App Marketing by App Masters

There are a handful of genuinely useful observations - the AI micro-drama format running 8+ minutes before product reveal, the 92% remarketing stat, and the advice to start with cheaper generation models - but the bulk of the runtime is a guided product demo, dad joke competition, app-audit filler, and audience shoutouts. The ratio of actionable signal to noise is low.

“they play for sometimes eight minutes before they show you the product”

“don't start with the most expensive model...start with the cheaper models like Minimax H3 or Clink, uh, 4”

The observation that virtually all high-spend drama ads are now AI-generated (tracing the lineage from Korean/Chinese micro-dramas) and the curiosity-gap framing of Rise's 'we owe you an apology' format are genuinely fresh. However, most other points - test creatives, algorithm rewards engagement, don't deceive consumers - are recycled performance-marketing orthodoxy.

“It started with the micro dramas in uh, in China where they could afford to produce in a studio. All of these quick, super emotional, extremely emotional, manipulative dramas for cheap. They made a lot of money. Now people scale this with AI. All of the dramas are AI.”

“algorithmic wise, the view time, it's insane. So what does it tell to the algorithm? It says, look, this is a great content”

Stav is a founder with genuine enterprise clients and real performance-marketing knowledge, and he demonstrates hands-on familiarity with the creative testing stack. However, the entire appearance functions as a product demo/sales pitch for his own tool, which limits how much practitioner-level wisdom he actually shares versus feature-showcasing.

“we have so many performance agencies that work with us”

“I have like 50 slack channels with our enterprise clients that complain and cry about the things that pains them and we just add it extremely fast”

The episode includes some real named examples - Rise mushroom coffee's AI singing drama ads, Cantina Labs as a Sean Parker-backed client, specific model names (Minimax H3, Kling 3/4, Wan 1.3.0), and a rough $2-per-localization cost figure. These are useful anchor points, but many supporting claims remain anecdotal ('millions of downloads,' 'massive amount of dollars') and the examples are instrumentalised to sell MaxFusion rather than to educate neutrally.

“Rise, which is um, this mushroom coffee brand...More than half of the ads they run right now are these singing Pixar style drama ads”

“Cantina Labs, which is a massive Sean Parker company...they are running their stack on MaxFusion and they make a lot of AI ads”

Steve asks a few decent framing questions and occasionally nudges for examples, but he functions primarily as an enthusiastic demo assistant rather than a probing interviewer - repeatedly responding with 'That's insane,' 'Wow,' and 'I love it' without ever challenging a claim, pressing on ethics beyond surface-level acknowledgement, or asking for failure cases. The format is structurally a product walkthrough, not a substantive interview.

“Give me an example. I know you kind of shared with me before we went live”

“Wow. That's so simple.”