BlastRadar — Production Risk Score from Git Diff in 10 Seconds
Before every code review or canary release, you need to know: how risky is this change? BlastRadar keeps it simple — paste a Git diff, wait ten seconds, get a risk score. No configuration files, no YAML to write, just pure diff analysis.
For SRE teams and anyone reviewing production code, this removes the “guess by experience” step. A quick glance at the score before deployment gives you confidence.
Meanwhile — Anti-Brainrot Side Pane for Claude Code
When Claude Code is busy thinking, compiling, or testing, your terminal sits idle. Meanwhile puts that downtime to work: it shows daily AI news summaries (no API key required) and spaced-repetition concept cards.
It runs in a side pane alongside Claude Code — popping up when CC is busy, hiding when it’s done. Pure Python, zero external dependencies. Great for anyone who spends a lot of time waiting for Claude Code to finish tasks.
claude-real-video — Let LLMs Actually Watch Videos
Most LLMs “process” a video by reading its filename or description — they never actually see the content. claude-real-video takes a different approach: it uses scene detection to extract key frames, pulls the transcript, and feeds both to the LLM.
Supports local files and URLs. Output works with Claude and other major models. 173 stars, MIT licensed. If your workflow involves LLM analysis of video content (meeting recordings, tutorials), this is far more reliable than just handing the model a filename.