The computer room: a new way to think about AI

Alex Wlchan puts forward a simple metaphor — the computer room — for how LLMs work. A prompt is a letter. The model is a clerk who reads it, flips through its shelves, and writes back. This isn't new knowledge; it's a new way of saying what's been obvious for a while.

The point of the post is that this metaphor is actually useful. You can reason about what the model is doing, where it might go wrong, and what to expect when you change the letter. It turns the black box into something you can point at.

It's a short piece. No hype, no vendor pitches. Just a good idea for a good price.

Why this matters for us: when we explain Obsidian products to la gente, we should use the same kind of plain metaphor — not the usual silicon valley jargon.