Lenny Rachitsky's newsletter just walked operators through cloning themselves with Gemini Omni in under 15 minutes. Voice, face, knowledge base, response patterns. Fifteen minutes.
The cost of not paying attention to this is steep. The average consultant charges $250-$500 an hour. A clone that handles even 20% of inbound questions at 2 AM in Singapore is a five-figure monthly asset. A clone built inside someone else's product, trained on your IP, with their terms of service governing the output is a five-figure monthly liability you're paying to build.
So the real question isn't whether you'll be cloned. The tools are here, they're cheap, and your competitors are already doing it. The question is whether the clone runs on infrastructure you own — your knowledge graph, your MCP server, your billing — or whether it sits inside a consumer product where the platform sets the rules, takes the margin, and owns the relationship with the customer who used to be yours.
I'm Matt Cretzman, and this is the exact fork in the road my book On Whose Terms names. Lenny's tutorial just made it consumer-grade. Let me walk you through what most people are about to get wrong.
The Three Extraction Economies, Now With a 15-Minute Onboarding
There are three ways expert knowledge gets extracted right now.
One: LLMs ingested it without consent. The UK copyright fight made this plain. 88% of creators wanted protection. The government punted. Your blog posts, your podcast transcripts, your book chapters — already in the training set. That ship sailed.
Two: Mercor and similar platforms pay experts hourly to dump IP directly into training pipelines. Jacob selling his birthright for stew. The expert trades a one-time hourly rate for permanent loss of the compounding asset. The platform turns that knowledge into a product the expert will never own a share of.
Three: Expert-owned extraction. The expert owns the knowledge graph, the agent, the distribution, the revenue, and the customer relationship. The infrastructure to do this didn't exist at consumer scale until 2024.
The Gemini Omni clone tutorial is extraction economy number two with a friendlier UI. It feels like number three because you're the one clicking the buttons. You're not. You're building an asset inside someone else's walls, governed by someone else's terms, monetized on someone else's roadmap.
This is the moment the book was written for.
Why a 15-Minute Clone Is a Trap if You Don't Own the Stack
Let me be specific about what you give up when your clone lives inside a consumer AI product.
You don't own the memory. Every conversation your clone has is data the platform keeps. You see a transcript if you're lucky. They see patterns across millions of users. Guess who builds the next product off that.
You don't own the distribution. Your clone is discoverable inside their ecosystem on their terms. The day they change the algorithm, change the pricing, or launch a competing first-party agent, your asset evaporates.
You don't own the billing relationship. If your clone charges $50 per consultation, the platform takes a cut. More importantly, the customer's credit card is on file with them, not you. You don't have an email list. You have a permission slip.
You don't own the knowledge graph. This is the one that matters most. Your clone is only as good as the structured knowledge behind it. If that knowledge lives in their format, in their database, you can't move it. The day you want to leave, you start over.
The alternative is what I've been building at Skill Refinery for the past two years.
The KDS + MCP Stack: Your Clone, Your Terms
A Knowledge Delivery System is the infrastructure layer underneath any expert-owned clone. It has four parts and they all have to be yours.
1. The Knowledge Graph. Structured, versioned, queryable. Not a folder of PDFs. Your frameworks, decisions, edge cases, case studies, and reasoning patterns organized so an agent can retrieve them in context. This is the asset. Everything else is interface.
2. The Agent Layer. The clone itself. Voice, response style, escalation rules, refusal patterns. Trained on your graph, not on a generic foundation model's guess at who you are. The agent is replaceable. The graph is not.
3. The MCP Server. Model Context Protocol launched from Anthropic in November 2024 and got donated to the Linux Foundation's Agentic AI Foundation in December 2025. That second move matters. MCP is now neutral infrastructure, not a one-vendor bet. Your expertise exposed as an MCP server means any agent — Claude, Gemini, ChatGPT, the next thing — can access your knowledge on your terms, with your rate limits, your pricing, your authentication.
4. The Distribution and Billing Layer. Your customer list. Your Stripe account. Your domain. Your email sequences. The clone is a product. Products need a business around them.
When you have all four, the 15-minute Gemini clone becomes a front door — not a business model. You can ship a Gemini Omni version, a ChatGPT version, a Claude version, and a version that lives on your own site, and all of them call back to the same knowledge graph and the same MCP server. You own the spine. They rent the limbs.
The Numbers Nobody Is Running
Here's the math operators aren't doing before they click "Create Clone."
A mid-tier expert with 10,000 newsletter subscribers and a $300/hour consulting rate gets roughly 40-80 inbound questions a week they don't have time to answer. Call it 60. At a 30% close rate on a $50 async consultation, that's $900 a week left on the table. $46,800 a year.
A clone built inside a consumer product, assuming the platform takes 30% and you have no email capture, nets maybe $25,000 of that. And you don't own the customer.
A clone built on your own KDS with an MCP server nets $46,800, builds your list, and creates a knowledge graph that compounds in value every quarter you feed it. Three years in, that same graph is powering an enterprise licensing deal, a course, and a co-pilot embedded in a partner's product.
Same 15 minutes of setup time at the front. Completely different asset at the back.
This is what compounds. This is what gets stewarded. The knowledge you've spent a decade building is a gift — and how you house it determines whether it becomes a legacy or a line item in someone else's quarterly report.
What to Do This Week
If Lenny's tutorial got you excited, good. The instinct is right. The execution is where most experts will give away the store.
Before you clone, audit. What are the 20 questions you answer most? What are the 10 frameworks you teach? What are the 5 case studies you reference constantly? That's the seed of your knowledge graph. Get it out of your head and into structured format first.
Pick infrastructure you can move. If the tool doesn't let you export your knowledge graph in a portable format, it's not infrastructure. It's a hostage situation with good UX.
Stand up an MCP server, even a simple one. This is the protocol layer that makes your expertise portable across every agent platform that exists now and every one that ships next year. Treat it like a domain name. You wouldn't run your business on a subdomain of someone else's website. Don't run your expertise on a subroute of someone else's AI.
Then build the clones. Plural. Ship a Gemini Omni front door. Ship a ChatGPT version. Ship a widget on your own site. They all call home to your graph. You own home.
The Fork
The extraction is happening. That part isn't optional anymore. The LLMs already have your old work. The Mercor-style platforms are bidding for your next decade. The consumer cloning tools just made it possible for any operator to set up extraction economy number two in 15 minutes and feel like they're winning.
The fork is whether you're the expert who owns the cards, the revenue, and the relationship — or the expert who rented their face to a product roadmap they don't control.
I built Skill Refinery as the infrastructure layer for experts who want to be on the right side of that fork. KDS, MCP, agent stack, billing, distribution. The full spine. The clones plug in on top.
I'm writing a book about this — On Whose Terms: The New Expert Economy and the Fight for What You Know. If the thesis resonates, join the launch list.
The tools to clone you in 15 minutes are here. The question is whose name is on the lease.
Keep Building,
— Matt