How DEV//REALITY Works
Every score, roast, and recommendation is calculated deterministically in application code. AI models never alter your scores.
Deterministic Score Model
Version 1.0.0
When you submit a profile, DEV//REALITY fetches public data via GitHub GraphQL and PageSpeed Insights. Findings are normalized into an evidence array. Categories are weighted according to clear, published criteria:
GitHub Presentation (100 pts)
Portfolio Quality (100 pts)
If an external check fails (e.g. your website blocks automated PageSpeed checks), your score is normalized strictly over the available categories using the formula:
score = (sum(categoryScore) / sum(maxScore of available categories)) * 100Score Interpretations
Outstanding public technical presentation with clear documentation and attention to detail.
Good developer presence with a few areas that would benefit from polishing.
Core projects and profile are present, but presentation weaknesses dilute first impressions.
Noticeable documentation, organization, or portfolio gaps that hinder recruiter discovery.
Public technical presence is sparse, outdated, or incomplete.
Trust & Accuracy Charter
Core Evaluation Philosophy
DEV//REALITY evaluates public developer presentation: documentation quality, repository organization, portfolio quality, website quality and other publicly observable signals.
It does not measure real programming skill, system design ability, interview performance, employment readiness, professional experience, team collaboration, or private code quality.
Scores represent presentation quality, not developer worth. Low public activity does not mean low ability. Stars and commit counts are not proof of skill.