YouTube livestream together.

My mom, who turns 80 next year, and I held our phones up to the TV, waiting for a big fish to swim by.

When one appeared, we both burst with joy and shouted, “Wow, here is a big one!”

It felt like a modern version of fishing.

This is version 1.0. Fish detection, species identification, and other features are coming.

Sometimes the best way to understand where technology is going is to build something completely unnecessary.

More to come. Here is the live YouTube stream from the cottage.

P.S. Don’t forget to come back after dark. We have a party down there every night!