The hard part of edge compute isn't the compute. It's explaining the model to developers who grew up on servers.
When I tell people what Halcyon Edge does, the technical questions are always the same: which sandbox, which language, what's the cold start. Those are the easy questions.
The hard question is the one that doesn't get asked out loud: "if my function lives in 18 places at once, where does my data live?"
The honest answer is "wherever you put it" — but for a developer who learned the craft on a single Node process talking to a single Postgres, that answer is a small earthquake. It implies a mental model that needs to be taught, not documented.
So most of the engineering I do these days isn't engineering. It's diagrams, naming, the order of the keys in a config file. The compute is the easy part.