Skip to main content
All experiments
005 ● Live

Content Idea Generator

Building an AI system that generates content ideas based on industry trends, SEO data, and audience pain points.

Interactive tool

Format

How it works

The generator holds two fixed framework pools: 15 blog title structures and 8 LinkedIn hook structures. Each framework is a template with two placeholders, one for your topic and one for your pain point. When you submit the form, the tool interpolates your inputs into the selected frameworks to produce ready-to-edit titles and hooks. The current year is also injected where a framework calls for it. Selection is deterministic, not random. A seed number is computed from the first three characters of your topic input by summing their ASCII character codes. The audience type adds a fixed offset to that seed: 0 for SaaS founders, 3 for marketing teams, 6 for sales teams, 9 for agency owners, and 12 for e-commerce brands. The combined seed is used as a starting index into each pool, and the tool picks five consecutive blog titles and three consecutive LinkedIn hooks from that position (wrapping around if needed). This means the same topic and audience always produce the same results, which makes the tool predictable and repeatable. It also means changing your topic slightly can shift the entire selection, since the seed changes with the first three characters. The frameworks themselves are written to be direct and slightly contrarian, avoiding the generic question-based headline patterns that fill most content guides. Formats like "Stop blaming X on budget" or "Why your Y efforts are not fixing X" are designed to signal a point of view, not just describe a topic. The goal is to give you five working hypotheses for a piece of content, not five polished headlines. You still need to decide which angle is accurate and worth arguing.

▸ Beyond the simulation

This is the back-of-the-envelope version.

The real model needs your actual stack: attribution, channel mix, conversion economics, and what is already broken. Twenty minutes will tell us if it is worth scoping.