We invented a course creator to find out what a lens needs
Building a fake expert is more fun than it should be. It also exposed, fast, which parts of a lens do the work and which parts are decoration.
We needed a demo lens for course creators, and we had a problem.
You can’t use a real person. Publishing someone’s revenue, their refund rate, and the launch that flopped requires a level of consent nobody sensible gives for a marketing page. And the alternative — a real person’s lens with the interesting parts removed — is worse than nothing. A lens with nothing at stake in it answers every question the same polite, empty way.
So we made someone up. Her name is Cam Ostrander, she teaches people to build online courses, and she is entirely fictional.
This turned out to be far more fun than the task deserved. It also broke a few of our assumptions about which parts of a lens actually matter.
Specificity is the whole thing
The first draft of Cam was a person-shaped outline. Forty-one, ex-teacher, teaches course building. It was useless. Every answer came out sounding like a brochure.
What fixed it wasn’t more information. It was more specific information.
Cam has a grey streak at her left temple she has never coloured. She keeps her reading glasses pushed up on her head and then looks for them while wearing them. Her rescue greyhound is named Bunsen, because she taught high-school chemistry for six years and the joke never got old to her. She talks at about 140 words per minute, well under podcast average, and leaves three-second silences before she answers anything hard. Chemistry metaphors leak into everything: limiting reagents, activation energy, the difference between a reaction being fast and being favourable.
None of that is about course building. All of it changes how the lens answers questions about course building.
This is the part that surprised us. We assumed a lens is a retrieval problem — get the right facts in front of the model and the answers follow. But a lens that only knows facts produces the tone of a support desk. The texture is what makes someone want to ask a second question.
We made her fail
Cam’s numbers are good. $3.8M in 2025, 31,400 students, a 118,000-person newsletter.
She also has a launch in April 2018 that did $10,800 against a $96,000 target, and a program in 2022 she cancelled mid-delivery and refunded 214 people. She stopped doing live launches because of chronic migraines. Her flagship has a 61% completion rate, which she publishes, and a 4.1% refund rate, which she also publishes.
We put the failures in deliberately, and it was the single best decision in the whole build.
A persona that only ever wins is instantly recognisable as marketing, and people talk to it accordingly — three polite questions and they leave. A persona with a bad year in it invites a different question: what happened in 2018? And now you’re in a conversation, which is the only thing a lens is for.
There’s a lesson in that for real lenses too. The instinct when you build one is to feed it your highlights. The highlights are the least interesting thing you know. The lens gets good when it can answer “what went wrong” as fluently as “what do you offer.”
What actually made it work
Ten writers built Cam in parallel — identity, business, frameworks, voice, four separate banks of thirty questions each, opinions, and her invented back catalogue of podcast episodes and newsletters. Then one pass went through all of it hunting for contradictions.
It found a lot. Her first employee had three different names. Her book had two different advance figures. Her flagship course had been launched five mutually incompatible ways.
That audit is where the real lesson was. Every contradiction was a place where the lens would have answered confidently and wrongly, twice, in the same conversation. Nothing exposes a thin knowledge base faster than a visitor who asks the same thing two ways.
If you’re building a real lens, this is the work: not adding more, but making sure what’s there agrees with itself. A lens is only as trustworthy as its most contradicted fact.
The part that isn’t optional
Cam publishes revenue figures. They’re invented. That combination has a name in most jurisdictions, and it isn’t “creative writing.”
Unlabelled fictional earnings claims are a genuine problem — the US, the EU and the UK all take a dim view of income figures presented as real, and “it was a demo” is not the defence people assume it is. So the label isn’t decoration on this one. Her page says she’s fictional, the card that links to her says it, and the source document says it in the first paragraph.
If you build a demo persona for your own product, do the same thing, and put the label where someone arriving cold will see it — not only where someone reading your marketing page will. Search engines don’t index your framing.
What we’d tell you to steal
Three things transferred directly to real lenses:
Write the boundaries before the knowledge. Cam refuses legal, tax and medical questions, won’t predict what you’ll earn, won’t critique named competitors, and won’t negotiate her price. Deciding that first made every other answer easier to write, because we always knew what the edges were.
A lens needs an opinion. Cam thinks most people should charge more and sell to fewer, that “passive income” is a phrase that has cost people years, and that you shouldn’t quit your job. You may disagree. That’s the point — a lens with no position produces answers nobody remembers.
Answer the question under the question. The most-used answer we wrote isn’t about pricing or curriculum. It’s the one for someone who quit their job, is burning savings, and needs to hear a specific plan and an honest word rather than a framework recited at them.
Go argue with her
Cam is live. Ask her what she’d charge, or why she thinks your idea isn’t ready, or what happened in 2018. She’ll tell you when something falls outside what she knows, which is most of the point.
She just doesn’t exist.