How Doublespeed grows

Each account lives on a real phone controlled by custom software that swipes, reposts and comments like a person (emulators get caught by device fingerprinting). Account setup: describe the product and target (e.g. 20-44, mostly male), AI generates usernames, names, bios and gender, and each account gets a different set of niche search terms (self-care, self-improvement, lifestyle, motivation) so accounts aren't identical but don't vary too much. Warm-up searches those terms, screenshots what it watches, asks an LLM whether it's relevant, then reposts/comments if so or keeps swiping. Specific warming algorithms are not shared.

Slide 1 is a hook naming the pain ('I used to struggle with tracking my calories'), fronted by an AI creator, a real creator or a faceless slide (only slide 1 is AI-generated). Middle slides give real but high-effort solutions (e.g. weighing meals on a scale); a later slide (e.g. #5) is your app as the cheaper, lower-effort fix. Each slide position has its own image bank: if slide 2 always clarifies the problem, a 50-image bank can run ~300 slideshows.

Find formats by searching your product's terms on TikTok (e.g. StudyTok for a study tool) and picking viral hits from the last 1-2 months. Paste the TikTok link to generate a template, write the template prompt by hand (e.g. 'five Spanish phrases you should know'), then generate variants: AI writes each slide's text, images are randomized from an image bank, hooks and CTA ('Download this app') can be changed in bulk. Always change something versus the original so you have a chance to win. For hook + demo videos, auto-match a library of hook clips with app screen recordings and add copy rather than paying creators. Post results feed back into what gets made next.