growth minded SUPERHEROES

A handful of non-trivial observations (CRO still perceived as spaghetti-testing after 20 years; needing the process before AI is the prerequisite for AI value; AI as a rearview mirror of historical best practices) are buried under a long, explicitly unplanned conversation full of tangents, anecdotes, and circular agreement. The ratio of filler to signal is very high.

“you had the process before AI. That's the whole thing. That's what matters.”

“AI is a rearview mirror. It's looking at a history of all the problems that. It's like a compilation of best practices.”

The dominant takes - AI creates Dunning-Kruger false confidence, don't use AI blindly, LinkedIn is enshittifying, people want shortcuts - are among the most recycled opinions in tech media in 2025. The CrossFit/GenAI analogy and the BMW user research social-desirability-bias angle show flashes of fresher thinking but are not developed into actionable frameworks.

“Generative AI is like saying CrossFit is all fitness. It's just the loudest, most cultish kind of bunch in fitness.”

“AI is Elon Musk, and that's donning karaoke.”

All three speakers are niche CRO practitioners and podcasters who have clearly done real work (agency CRO, in-house experimentation at Motive), but none is presented in the transcript as having operated at significant scale or having a credential that would command authority in a broader B2B audience. The episode references higher-caliber guests from other episodes but those people are not present.

“I have a chameleon second place award downstairs.”

“my favorite episode was with Kelly Hopping, who's the CMO at Demandbase. And she's like one of my. I worked with her at Gartner.”

Concrete claims are almost entirely absent: a single 'nearly 2% uplift' on a button colour test, a vague '4x acceleration' for agency workflows, and a factually incorrect ChatGPT launch date. The BMW research anecdote and the experimentation database use-case are described in hand-wavy terms with no numbers, client names, or measurable outcomes offered.

“we just did a button color test. It was, uh, nearly 2% uplift.”

“I could accelerate it by four times.”

The host openly admits to preparing no questions, and the conversation spends roughly ten minutes on a tangent about the word 'hacking' before circling back. There are occasional moments of genuine pushback (Shiva pressing Andrej on whether AI is really solving the research problem vs. better UX questioning), but these are quickly dissolved into mutual agreement rather than productive disagreement.

“This is why I did not prepare any kind of questions or.”

“Andre, I'll put it back on you. If AI is not the most important thing to talk about in 2025, what was the most important thing for you? And you can't use the words AI at all.”