Note: I discuss C.J. Cherryh's Foreigner below. No real plot spoilers, but if you're planning to read it and want to go in completely fresh, bookmark this for after.

There's a Google I/O every year. Developers line up for keynotes, product announcements, the whole ritual. I/O -- input and output -- is the founding concept of computer science, the basic grammar of how information enters and leaves a system.

There are also IO towers on reservoirs. Physical structures. Pipes and valves that manage the literal intake and release of water from storage. Same abbreviation, same underlying logic, completely separate worlds of practice.

Like most people outside the water industry, most of the engineers at Google I/O have never thought about a reservoir tower. Most of the water operators who've inspected one have never attended a web developer conference. But they're both reasoning about input, throughput, and output every day. The concepts rhyme across domains in ways that neither side tends to notice.

This kind of double-meaning is everywhere once you start looking, and it does more than just create puns. It creates membership. It's the logic of the shibboleth.

"Developer" means one thing in a real estate office -- someone who acquires land and builds on it -- and something else entirely in a software shop -- someone building on top of a platform or API. Both meanings share the underlying idea: development is building on top of an existing substrate. But if you use the word wrong in context, you immediately mark yourself as an outsider. You don't belong to that conversation yet.

The duck curve works the same way. If I drop that phrase in front of a utility engineer, it's a recognition signal, an automatic credentialing. She knows I know something about grid management and solar intermittency. If she starts explaining what the duck curve is, she's decided I don't. The term itself is doing social and epistemic work that goes far beyond its technical content.

These are shibboleths. But they point at something more interesting than mere gatekeeping.

C.J. Cherryh's 1994 science fiction novel Foreigner is built around this problem at species scale.

The setup: a human ship falls out of hyperspace at the wrong destination, crashes into contact with an alien civilization -- the atevi -- at roughly steam-engine levels of technology but with their own fully developed culture, society, and psychology. The first contact goes badly. War follows, then a long and uneasy peace.

To manage that peace, the two sides build an institution: a single official human translator, the paidhi, who mediates all contact between civilizations. One channel. One diplomat with full language fluency, living in the atevi world, serving as the only permitted conduit between species. The protocol is designed to prevent misunderstanding from escalating back into conflict.

The protagonist, Bren Cameron, is that translator. Fluent. Credentialed. He knows the words.

And the book is, in part, about everything that words don't carry.

Language learning confronts you with this. When you learn another language seriously -- not phrasebook competence but actual fluency -- you start to realize that each language isn't just a different encoding of the same content. It's a different structure of perception.

The Eskimo-words-for-snow observation has been overcited and partly debunked in the details, but the underlying point holds: cultures accrete vocabulary around what matters to them. Distinctions that one language makes in a single word, another distributes across a whole phrase, because the distinction either mattered more or mattered differently. Some cultures have had number systems that categorize the world in ways that make Western arithmetic look like one option among many. Language shapes what's easy to think, not just what's easy to say.

What Cherryh explores is the layer beneath that. Even after Bren achieves genuine fluency -- not just vocabulary and grammar but the cultural context, the trust, the implicit value hierarchies -- there's still a gap. The atevi have a different psychology around loyalty and association. They experience something that doesn't translate cleanly into the human concepts of friendship or love. Bren can define the term intellectually. He can use it correctly in sentences. He still can't feel it from the inside.

It's only through high-stakes forcing functions -- near-death situations, moments of genuine crisis where the usual protocols break down -- that something gets through that symbolic systems can't carry. Actions speak louder than words, yes, but the book earns that cliche by making you feel what it actually costs to close the gap.

AI is, among other things, the most powerful translation technology in human history. Real-time, across most major language pairs, at a quality that would have seemed like magic twenty years ago. The technical layer of translation -- converting words and phrases and grammatical structures from one encoding to another -- is increasingly solved.

What's left is everything Cherryh was writing about.

The embedded assumptions about the relationship between individual and family and society. The implicit hierarchies of what's polite and what's presumptuous. The specific cultural logic of what you owe someone versus what you choose to give them. The thousand small things -- and I keep thinking about something as mundane as how cutlery is arranged on a table, because it stands for the whole enormous category of practices that are completely invisible until you get them wrong -- that don't live in words at all.

Translation across those registers requires something closer to what Bren Cameron was doing. Not just decoding, but dwelling. Not just knowing the term but having the context it grows from.

The IO tower and the Google I/O session might share a word. But to move fluently between those worlds -- to actually carry meaning back and forth, to let insights from one domain illuminate the other -- takes something more than vocabulary. It takes enough of a shared substrate that the concepts can actually land.

We're building ever more powerful systems for the easy half. The harder half is still mostly patient, slow, human work.

Writing coda: This post began as a voice memo. A bespoke Claude Sauntercast skill cleaned the transcript, preserving the walking cadence while making it readable. Then a second skill, trained on past Pioneering Spirit posts, shaped the cleaned sauntercast into the essay above. Everything is translation, including this.