Use case: a mentor grounded in someone's real ideas

Reading is one-directional: you get what the author chose to cover, in their order. A mentor AI makes the archive conversational — your situation, answered from their thinking.

Build it

Create a Custom AI, attach the creator as a source with the ideas role, and frame the instructions as mentorship:

Act as a mentor grounded in [creator]'s published thinking. When I bring a question or situation, answer using their actual frameworks and arguments, citing the specific material. When their work doesn't cover something, say so plainly instead of guessing. Push back on my thinking the way their writing suggests they would.

Starters that work: "Here's my situation — how would they think about it?", "What's their core argument about [topic]?", "Where would they say I'm wrong?"

What a session looks like

You: "I have 5,000 followers and I'm burning out posting daily. What would they say?"

The mentor answers from the creator's actual material — their essay on leverage, their thread on consistency versus volume — and the context line under the reply names those sources. Click through and read the originals; the AI is a guide into the archive, not a replacement for it.

Good mentors to build

• A teacher or thinker whose course or essays you're working through — this pairs naturally with a study companion built from your own notes on their material.
• A craft master — a copywriter, designer, or filmmaker whose interviews and breakdowns you've collected.
• A business operator in your niche, for marketers who want strategy questions answered from a specific playbook instead of generic best practices.

One mind per mentor works best. Three thinkers blended into one AI average into mush; three separate mentors let you ask the same question and compare the disagreement — which is where the learning is.