Short-form content engine

An end-to-end short-form content engine, built and run daily.

A content database tracks performance across the niche. Every stage of the pipeline reads from it and writes back, so the system learns what works.

See how it works What’s inside

DB→BRIEF→SCRIPT→HOOKS→RENDER→DELIVERY

How it works

The pipeline as a system diagram

Six stages, in order. Every stage reads from the database and writes its results back.

Feedback loop

System readout

EVERY STAGE WRITES BACK. THE SYSTEM LEARNS.

The system

The machine, running

The intelligence and ideation layers of the pipeline, as they run. Each block below is indexed to the stage it belongs to.

Stage 01 / Intelligence

Tracking performance across the niche

The content database tracks post performance across the niche. The social graph is that same niche seen structurally: tracked creators set against the untracked field around them, each node sized by how connected it is.

Force-directed social graph of the niche. Tracked creators are highlighted, untracked creators are muted, and node size scales with connection count, forming visible hub-and-spoke clusters. — FIG. 01 / Social Graph

Intelligence dashboard combining a full-width creator velocity chart across the top, niche momentum at bottom left, and a top trending list at bottom right. — FIG. 02 / Overview

Cumulative trajectory per tracked creator, over a selectable week or month window.

Creator velocity view: cumulative trajectory plotted as a line per tracked creator over a selectable week or month window, with a leaderboard panel at the right. — FIG. 02.1 / Creator Velocity

The cohort's mean plotted against its median, over a rolling window.

Niche momentum view: a filled mean line plotted against a dashed median line across the tracked cohort, with toggles for the length of the rolling window. — FIG. 02.2 / Niche Momentum

What the niche surfaced this week, ranked, each carrying its source.

Top trending list for the current week: ranked items from across the niche, each with its source and date. — FIG. 02.3 / Top Trending

Stage 02–03 / Ideation & scripting

From what the niche is doing to a script

Ideation triage runs on acceleration rather than accumulated stars. Repositories and MCP servers are triaged by how fast they are moving, and the ones heating up become candidates.

Ideation triage board: repository cards each showing a velocity sparkline and a steady or heating state, sorted across triage, brief-created, rejected and archive tabs. — FIG. 03 / Ideation Triage

The idea queue starts from posts instead. High-engagement posts by other creators in the niche are reverse-engineered into pitches: what the post did, why it worked, and what could be made from it.

Idea queue: scored cards mined from high-engagement posts in the niche, each reverse-engineering its source post into a content pitch with tags and a generate-script action. — FIG. 04 / Idea Queue

A brief is the meta piece: it captures the essence of everything a piece of content is about. From one brief the system generates the script and its hook variants, each tagged by strategy.

Brief detail view. The left column holds the brief's labelled fields — title, thesis, angle, key points, target audience, tone, CTA intent and trigger word. The right column fans out into format cards and a generated script with hooks tagged by strategy type. — FIG. 05 / Brief Detail

Stage 04 / Validation

Hooks are validated before promotion

Hook strategies are compared on retention and engagement, with low-sample rows flagged as such.

Hook type breakdown: horizontal retention bars per hook strategy above a table of watch, completion, engagement and reach columns, with low-sample rows badged. — FIG. 06 / Hook Breakdowns

What’s inside

The components

Six machined modules. Each one reads from the system and writes back to it.

C-01

Content Database

Tracks post performance across the niche. Every stage reads from it and writes results back.

in: performance data → out: niche-level intelligence

C-02

Brief Engine

Develops briefs against the database, with angles validated against real share-rate comps before anything gets scripted.

in: performance data → out: validated brief

C-03

Scripting Engine

Generates the script plus five hook variants per video from the validated brief.

in: validated brief → out: script + five hook variants

C-04

Trial-Reel Validation

Runs hooks as trial reels before promotion. Nothing ships on intuition.

in: hook variants → out: validated hooks

C-05

Dual-Renderer Pipeline

Renders each video through two renderers: a Remotion split-screen with a HyperFrames overlay.

REMOTIONHYPERFRAMES

in: validated script → out: rendered video

C-06

Delivery Automation

DMs artifacts to viewers who comment a trigger word, and writes delivery results back to the database.

in: rendered video → out: delivered artifacts

Operating principles

How the machine decides

Two principles guide content through all stages of the cycle: niche-level intelligence, ideation, production, and publishing.

Principle 01

Data-first

Every decision traces to the performance database. Angles are validated against real share-rate comps before anything gets scripted.

Principle 02

Validate before promoting

Hooks run as trial reels before promotion. Nothing ships on intuition.

Built and run daily.

Every stage writes back to the database, so the system learns what works.

See how it works

INSTAGRAM / X / LINKEDIN / NEWSLETTER