How Debloat AI grows

From doomscrolling looksmaxxing content: under bloated-to-not-bloated transformation videos, ~100 people ask 'How do I debloat?'. The comment section is a strong indicator of how real a problem is; the only existing solutions were products like pills people don't trust. Built an app for it.

Outsource editing as early as possible (low-leverage task) but keep scripts in-house: outsourced script writers weren't good. Hire one editor per whole video, judged on making it entertaining rather than on captions or motion graphics; you send the script and idea. Upwork gave few gems; better sources are a Twitter post asking for editor referrals, friends' editors, and Discord servers where people seek gigs and will do a free sample video.

First proper paid campaign: make a bunch of TikTok ad videos and scale the ones doing well, for an app concept no one else had built yet.

Onboarding asks the user to take a face picture before the paywall (sunk cost), then shows a slightly exaggerated bloat rating ('you're really bloated... you have to fix it now') followed directly by the paywall. Framing is aimed at the user's insecurity; social proof also used.

About eight faceless TikTok accounts posting slideshows, run by hired people paid on a CPM basis and left to figure out views themselves. Found them through a friend running the same strategy for a height app, who passed on a contractor who knew more.