Green Irony

The Brilliant Stranger

A field guide to LLMs, explained like you run a business.

An illustrated map on a desk tracing a winding path from mountains to a city, beside a brass compass

Most of what people call AI at work is one specific thing: a large language model, or LLM. It's the engine inside the chat assistants and agents your team already uses.

An LLM has a strange shape. It has read more than any person ever could, and it knows nothing about your business. Not your customers, not your pricing, not what happened yesterday. Every conversation, it meets you for the first time.

That's why AI so often gets your business almost right. The model isn't broken. It's a brilliant stranger, and nobody has introduced it properly.

This four-part field guide explains how that stranger works, in plain language, one analogy at a time. Each part starts with a question people ask about AI and ends with what it means for how you set up AI in your business. By the end, you'll know what the industry calls the context layer, and why it matters more than which model you choose.

Four questions, in the order people ask them

  1. An expert reads in a sealed room while an assistant outside slides a single page under the door

    Part 1

    "Why doesn't it remember what I told it?"

    The Expert Behind the Door

    It meets you for the first time, every time. Everything it knows about you is whatever gets slid under the door.

    Read part 1

  2. A man studies a guitar whose headstock fans into hundreds of tuning pegs, with a filing cabinet standing separately nearby

    Part 2

    "Shouldn't it get smarter the more we use it?"

    The Guitar, Not the Filing Cabinet

    The model stopped learning before you ever met it. What changes its answers is what it gets handed each time.

    Read part 2

  3. A man taking a timed exam reaches for a slim, organized notebook instead of a chaotic stack of binders

    Part 3

    "We connected it to our docs. Why is it still wrong?"

    The Open-Book Exam

    A brilliant student still fails with the wrong book. Your documents were written for people, not for a model.

    Read part 3

  4. A chief of staff places two pages on an expert's desk, with a stack of unused documents set aside behind her

    Part 4

    "How does it know what to look at?"

    The Chief of Staff

    Someone has to decide what lands on the desk. That job has a name: the context layer.

    Read part 4

Every LLM starts as a brilliant stranger. The work is introducing it to your business.

This is the work we do.

Start a conversation