Previously at the Volcano Base I’d been praising constraints as a way of developing momentum in Systems Tai Chi. Since then, I’ve been waiting for a login in order to get a project moving. I still find it amazing that organisations can take weeks to do something an individual can achieve in moments.

In theoryland, an edge case is a scenario that sits at the boundary of what a system expects. A little gremlin perched at the perimeter, clutching a clipboard, waiting for someone to trip over it. These are the situations designers knew might happen but sincerely hoped never would, like a toaster receiving a fax, or a user entering an emoji into a field marked “postcode”.

In practice, "edge case" often means:
“We didn’t think about this and now we really wish you’d stop bringing it up.”

It’s a polite way of saying:
“Our beautiful model of the world has a hole the size of reality in it.”

With a sufficiently philosophical tilt and a slightly malevolent spirit, yes. Everything is an edge case of something. Take breakfast. For most people it involves food. For others it involves intermittent fasting, or a smoothie consisting mostly of regret. These are edge cases too, relative to the supposed “normal” breakfast.

The moment you define a centre, you create a periphery. And the periphery breeds exceptions like fungus in a misplaced lunchbox.

But systems, unlike humans, need fences. They can’t accommodate every possible permutation of reality unless you want them to run at the speed of continental drift. So designers draw boundaries. Inside the fence: the happy path. Outside: the land of edge cases, where dragons and out-of-office replies roam.

Because edge cases expose what systems assume about the world. Automation runs on tidy rules and tidy data. Edge cases are situations where the world turns up in a crumpled raincoat with a suspicious bulge and says, “Surprise.”

AI, even the fancy kind, is essentially a probability machine. Edge cases live in the low-probability wilds. They are the strange corners where patterns fall apart, where the system has only seen three examples, two of which were mislabeled by someone after a Friday pub lunch.

Absolutely. Humans fail at edge cases constantly. The rules of etiquette collapse as soon as someone brings a badger to a dinner party. Nations crumble when an unforeseen variable arrives, such as a giant horse containing warriors, or an orange fascist.

Craftsmanship, policy, romance, public transport schedules: all unravel when faced with situations they weren’t designed to handle. Edge cases are simply where assumptions meet reality and reality wins.

If anything, human history is one long sequence of hilarious and occasionally catastrophic edge cases. The difference is that humans improvise in a way that systems can’t (yet).

Some interesting stuff I discovered on my adventures.

A newly launched 301 km circular walking route across the North Pennines UNESCO Global Geopark, stitched together from uplands, waterfalls, wildlife and historical remnants. Why care: It’s a reminder that edges aren’t just conceptual. Some of them are windy, beautiful, and require sturdy boots.

A look at Google’s latest developments in extended reality features for Android, including adaptations for Galaxy XR hardware and glimpses of immersive futures. Why care: When the edges of the physical interface dissolve, every design assumption gets stress-tested in 3D.

Google Research unveils architectures designed to give AI persistent, updateable long-term memory that learns as it goes, beyond the cramped limits of a context window. Why care: If AI stops forgetting everything it learned five minutes ago, we’ll finally have systems that behave less like goldfish with Wi-Fi.