Content teams went from producing 25 posts/week to 250+, but were publishing a lot of generic fluff—phrases like "in today's digital landscape," unsubstantiated claims, surface-level analysis.
Existing tools (Grammarly, Clearscope, etc.) focus on mechanics and SEO keywords, not substantive quality.
What FluffFilter Does:
* Analyzes content against 20+ specialized evaluators
* Automatically detects content type (blog post, case study, email, etc.)
* Returns 3-5 surgical fixes with exact locations and suggested replacements
* Batch processing for reviewing 50+ documents at once
* Rails 8.0 + SQLite: Decided against starting with Postgres + Redis to launch with zero-dependency stack. Deploys in 30 seconds, is cheap to host.
* Solid Queue replaced Redis: One less service to manage, job processing works identically.
* Claude Sonnet 4.5: ChatGPT did pretty well too, but found Claude's analyses felt better.
* Turbo Streams for real-time updates
I analyzed an AI-generated "marketing trends" blog post. It would score well on SEO tools (good keywords, readable), but FluffFilter found: * Zero concrete examples or verifiable data * Vague audience (unclear who it's for) * Ends with "Would you like that?" like a personal email
See the full analysis: https://flufffilter.com/examples/blog-post
More examples: https://flufffilter.com/examples
Try it yourself: 7-day trial with 15 analyses at https://flufffilter.com
Built this nights and weekends. Would love HN's thoughts on:
1. The technical approach
2. Whether the AI feedback quality is genuinely useful vs. generic
3. What other content types would be valuable to evaluate