Live work · Content Pipeline
Content Pipeline.
One blog post in. Nine platform-native assets out.
Tone analysis, per-platform format constraints, and accuracy checks before each output ships. Run on every long-form post I write.
Demo · Try it
See it in action.
Pick a source post to repurpose
Pick a post above to start the pipeline
I · About
Repurposing one piece of writing across nine platforms used to take me three hours. Now it takes 90 seconds. The pipeline reads my post, extracts the core argument, the data points, and the quotable phrases, then generates platform-native versions for Twitter, LinkedIn, Instagram, email, scroll hooks, thread openers, newsletter blurbs, quote cards, and CTA variants. Each output passes a tone-and-accuracy check before it lands.
II · How it works
The pipeline.
- 01Input: a blog post URL or raw markdown
- 02Step 1: Claude analyzes tone, key points, audience, and quotable phrases
- 03Step 2: Nine parallel generation calls, each constrained by the target platform's format (character limits, hashtag conventions, hook style)
- 04Step 3: Each output is scored for tone-match and accuracy against the source
- 05Step 4: Outputs that fail the score are regenerated with feedback in the prompt
- 06Output: structured JSON of nine ready-to-publish assets
III · Sample
Sample run output (excerpt).
| Platform | Output | Length |
|---|---|---|
| Most teams test their AI features by vibes. The ones who ship reliably write the verifier suites first. | 112 chars | |
| Three years ago, evals were a 'nice to have.' Today they're table stakes. Here's why every AI product needs an eval flywheel... | 1240 chars | |
| Email subject | What 'looks right' isn't 'works' | 31 chars |
| Scroll hook | I tested every coding agent on the same task. Here's what failed. | 61 chars |