Overview & result
Clarity Judge checks writing against separate named checks with Jev: hedging, em dash overuse, clarity, filler phrases, tone, passive voice and actionability. Each check shows its own verdict, confidence signal and supporting sentence instead of one overall score. The README describes it as an independent community project whose public demo shows simulated results unless you add your own TypeSafe key.
From the original post
Multi-axis writing quality checker powered by TypeSafe AI's Jev model. Separate named checks, each with its own verdict and confidence.
How Jev fits in the loop
- Paste or pick a piece of writing in the editor.
- The app's /api/judge endpoint sends it to TypeSafe with separate named checks such as hedging, filler phrases, passive voice and actionability.
- Each check comes back with its own verdict, a confidence signal and the supporting sentence.
- Without a TypeSafe key the app shows deterministic mock results labelled as demo mode.
How to reproduce
- Inspect the original showcase and source material at https://github.com/TypeSafeAI/clarity-judge
- Verify the bounded prompt and input candidate schema configured for Jev
- Benchmark decision latency and classification accuracy against baseline models
Why this build matters
It keeps every check separate and inspectable instead of collapsing writing quality into one opaque score.
Limitations
- An independent community project, not an official TypeSafe product.
- It evaluates supplied writing only: it does not rewrite text or establish that a passage is factually correct.
- The public demo carries no server key, so its results are simulated and do not show Jev's measured performance.