The Mirror Test in Practice: When the Machine Wrote His Investor Update
- Jeff Abbott
- Jul 10
- 3 min read
Daniel runs a five-person company out of a co-working space in Phoenix — the kind of founder who does his own investor updates at eleven at night because nobody else is going to. He'd been using an AI assistant for months: meeting notes, draft emails, the occasional landing page. Useful, unremarkable. Then one Tuesday he asked it to draft the October update, and what came back stopped him.
It had his cadence. The short opening line he always uses. The optimism-with-a-caveat structure he thought was a personality, not a pattern. It even reached for the small self-deprecating aside he would have written in exactly that spot — the one his lead investor once said was the only part of these updates she reads twice. He sat there with the feeling many of us are starting to know: it sounds more like me than I do.
Founders in AI Salon rooms from Phoenix to Tokyo have described some version of this moment. Most respond in one of two ways — they get spooked and pull back, or they shrug and ship it. Daniel did the third thing, the one the book's Prelude suggests. He paused and took the Mirror Test: three questions, answered in writing, before touching the draft again.
1. What did I just see?
He kept this factual, which is harder than it sounds.
The model reproduced my sentence rhythm, my sign-off, and my habit of pairing every piece of good news with a caveat. It has learned this from months of my own prompts, drafts, and corrections. It produced the genre — a founder update — flawlessly.
Not magic, not mind-reading. A very good mirror, reflecting what he had been putting into it all along.
2. What assumptions does this reveal?
This is where it got interesting.
It assumes I am my patterns — that the way I've written is the way I'll write. It assumed this quarter ended the way quarters usually end. It assumed the caveat is a style. Sometimes the caveat is the whole message.
The draft was confident about things Daniel wasn't confident about, in his own voice — which is precisely what made it unsettling.
3. What remains uniquely mine?
His answer, lightly edited:
The decision about what not to say yet. The 2 a.m. doubt that came before the optimism. Knowing which investor reads only the first line, and which one calls after every update. The relationship is mine. The stakes are mine. The mirror has no skin in this.
Seven minutes, one notebook page. Then he rewrote the update — keeping perhaps half of the draft, cutting the manufactured confidence, writing the hard paragraph himself.
Here is the part worth underlining: Daniel didn't use AI less after that. He used it more, and better. Once he could see what the mirror does — reflect patterns, reproduce genres, assume yesterday predicts tomorrow — he stopped asking it to be him and started asking it to help him think. The drafts got faster. The judgment stayed home, where it belongs. As we say in the book: AI is a mirror. You get in what you put out.
Daniel is a composite — drawn from real conversations with founders in AI Salon gatherings, with details changed. The moment, we suspect, you will recognize as your own.
Try it yourself
The next time an AI's output feels uncannily you, don't shrug and don't flinch. Take out a pen and answer the three questions above. Then decide — with full knowledge of what the mirror can and cannot hold — what to do with the draft.
Comments