Small decisions. Interesting possibilities.Submit contentSubmit

Sniff Test

A prose linter that combines local countable rules with Jev-powered judgment rules for writing checks.

Added to Jevfast

How it uses Jev

Sniff Test runs deterministic countable rules locally and, only after consent, sends eligible prose paragraphs with judgment-rule wording to Jev for one probability per rule. It reports flags with the file, line, rule, and confidence for the writer to fix.

What Jev decides

Paragraph writing checkExample answers · not a recorded Jev response · Source ↗
Question 1 · not x but y
YOUR APP
INSTRUCTION

Does the paragraph use a sentence whose purpose is to deny one label and replace it with another label for the same subject?

STATE

One eligible prose paragraph from the draft, evaluated against the checked-in not-X-but-Y rule.

JEV · NOUL
YesNo

App workflow

  1. Load the ruleset

    Read default and project rules, including countable patterns and judgment criteria.

  2. Run local checks

    Evaluate countable rules on the machine and skip fenced code and excluded document structures.

  3. Request consent

    Before judgment checks, disclose what text and rule wording will be sent and wait for approval unless consent is already configured.

  4. Report flags

    Send eligible paragraphs to Jev, apply the configured threshold, and print results without changing the text.

Why it is interesting

It separates locally countable checks from semantic judgments, preserves an explicit consent boundary, and reports findings without rewriting the draft.