Methodology
How we compute RadarScores — transparent, automated, and opinion-free.
How It Works
Crawl
Every Monday at 6AM UTC, our automated pipeline crawls 6 data sources for each tool, collecting fresh signals about adoption, sentiment, and activity.
Score
Raw signals are normalized into 5 sub-dimensions (0-100 each), then combined with equal weights into a single RadarScore.
Compare
Week-over-week deltas track momentum. AI-generated comparison blurbs help you understand relative strengths between tools.
Data Sources
GitHub
Stars, forks, commit frequency, release cadence. Measures open-source health and developer adoption.
Post count, comment volume, upvote-ratio sentiment across AI subreddits. Captures grassroots developer opinion.
Hacker News
Front page appearances, total points, comment engagement. Reflects technical community interest.
ArXiv
Academic paper mentions and recent publications. Tracks research-level relevance and innovation.
G2
Enterprise reviews, star ratings, satisfaction scores. Represents real-world business adoption.
Changelog
Release frequency, days since last release, version tracking. Measures active development momentum.
Scoring Dimensions
Adoption Momentum
20%Growth velocity — how fast a tool is gaining users. Driven by GitHub stars growth and G2 review growth.
Developer Sentiment
20%Community opinion — what developers actually think. Aggregated from Reddit upvote ratios, HN sentiment, and G2 satisfaction.
Enterprise Readiness
20%Business viability — can enterprises adopt this? Based on G2 rating, review volume, and project maturity signals.
Recency
20%Active development — is the project alive? Measured by days since last release, release cadence, and commit frequency.
Buzz
20%Mindshare — how much is the community talking about it? HN front page hits, Reddit post volume, and ArXiv papers.
Update Cadence
Top 50 Tools
Updated every week (Monday 6AM UTC)
Remaining Tools
Updated monthly (every 4th Monday)
Formula
RadarScore = (Adoption × 0.2) + (Sentiment × 0.2) + (Enterprise × 0.2) + (Recency × 0.2) + (Buzz × 0.2)When a sub-score has no available data, its weight is redistributed proportionally across the available dimensions. All sub-scores are clamped to 0-100.