jev-writer
Analytics engine that rates published posts on pre-registered rubrics with Jev and tests ratings against real engagement.
Overview
jev-writer finds out which qualities of your writing actually predict engagement. It rates published posts against pre-registered writing rubrics using calibrated probabilities from Jev, then tests those ratings against real outcomes such as impressions, reach rate and conversion rate. It analyses posts and does not write them.
Unlike writing assistants that flatter drafts, it enforces statistical safeguards: an observational power gate, controls for date confounds and Benjamini-Hochberg false-discovery corrections, and it refuses to state findings when sample sizes cannot support them, reporting results in plain English. It works with LinkedIn exports today and with any platform for which you write a rubric pack, and accepts Vercel AI Gateway, OpenRouter or TypeSafe keys. It can be installed as an agent skill that walks you through exporting data, getting a key, running the pipeline and reading results without overstating them, or driven directly through a CLI with doctor and run commands. It has CI and tests.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Score | How well does this post meet each pre-registered writing rubric criterion?(inferred) | — |
How it works
- Export published posts and engagement data
- Rate each post against rubric criteria with Jev
- Test ratings against engagement with power, confound and FDR controls
- Report findings in plain English or decline when underpowered
Strengths
- Pre-registered rubrics reduce fishing
- Refuses to overstate underpowered results
- Agent skill guides the whole workflow
Caveats
- Observational analysis, not causal
- LinkedIn is the main supported export today
Getting started
npx skills add Kaos599/jev-writer --skill jev-writer, or `npx github:Kaos599/jev-writer run ./exports`.At a glance
- Type
- Project
- Status
- beta
- Language
- TypeScript
- Runtime
- Node CLI / agent skill
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Unlike generative writing assistants that flatter drafts, jev-writer enforces strict statistical safeguards: an observational power gate, date-confound controls, and Benjamini-Hochberg false-discovery corrections. It reports what it finds in plain English and refuses to state findings when sample sizes cannot support them.
logicrw/awesome-jev-projects
Frequently asked questions
Does it write content?
No, it only analyses published posts.
What statistical safeguards does it use?
A power gate, date-confound controls and Benjamini-Hochberg false-discovery correction.
Which platforms are supported?
LinkedIn exports today, plus any platform you write a rubric pack for.
Related projects
Community tool that fetches a SaaS homepage as Markdown and asks Jev ten questions about first-visit clarity, with a leaderboard.
ServicebetaTypeScriptChrome extension that grades readable page sections on ten rubric scores with Jev via Vercel AI Gateway and gives the page an A-E grade.
ProjectbetaTypeScriptDocument evaluation CLI and CI gate using Jev or OpenAI Decisions, with pre-registered validation and documented reliability limits.
CLIbetaPythonRussian-language Letta agent skill that uses Jev as a fast external judge, with rules for treating mid-range probabilities as 'unsure'.
ProjectprototypePythonAudit of 300 web pages for AI citability with nine Jev decisions each, checked against an LLM judge for agreement, cost and latency.
ProjectresearchPythonWriting checker running separate named Jev checks such as hedging, filler, tone and passive voice, each with its own verdict.
DemobetaTypeScript