Lesson 4 of 25
The Law of the Master Persona
Telling the model who to be changes what it notices. A venture capitalist and a logistics manager looking at the same business idea will raise entirely different objections, and asking for both gets you two sets of problems instead of one bland summary. The useful part of a persona is not the voice - it is the priorities and the blind spots that come with the job.
Gems are the persistent version of a persona on Gemini - a saved persona with its own instructions and files that you open like an app, so your mentor or editor does not have to be rebuilt in every chat. For a one-off multi-expert critique the prompt is much as the book has it, with one addition worth making: let the personas ground their claims in Search, so the VC is arguing with real market figures rather than plausible ones.
- Save a recurring expert or mentor as a Gem.
- Let the sceptical persona ground its figures in Search.
- Put must-know background in the Gem's files, not the chat.
- Ask for the critique in a Doc so you can reply to each point.
- Keep factual lookups out of persona chats.
Instead of
What do you think of my business idea? (Context: My idea is a subscription box for rare indoor plants.)
Perplexity version
Critique my business idea as three distinct experts, each caring about things the others would consider someone else's problem. The idea: a monthly subscription box delivering rare and exotic indoor plants to customers in Indian metros - Hyderabad, Bengaluru, Mumbai, Delhi NCR - at around ₹1,200 a month. Persona 1 - sceptical early-stage VC. Before you argue, use Google Search to find what you can about the Indian subscription-box and houseplant markets, what comparable Indian subscription businesses charge and whether any have shut down. Cite what you find. Then attack: cost per box, plant mortality in transit, customer acquisition cost versus lifetime value, churn after month three, real addressable market. End with the two numbers that decide this and what they would have to be. Persona 2 - D2C marketing head. Use Search to see how Indian plant and gardening brands currently position themselves on Instagram, then give me a brand name and positioning that is not already taken, the customer who buys twice, three zero-budget campaigns, and the one message that beats 'just go to a nursery'. Persona 3 - logistics lead. Use Search on two specific things and give me links: the rules on transporting and selling plants across Indian states, including any plant quarantine or phytosanitary requirement, and what courier services will and will not carry live plants. Then give me the three hardest operational problems and a solution for each. Finally, a section where the three disagree by name, ending with the one question all of them want answered first. Put the whole critique in a Google Doc called 'Plant box - expert critique' with each persona as its own heading, so I can comment under each point.
Instead of
Explain what blockchain is.
Perplexity version
Explain blockchain to me through two teachers, then let them argue. Part 1 - the patient professor. You teach introductory technology to students with no computing background. Start from the problem that existed before the solution. Cover what a distributed ledger is, why immutability matters, and how consensus works. Use one analogy and keep it consistent throughout - do not switch metaphors. Define each unavoidable term at first use. End with the three things learners most often get wrong and the correction for each. Where a diagram would help, describe it precisely enough that I can picture it - then generate that diagram as an image. Part 2 - the conference evangelist. Make me care rather than make me knowledgeable. Concrete finance and supply-chain examples. Ground the examples in Google Search so they are real deployments rather than hypotheticals, and give me the links. Part 3 - the professor replies, flagging where the evangelist's enthusiasm outran the evidence. Use Search here too: what blockchain projects have been quietly abandoned, and where would an ordinary database have been the better choice? Cite them. Do not soften this part - it is the point of the exercise. Then put parts 1 to 3 in a Google Doc called 'Blockchain - two explanations' for me to re-read. If I am going to study this properly, tell me exactly what to put in a Gem called 'Tech Professor' so I can come back to this teacher for the next topic without rebuilding the persona.
Instead of
Should I quit my job to start a business?
Perplexity version
Act as a career coach who has guided people through this exact jump and seen it go both ways. Be direct rather than encouraging. My situation: ten years in accounting in India, stable job, a salary my household depends on. ₹10 lakh in savings. No professional food-industry experience - catering is a long-standing passion. A family to support. Do not give me a yes or no, and do not tell me to follow my passion. 1. Ask me five questions I would struggle to answer comfortably, each with a line on what my answer would reveal. 2. Use Google Search for the factual half of this and cite it: what licences and registrations does a catering business actually need in India - FSSAI registration, GST, local municipal permissions, kitchen requirements - and what do they cost and take? What do current sources say about why small food businesses in India fail in their first year? 3. Using that, lay out a 90-day test I could run while still employed: what I would spend, what I would need to register before taking a single paid order, the first job I should take and how I would get it. 4. Define kill criteria - what I would have to see, by when, to decide against this. Put the 90-day plan in a Google Sheet called 'Catering - 90 day test' with columns Week, Action, Cost, What It Proves, Decision Point. Keep the five questions in the chat so I answer them here. Then tell me what to save as a Gem called 'Career Coach' so this persona and my situation are there when I come back next month.