Stream 11 opens with a diagnosis: analyzing messages from the ten previous streams, it's clear the biggest pain for people following the project isn't building the vertical agent, it's selling it. With Email Hacker AI hitting R$11,879.51 in measured sales — driven by concrete cases like Gesael (R$2,600 sold in one day using the tool itself) and Enzo Neto, run by Eric, passing R$7,500 — the day's agenda turns to distribution: build a worker that generates content 24 hours a day on its own, without depending on someone sitting there editing video.
The first hack isn't even code: it's choosing the VPS location. A new Hostinger box gets contracted, KVM1, Ubuntu, hosted in Brazil — on purpose, because the algorithm delivers content to the audience where the server is hosted. Next comes the decision to create a new, separate repository (not mixing it with the old video project) and design an eight-agent pipeline running in sequence: Segmenter (breaks the stream transcript into blocks by topic), Analyst (evaluates viral potential), Editor (cuts silence and dead air), Director (sets the visual script for each clip), Render, Capture (word-by-word synced captions), Scout (hashtags), and Composer/Publisher (posts automatically). The idea: the user pastes the channel link, hits play, and the rest is automatic.
The turning point of the stream wasn't technical, it was conceptual. The first version of the Remotion-generated motion design only illustrated what was being said — it turned into a pretty caption. The turn happened when it became clear the right approach was for the motion design to represent what's happening on screen, not what's being said: if the person says "I'm going to build the funnel scene," the top of the frame should show code being written in real time, blocks connecting, the scene being assembled — not an infographic of the sentence. Along with that came a fine-tuning pass on framing (crop centered on the face, no cut-off head, no leftover black corner) solved literally in a loop: screenshot, analyze, adjust, screenshot again, until it's 100% centered.
While the pipeline ran, the stream also handled real bureaucracy: creating an Instagram account from scratch ("email hacker"), converting it to professional, creating a Facebook page, linking the two, and generating a token via Meta Business — all live, complete with a password-not-applied scare and a lost login mid-process. In the end, the worker published, on its own, a real clip from the stream itself to YouTube Shorts and Instagram Reels, with the top video generated entirely by AI illustrating the screen in real time. TikTok stayed pending for the next day.
The stream closed at nearly five hours with the pipeline working end to end, an estimated cost of cents per clip, and a promise to refine curation and scripting on day 12 — the logic already validated, just needing polish.